Fisher College of Business
Working Paper Series
Charles A. Dice Center
for Research in Financial Economics
Financial Policies, Investment, and the
Financial Crisis: Impaired Credit Channel
or Diminished Demand for Capital?
Kathleen M. Kahle, Eller College of Management, University of Arizona
René M. Stulz, Department of Finance, The Ohio State University, NBER, and
ECGI
Dice Center WP 2011-3
Fisher College of Business WP 2011-03-003
This paper can be downloaded without charge from:
http://www.ssrn.com/abstract=1754660
An index to the working paper in the Fisher College of
Business Working Paper Series is located at:
http://www.ssrn.com/link/Fisher-College-of-Business.html
February 2011
fisher.osu.edu
Financial policies, investment, and the financial crisis: Impaired credit
channel or diminished demand for capital?
Kathleen M. Kahle and René M. Stulz*
February 2011
* Respectively, associate professor, Eller College of Management, University of Arizona, and Everett D.
Reese Chair of Banking and Monetary Economics, Fisher College of Business, Ohio State University,
NBER, and ECGI. We thank Christian Leuz, Cathy Schrand, Berk Sensoy, and Mike Weisbach for useful
discussions; Viral Acharya, Rudi Fahlenbrach, Jarrad Harford, Jean Helwege, Vince Intintoli, Francisco
Perez-Gonzalez, Robert Prilmeier, David Scharfstein, Andrei Shleifer, and participants at the Fall 2010
Corporate Finance meeting of the NBER, the AFE meetings in Denver, and at seminars at Harvard
University, Ohio State, the University of Amsterdam, and the University of Tilburg for useful comments;
and Matt Wynter and Jia Chen for excellent research assistance. We are especially indebted to Harry
DeAngelo for discussions and detailed comments.
Financial policies, investment, and the financial crisis: Impaired credit
channel or diminished demand for capital?
Abstract
Though much of the narrative of the financial crisis has focused on the impact of a bank credit supply
shock, we show that such a shock cannot explain important features of the financial and investment
policies of industrial firms. These features are consistent with a dominant role for the increase in risk and
the reduction in demand for goods that occurred during the crisis. The net equity issuance of small firms
and unrated firms is abnormally low throughout the crisis, whereas an impaired credit supply by itself
would have encouraged these firms to increase their net equity issuance. After September 2008, firms
increase their cash holdings rather than use them to mitigate the impact of the credit supply shock. Firms
that are more bank-dependent before the crisis do not reduce their capital expenditures more than other
firms during the crisis. Finally, the evidence is strongly supportive of theories that emphasize the
importance of collateral and corporate net worth in financing and investment policies, as firms with
stronger balance sheets reduce capital expenditures less after September 2008.
The conventional view of the financial crisis is that bank losses from toxic assets led to fire sales by
banks and a bank credit contraction (see Brunnermeier (2009) and Shleifer and Vishny (2010)). These
toxic assets were mostly securities backed by subprime and related mortgages, so their loss in value had
little to do with the performance of industrial firms, making the credit contraction an exogenous event for
these firms. Research in finance, including research on the recent financial crisis, shows that exogenous
credit contractions have real effects on firms by forcing them to reduce investment.1 However, an
exogenous contraction in the supply of credit was not the only adverse development in the recent crisis.
Credit contracted endogenously as well because of a demand shock that led to lower cash flows, loss of
investment opportunities, and weaker balance sheets. Further, risk increased sharply in the fall of 2008,
making firms less credit worthy and leading to a flight to quality that increased risk premia. In this paper,
we assess the relative economic importance of the direct impact of the exogenous bank credit contraction,
the decrease in demand and thus expected cash flows, and the increase in risk in explaining the financial
and investment policies of firms during the financial crisis. Most of the evidence is inconsistent with the
view that the direct impact of the bank credit supply shock on firms was the dominant factor. Contrary to
the predictions of the exogenous credit supply shock hypothesis, net debt issuance does not fall during the
first year of the crisis and firms most likely to be bank dependent decrease their net equity issuance rather
than increase it. After September 2008, net debt issuance drops sharply, but firms hoard cash, and firms
more likely to be bank dependent do not decrease capital expenditures more than other firms. Overall, our
evidence shows that the demand shock, the resulting endogenous credit contraction, and the reaction of
firms to the increase in risk play a dominant role in explaining firms’ financial and investment policies.
There is much evidence that financial policies depend on both financial market conditions and
macroeconomic conditions (see references in Baker (2009), Erel, Julio, Kim, and Weisbach (2010), and
Korajczyk and Levi (2003)). Based on this evidence, we would expect firms to exploit advantageous
1
References to this large literature include work focused on the impact of monetary policy (e.g., Gertler and
Gilchrist (1994)), Kashyap, Stein, and Wilcox (1993), and Kashyap, Lamont, and Stein (1994)) as well as more
recent work focused on specific events associated with changes in the supply of credit (Leary (2009) and Lemmon
and Roberts (2010)). Papers on the financial crisis are discussed later in this introduction.
1
conditions in credit markets and to increase their leverage when such conditions obtain. If the availability
of credit becomes temporarily restricted, firms should borrow less. Further, with suddenly less welcoming
credit markets, firms would use their cash holdings to finance investment since precautionary holdings of
cash exist precisely to mitigate the impact of adverse shocks (see, for instance, Opler, Pinkowitz, Stulz,
and Williamson (1999)). Finally, to the extent that a credit contraction is of an unprecedented magnitude,
we would expect an unprecedented decrease in net debt issuance. As credit becomes harder to obtain,
firms should reduce their dividend payouts, repurchase less stock, and issue more equity (see Leary
(2009) for evidence of the latter). In summary, we expect a credit supply shock to be accompanied by a
drop in net debt issuance, by an increase in net equity issuance, by a drop in dividends and share
repurchases, and by a decrease in cash holdings. Further, the credit supply shock should lead to a greater
decrease in investment for the firms that are most affected by the shock. We call this set of predictions the
credit supply shock hypothesis.
The demand for goods produced by industrial firms fell and growth opportunities for these firms
disappeared during the crisis for several reasons. Retail sales peaked in November 2007.2 The saliency of
the crisis in financial markets makes it easy to forget that housing has always played a critical role in
economic activity and that the housing sector was in recession before the financial crisis.3 Further, it is
well-known that oil price increases are a drag on economic activity, and the price of oil peaked in 2008.4
Campbell, Giglio, and Polk (2010) document that one feature of the financial crisis is a sharp decrease in
corporate expected cash flows that takes place before September 2008. Finally, consumption fell sharply
during the last quarter of 2008, partly because of the increase in risk and partly because of disruptions and
decreases in consumer credit. Over the period for which the monthly data is available from the U.S.
Census Bureau (1992), retail sales had their worst percentage drop in October 2008, their second worst
drop in December 2008, and their third worst drop in November 2008. The next worst change was after
2
See U.S. Census Bureau, monthly retail sales adjusted.
Leamer (2007) concludes that “Of the components of GDP, residential investment offers by far the best early
warning sign of a recession. Housing starts fell roughly by half from the end of 2005 to the end of 2007.”
4
Hamilton (2009) concludes that the U.S. would not have been in a recession during the period from the fourth
quarter of 2007 to the third quarter of 2008 in the absence of the runup in oil prices.
3
2
September 11. Using data from the National Income and Product Account (NIPA) tables from the Bureau
of Economic Activity, seasonally-adjusted personal consumption peaked in the second quarter of 2008.
The drops in personal consumption in the third and fourth quarters of 2008 were the seventh and sixth
worst, respectively, since 1959. The drop in consumption was extremely abrupt immediately after the
events of September 2008. For instance, sales of the retailer Neiman Marcus were 25% in October 2008
lower than in October 2007. In contrast, sales in September 2008 were only 11% lower than the previous
year. This shock to demand decreased both revenues and cash flows. These dramatic changes in
consumption spending around the events of the fall of 2008 may have multiple causes, but assuredly part
of these changes reflect consumer fear caused by the financial chaos.
As a result of a decrease in demand, the value of firms’ tangible assets decreases: excess supply of
goods at current prices leads to a decrease in net worth and collateral, so firms cannot borrow as much as
they could previously (Kiyotaki and Moore (1997)). Firms become more highly levered, which worsens
agency problems between creditors and shareholders (Jensen and Meckling (1976); Bernanke and Gertler
(1989)). For firms whose debt is already risky, the increase in leverage leads to a debt overhang, which
makes equity issuance unattractive for shareholders (Myers (1977)). At least since Miller (1963), we
know that equity issues are particularly sensitive to the business cycle. In particular, firms tend to issue
equity when equity prices are increasing, not when they are falling. Thus the fall in the demand for goods
and growth opportunities leads to a decrease in net debt issuance and a decrease in net equity issuance.
With the demand shock, cash increases only to the extent that the cash savings from lower investment
exceed the cash losses from lower cash flows. We refer to this scenario as the demand shock hypothesis.
A striking feature of the recent crisis is the extent to which risk measures increased to unprecedented
levels. For instance, the VIX reached levels never before seen. An increase in risk increases risk premia,
which decreases firm values and increases the cost of external finance. Uncertainty about the environment
also increased, as aspects of the financial system that were believed to be robust (such as the commercial
paper market and the repo market) seemed to be failing. Caballero and Krishnamurthy (2008) emphasize
the impact of an increase in uncertainty about the environment in creating a flight to safety and cite
3
Greenspan, stating that “When confronted with uncertainty, especially Knightian uncertainty, human
beings invariably attempt to disengage from medium to long-term commitments in favor of safety and
liquidity…” The increase in risk and uncertainty led to a drop in the supply of capital across debt and
equity markets, and valuations fell as risk premia increased. As in our discussion of the demand shock,
the decrease in firm values worsens information asymmetries and agency problems, making external
financing relatively more expensive for firms in which information asymmetries and agency problems are
more prevalent. The risk increase also increases default probabilities for levered firms, raising their cost
of debt financing. Further, it provides incentives for firms to postpone investment and strengthen their
balance sheet by hoarding cash (see, for instance, Bernanke (1983)).
After the collapses of Fannie Mae, Freddie Mac, Lehman Brothers, AIG, and Washington Mutual in
September of 2008, corporate executives and investors had genuine concerns about whether the financial
system would still be functioning when they awoke the next day. There was also considerable uncertainty
about government policy, given that Bear Stearns had been rescued but Lehman was allowed to fail.
Pastor and Veronesi (2010) show that policy changes are accompanied by falling stock prices, increased
volatility, and increased risk premia. When the chairman of the Federal Reserve System and the Secretary
of the Treasury testified before Congress that the financial system was close to a collapse, corporate
executives would have been foolish to ignore that information. Economists describe the events that
happened in the fall of 2008 as a panic (Gorton (2010)). Whether or not the fear was rational, it would not
make sense to assume that business executives were unaffected and that this fear did not worsen the
impact of the increase in risk. We would expect an increase in risk to raise a firm’s cost of capital, making
financing more costly. This increase would be even worse from the perspective of executives if they felt
that the market was overreacting. It would further decrease the liquidity of markets, which increases the
cost of raising funds. Firms would seek stronger balance sheets, with more liquid assets. We would
therefore expect an increase in risk to lead to a drop in net debt and net equity issuance, a drop in payouts,
and an increase in cash holdings. We call this the risk hypothesis.
4
These three hypotheses are not mutually exclusive and all three play a role during the crisis. Our
focus is on whether the financial and investment policies of industrial firms help us understand the
relative importance of the three hypotheses. However, the importance of these hypotheses may differ
across firm types. For instance, a temporary restriction to credit would have little effect on firms with no
debt, and firms whose investment opportunities are already poor could not suffer much from a loss of
growth opportunities. There is also no reason for the relative importance of these hypotheses to be
constant throughout the crisis. It could be that one dominates at one stage and another at another stage.
Finally, a caveat is necessary at this point. Though drawing sharp differences between our three
hypotheses helps with the presentation, there are some relations among these hypotheses that blur their
boundaries. For example, a restriction in the supply of capital that is expected to be long lasting could
lead firms to invest in cash. Similarly, firms that expect the credit markets to be closed must still raise
funds to pay off loans or maturing debt issues. They would want to increase cash holdings for that
purpose.
We investigate firm financial policies using quarterly data, which is the highest frequency corporate
data available, from the start of the crisis to its peak. We use the second quarter of 2007 as both the end
and the peak of the credit boom since credit spreads for high yield bonds reached their lowest value in
that quarter. The third quarter of 2007 is generally considered to be the first quarter of the crisis and the
first quarter of 2009 is the peak of the crisis as it corresponds to the lowest value of the S&P 500 index.
We begin by examining asset-weighted averages. Equally-weighted averages for U.S. firms are
dominated by small firms and, consequently, are not a good proxy for the behavior of the economy as a
whole. Using asset-weighted data, there is no evidence of a crisis in the year after the peak of the credit
boom. Net debt issuance increases. Net equity issuance does not fall. The ratio of cash to assets does not
fall. The ratio of capital expenditures to assets does not fall. Slightly more than one year after the
beginning of the crisis, markets were further disrupted by the events of September 2008. In the following,
when we discuss financial policies after these events, we consider financial policies for the last quarter of
5
2008 and the first quarter of 2009.5 After the September events, net debt issuance falls, but not to levels
that are extreme outliers in the sample and the drop in net debt issuance is more than offset by a reduction
in equity repurchases. There is no evidence of changes in dividend payouts. However, cash as a percent of
assets (the cash ratio) increases by 1.30 percentage points in the two quarters following September 2008,
a two-quarter increase unmatched since the start of our sample. The asset-weighted evidence is not
consistent with a dominant role for the credit supply shock hypothesis. It is supportive of the demand
hypothesis and the risk hypothesis.
After considering the aggregate evidence, we examine various subsets of firms whose financial and
investment policies are expected to differ according to the three hypotheses. To isolate the impact of the
financial crisis, we use models for debt issuance, equity issuance, cash holdings, and investment
developed in the literature before the financial crisis. We estimate these models quarterly and use firm
fixed effects to account for unobserved firm characteristics. We then estimate coefficients for a crisis
indicator variable and an indicator variable for the last two quarters of the crisis. Using a difference-indifferences approach, these indicator variables are interacted with firm characteristics related to the
predictions of the three hypotheses.
We start with an analysis of net debt issuance. For the whole sample, net debt issuance as a
percentage of assets is significantly higher during the crisis by 11 basis points per quarter. Net debt
issuance is similar for small firms or unrated firms. Compared to other firms, it is significantly lower for
firms that are in the top quintile of cash holdings in June 2006 and for firms that are highly financially
constrained as of June 2006; we define firms as highly constrained if they were net borrowers in the
previous year, did not make payouts, had a Tobin’s q above one, and did not have access to public
markets. Except for the highly constrained firms, none of the evidence is supportive of the credit supply
shock hypothesis. After September 2008, net debt issuance collapses for the whole sample, falling by 96
basis points after accounting for firm characteristics used in the literature to predict net debt issuance. Net
5
One could argue, however, that the financial data for the end of the third quarter of 2008 already partly reflects the
impact of the events of September 2008. By considering only the two quarters we do, our analysis is conservative
and may ignore some of the impact of these events.
6
debt issuance falls less for large and investment-rated firms, but more for highly levered firms. Small
firms, unrated firms, and highly financially constrained firms do not appear to experience a collapse after
September 2008 that is worse than for the sample as a whole. Though a drop in net debt issuance is
consistent with all three hypotheses, the credit supply shock hypothesis predicts a worse drop for more
bank-dependent firms, which is not the case.
For the whole sample, net equity issuance is lower by 26 basis points per quarter during the crisis.
Except for the highly financially constrained firms, we find no evidence supporting the prediction of the
credit supply shock hypothesis that firms substitute equity for debt. Small firms and unrated firms
decrease net equity issuance more rather than less. However, highly financially constrained firms
decrease their net equity issuance abnormally after September 2008, but not before. In contrast to the
credit supply shock hypothesis, the demand shock and the risk hypotheses are consistent with a reduction
in net equity issuance.
The credit supply shock hypothesis predicts that firms use cash to offset the lack of availability of
bank financing. For the whole sample, cash holdings decrease during the crisis, but they increase
following September 2008. For large and investment-grade firms, however, cash holdings are not lower
during the crisis, and in fact increase during the crisis because of the post-September 2008 increase.
Strikingly, the highly constrained firms also increase cash holdings abnormally during the crisis. The
explanation for this is that these firms have surprisingly high operating cash flows during the first year of
the crisis. Firms with high cash holdings decrease cash holdings throughout the crisis. If firms increase
their cash holdings after September 2008 because they expect credit to be restricted, we would expect
such an increase to be higher for firms that are bank dependent. This is not the case. Unrated firms, small
firms, and highly financially constrained firms do not increase their cash ratio differently from other firms
after September 2008.
Finally, we examine firms’ investment policies. Small firms, unrated firms, highly constrained firms,
and firms with large amounts of short-term debt do not decrease capital expenditures more than other
firms during the crisis. Such a finding is not consistent with the credit supply shock hypothesis. However,
7
firms with stronger balance sheets (more cash, less leverage, higher debt rating) reduce capital
expenditures less after September 2008. In fact, firms in the highest quintile of cash holdings experience
no reduction in capital expenditures, while firms in the highest quintile of leverage experience a reduction
in capital expenditures that is twice the size of the rest of the sample. This evidence is strongly supportive
of the credit transmission mechanism highlighted by Bernanke and Gertler (1989) and Kiyotaki and
Moore (1997). With a shock to future cash flows, firms find it more expensive to borrow, which leads to a
drop in capital expenditures. We also examine the evolution of R&D expenditures. All firm types
considered spend more on R&D in the fourth quarter of 2008 than the last quarter of the credit boom, but
these expenditures drop in the first quarter of 2009.
Our evidence complements the literature on the implications of the financial crisis for industrial
firms. Most of this literature has focused on identifying the effects of the credit supply shock, while we
focus instead on assessing the economic importance for industrial firms of the credit supply shock, the
demand shock, and the risk shock. In particular, using survey data, Campello, Graham, and Harvey
(2010) provide evidence that firms that were credit constrained in 2008 reduce their spending plans,
bypass attractive investment opportunities, and burn more cash. Almeida, Campello, Laranjeira, and
Weisbenner (2009) demonstrate that firms which had a substantial proportion of their long-term debt
maturing immediately after the third quarter of 2007 reduce investment substantially in comparison to
other firms. Campello, Giambona, Graham, and Harvey (2009) show that credit-constrained firms draw
down credit lines during the crisis, but also face difficulties in renewing credit lines. The importance of
credit lines is also emphasized by Ivashina and Scharfstein (2010a, 2010b), who provide evidence that
firms draw down their credit lines after the September 2008 and hoard the funds drawn in cash. Ivashina
and Scharfstein (2010b) conclude that “the primary motivation behind the draw-downs was panic from
the financial market uncertainty.” Gao and Yun (2009) show that riskier firms experienced a sharp drop in
commercial paper issuance. Duchin, Ozbas, and Sensoy (2010) find that firms with greater cash holdings
one year before the start of the crisis reduce investment less early in the crisis but not later. They interpret
their evidence as consistent with the existence of a supply shock to external finance that is less costly for
8
firms with greater cash holdings. Iyer, Lopes, Peydró, and Schoar (2010) find no credit supply reduction
for large firms in Portugal, but find that younger firms with weak bank relationships are affected. Finally,
Kuppuswamy and Villalonga (2010) show that the financial crisis had less of an impact on diversified
firms.
The paper proceeds as follows. In Section 1, we introduce the dataset we use. In Section 2, we show
how debt and equity issuance, cash holdings, and capital expenditures evolve during the credit boom and
the financial crisis for the asset-weighted and the equally-weighted sample averages. In Sections 3
through 6, we investigate the predictions of the three hypotheses using regression models of debt and
equity issuance, cash holdings, and capital expenditures, respectively. We discuss the interpretation of our
results and conclude in Section 7.
Section 1. The sample
Most empirical work in corporate finance uses annual data. For our purpose, such data is unsuitable
since it would force us to ignore how corporate financial policies evolve during the crisis and would make
it impossible to examine the financial crisis from the top of the credit boom to the peak of the crisis
which, as already discussed, we define as the first quarter of 2009. We therefore use quarterly data
collected from the CRSP/Compustat Merged (CCM) Fundamentals Quarterly database for 1983-2009.
There are distinct problems with the use of quarterly data. First, many of the Compustat data items are
only provided annually, so less detailed data is available on a quarterly basis than on a yearly basis.
Second, many industries have seasonal factors. There is little we can do to deal with the lack of data
availability, but we can address the seasonality issue. The first approach we use to address the seasonality
issue is that we compare quarters to identical quarters in other years. The second approach is that we
estimate models that specifically allow for seasonality.
In our investigation, we use issuance data from the cash flow statement following Fama and French
(2008). The capital structure literature often uses other measures, such as changes in debt or equity above
a threshold (see, for instance, Leary and Roberts (2005)) or only public issues (e.g., DeAngelo,
9
DeAngelo, and Stulz (2010); Erel, Julio, Kim, and Weisbach (2010)). In this paper, we focus on the
funding obtained by corporations from all sources, not just banks or public markets, since substitution
across funding sources could help firms offset the impact of a bank credit contraction (see, e.g., Iyer,
Lopes, Peydró and Schoar (2010) for evidence of such substitution). We also want to understand the
magnitude of financing flows in comparison to more normal times, so that net issuance close to zero is of
interest to us.
The quarterly issuance data we need in our investigation is only available beginning in the third
quarter of 1983. Consequently, our sample effectively starts from that quarter and ends with the first
quarter of 2009. We delete observations with negative total assets (atq), negative sales (saleq), negative
cash and marketable securities (cheq), cash and marketable securities greater than total assets, and firms
not incorporated in the U.S. If a firm changes its fiscal-year end, and thus a given data quarter is reported
twice in Compustat (for both the old fiscal quarter and the new fiscal quarter) we retain the observation
for the new fiscal quarter only. Finally, we eliminate all financial firms, which we define as firms with
SIC codes between 6000 and 6999; we also eliminate utilities, which we define as firms with SIC codes
between 4900 and 4949.
Section 2. The aggregate evidence
At the start of the crisis, our sample includes 3,198 firms. When we divide firms listed on the NYSE
into quintiles based on total assets and assign non-NYSE firms to these quintiles, we find that 2,021 of the
3,198 firms are smaller than the largest firm in the bottom quintile of NYSE firms. Consequently, at the
beginning of the crisis, two-thirds of our firms are small firms with assets less than $715.7 million. The
average and median assets for these firms are $196.6 million and $127.6 million, respectively. In contrast,
the mean and median of assets are $31.9 billion and $16.1 billion, respectively, for the firms in the top
quintile. Equally-weighted averages reflect small public firms and are not indicative of the behavior of the
economy as a whole. Asset-weighted averages better reflect the economy as a whole. Though none of our
data includes private firms, asset-weighted capital expenditures increase in the first year of the crisis as
10
they did for the U.S. economy as a whole.6 In this section, we first investigate the implications of assetweighted results for debt and equity issuance, cash holdings, capital expenditures, and operating cash
flow for our three hypotheses. We then turn briefly to equally-weighted results. At the end of the section,
we examine the role of credit lines for a random sample.
Section 2.1. Asset-weighted results
Panel A of Table 1 shows the asset-weighted results from the first quarter of 2005 to the first quarter
of 2009. The first two columns examine net debt issuance. We use two measures to examine net debt
issuance. The first measure is obtained from the statement of cash flows and is calculated as long-term
debt issuance (dltisy) minus long-term debt retirement (dltry) divided by lagged assets.7 This measure is
higher in the first year of the crisis than in the last year of the boom, but not significantly so.8 Its value at
the peak of the crisis is not different from its value at the peak of the boom. Strikingly, this ratio has a
negative value 13 times in our dataset, so none of the values during the crisis are extreme values for this
variable. An obvious concern is that a relatively stable net debt issuance could mask large changes in
long-term debt issuance. This is not the case.
We also use a broader measure of debt issuance that includes short-term debt. We refer to this
measure as net total debt issuance. It is calculated from balance sheet data and includes changes in both
long-term debt (dlttq) and debt in current liabilities (dlcq) during the quarter. Net total debt issuance is
6
Using the September 30, 2010 version of the NIPA data, aggregate gross investment by domestic businesses is
$6,346 billion in the year before the crisis and is $6,577 in the first year of the crisis. Aggregate gross investment
falls sharply in the fourth quarter of 2008 and again in the first quarter of 2009. It reaches its lowest value in the
second quarter of 2009, but the drop from the first quarter to the second quarter is small. The same conclusions hold
for aggregate net investment by domestic businesses.
7
Many of the quarterly Compustat variables, including dltis and dltry, are reported on a year-to-date basis. For these
variables, in the second, third, and fourth quarter of each fiscal year, the quarterly value is calculated by subtracting
the lagged value from the current value.
8
We assess the significance of these changes using two different approaches. The first approach (seasonalityadjusted p-values) is extremely conservative in that we compare the change of interest to changes over identical
quarterly calendar periods to account for seasonality. For example, to investigate the statistical significance of a
change after September 2008, we use the distribution of two-quarter turn-of-the-year changes in our sample. This
approach has low power since it uses only 26 two-quarter changes, but it fully adjusts for seasonality. The second
approach (Newey-West p-values) uses all two-quarter changes but relies on Newey-West t-statistics to account for
overlap.
11
also not lower during the first year of the crisis than during the last year of the boom. This evidence is
consistent with the findings of Chari, Christiano, and Kehoe (2010) who, using macroeconomic data, find
no decrease in bank lending to firms.9 However, our measure is broader than theirs since it includes other
sources of borrowing besides banks, including public markets.10 Net debt issuance turns negative in the
first quarter of 2009. However, there are six quarters since 1983 in which this ratio has a lower value than
in the first quarter of 2009. Consequently, the asset-weighted data does not support the notion that the
crisis resulted in exceptionally low aggregate net debt issuance; not only are the values that obtain during
the crisis not the lowest in our sample period, they are not even in the bottom 5% of the distribution.
Importantly, however, these numbers reflect net debt issuance given the policy interventions that took
place and they say nothing about what the net debt issuance would have been in the absence of these
interventions.
It is interesting to compare the net total debt issuance in the first quarter of 2009 to the net total debt
issuance in the last quarters of the 1990/1991 and of the 2001 recessions (results not tabulated). For both
of these recessions, net debt issuance in the last quarter of the recession is closer to net debt issuance at
the peak of the boom in 2007 than it is to net debt issuance at the peak of the financial crisis. In the fourth
quarter of 2001, aggregate net debt issuance is 0.97% while in the first quarter of 1991 it is 0.65%. The
negative value for aggregate net debt issuance in the first quarter of 2009 is therefore quite different from
the experience of the previous two recessions.
We turn next to the aggregate ratio of net equity issuance to lagged assets, where net equity issuance
is defined as aggregate equity issuance (sstky) minus aggregate equity repurchase (prstkcy). For the assetweighted data, net equity issuance is negative because firms in the aggregate repurchase stock rather than
9
It is also consistent with evidence in Boyson, Helwege, and Jindra (2010) that bank balance sheets did not shrink
during the crisis.
10
With the NIPA data cited earlier, businesses are net borrowers until the end of the third quarter of 2008, but their
net borrowing is lower in the first year of the crisis than in the last year of the boom, a finding which is inconsistent
with our finding here. Since the NIPA data covers private firms, a possible explanation for the inconsistency is that
private firms face greater difficulties in borrowing than public firms. Using the flow-of-funds data, corporate credit
grows more in 2008 than in eight years since 1977. It also grows in each quarter of 2008 as well as in the first
quarter of 2009.
12
issue it. The net equity issuance in the first year of the crisis is similar to the net equity issuance in the last
year of the boom. However, after the first year of the crisis, net equity issuance increases because
repurchases (not tabulated) decrease sharply. Strikingly, therefore, firms repurchase the least when stock
prices are the lowest and the most at the top of boom when they are highest. Though we do not show
dividends in the table, dividends to assets do not decrease significantly during the crisis. In comparison to
the other recessionary periods, net equity issuance is positive in the first quarter of 1991 and hardly
changes during the recession of 2001.
Data on the asset-weighted ratio of cash to assets is reported in the next column, where cash is cash
and marketable securities (cheq). The cash ratio shows a u-shape during this period. It starts in 2005 with
a value of 10.42%, falls to 9.49% at the top of the credit boom, and ends the period at 10.18%. Thus, at
the peak of the financial crisis, firms hold more of their assets in the form of cash than at the start of the
crisis. After the top of the credit boom, the ratio keeps falling and reaches its lowest point (8.89%) at the
end of the third quarter of 2008. In the last two quarters alone, the cash ratio increases significantly by
1.30 percentage points, representing an increase in cash holdings of 14.5%.
There is no other two-quarter period in our sample where the cash ratio increases by more than in the
two quarters after September 2008.11 In the recession of 1990-1991, which was also associated with a
credit crunch, there are only trivial changes in the asset-weighted cash ratio. More specifically, the assetweighted cash ratio falls by 0.06 percentage points from the third quarter of 1990 to the first quarter of
1991. Nothing like the post-September 2008 increase in cash holdings takes place in that recession. Such
an increase does occur in the 2001 recession, which was not associated with a credit crunch. In that
recession, the aggregate cash ratio increases by 0.62 percentage points. However, just about all of the
increase in the cash ratio takes place after September 11. The aggregate cash ratio falls initially in the
crisis of 1998. This crisis started in August and the LTCM collapse was in September. The cash ratio falls
11
The next highest increase is 1.22 percentage points from the third quarter of 1999 to the first quarter of 2000. There was much
concern about potential software problems at the turn of the century and this concern most likely explains this hoarding. Because
cash holdings were lower at that time, the percentage change in cash is higher at the turn of the century (18.8%) than from the
third quarter of 2008 to the first quarter of 2009 (14.5%).
13
from 6.52% at the end of June to 6.28% at the end of September. It then increases to 6.63%. The total
increase around the LTCM collapse is therefore a trivial 0.11%.
There is little evidence that the crisis impacted capital expenditures in 2008. Asset-weighted capital
expenditures do not fall in the first year of the crisis. In addition, capital expenditures in each quarter of
2008 are roughly the same as in the corresponding quarter in 2007. Further, asset-weighted capital
expenditures do not fall in the last quarter of 2008. Though capital expenditures fall in the first quarter of
2009, so do operating cash flows. Our measure of operating cash flow represents the cash flow available
for discretionary investment and is calculated following the approach of Minton and Schrand (1999) (see
Appendix for details). Though we do not show the results in the table, we also investigate the evolution
of R&D expenses. There is no difference in R&D expenses to assets from the last twelve months of the
boom and the first twelve months of the crisis. However, R&D falls in the first quarter of 2009.
One concern is that our results could be influenced by the changing composition of the sample as
firms cease to exist. Consequently, in untabulated results, we construct a sample of firms that exist
continuously from the end of the first quarter of 2007 to the end of the first quarter of 2009. There are
2,547 firms that satisfy this requirement. The change in the aggregate cash ratio after September 2008 for
these firms is 1.47 percentage points, as compared to 1.30 percentage points for the whole sample of
firms. All the other patterns we discuss continue to exist for this sample. We also verify that the dollar
amount of cash holdings increases as well and find that, over the crisis, the aggregate dollar amount of
cash held by firms that are continuously in existence increases by roughly $100 billion from the top of the
boom to the peak of the crisis. For the whole sample, cash holdings increase by $89 billion in the two
quarters after September 2008.
Section 2.2. Equally-weighted results
We now turn to the equally-weighted results provided in Panel B of Table 1. While we require that
firms have data for all variables when examining the asset-weighted results, so that the denominator of
the ratios is the same, for the equally-weighted results we construct averages for each variable separately
14
and only require data for that variable. The number of observations reported is the number of firms for
which we have cash and assets data. We winsorize the equally-weighted results at the 1% and 99% levels
to mitigate the influence of outliers.
The story of net debt issuance differs substantially after September 2008 when we examine equallyweighted averages instead of asset-weighted averages. During the first twelve months of the crisis, net
debt issuance is the same as during the last twelve months of the credit boom, irrespective of the measure
of net debt issuance used. However, net debt issuance collapses after September 2008 when examining
the equally-weighted results. For both measures of net debt issuance, we find that net debt issuance is
negative in the two quarters after September 2008, i.e. firms are repaying debt on net rather than
borrowing. Further, the net total debt issuance measure (that includes short-term debt) reaches its lowest
value during our sample period in the first quarter of 2009, and the net long-term debt issuance measure
reaches its second lowest value in that quarter. In the two previous recessions, equally-weighted net debt
issuance falls, but much less; it stays positive in the recession of 1990-1991 and becomes trivially
negative in the recession of 2001.
In contrast to net debt issuance, net equity issuance falls sharply in the first year of the crisis. In the
last twelve months of the boom, net equity issuance is roughly 1% per quarter; it drops to 0.5% in the first
twelve months of the crisis. Strikingly, it turns negative in the last quarter of 2008, albeit by a small
amount – this is the only quarter in our sample where equally-weighted net equity issuance is negative.
The evidence is inconsistent with the view that the dominant effect was a curtailment of credit. Firms did
not substitute equity issuance for debt issuance. Instead, equity issuance drops substantially before debt
issuance drops. The net equity issuance drop is large compared to the two previous recessions. In the
recession of 2001, equally-weighted net equity issuance actually increases.
The cumulative shortfall in financing cash flow from the decrease in net equity issuance for the
equally-weighted results is on average more than double the cumulative shortfall from the reduction in net
debt issuance. To see this, suppose that from the start of the crisis to its peak, firms continued issuing
equity at the same rate as in the last quarter of the credit boom. Over these seven quarters, firms would
15
have issued equity corresponding to approximately 9.94% of assets.12 Instead, they issue equity
corresponding to 2.50% of assets, for a shortfall of 7.44% of assets. In contrast, had firms kept issuing
debt at the same rate as they did in the last quarter of the credit boom, they would have issued debt equal
to approximately 7.35% of assets. Instead, they issue debt equal to 3.79% of assets, for a shortfall of
3.56% of asset.
Firms decrease their cash to assets ratio by 1.1 percentage points from the first quarter of 2005 to the
top of the boom. The magnitude is similar to the 1.3 percentage point drop for the asset-weighted results.
However, instead of a u-shape pattern of cash holdings, the cash ratio falls from the top of the boom to the
peak of the crisis by 1.78 percentage points. The cash ratio increases by 0.58 percentage points after
September 2008. We also investigate the medians for the sample used for the equally-weighted average.
We find an increase in the median cash ratio after September 2008 of 0.91 percentage points. In
evaluating significance, we report p-values for paired t-tests using the firms that exist in the third quarter
of 2008 and the first quarter of 2009. For these firms, the cash ratio increases by 0.70 percentage points,
which is statistically significant at the 1% level. The evidence is consistent with the view that small firms
use their precautionary cash holdings to cope with adverse shocks. This behavior does not show up in the
asset-weighted results because, while there are many small firms, their weight in the asset-weighted
average is small. In the 1990-1991 recession, the equally-weighted cash ratio falls by 0.42% when we
average across the same firms. In contrast, in the 2001 recession, the average ratio increases by 1.28%.
However, in the 2001 recession, the cash increase is a post-September 11 phenomenon.
Finally, we consider capital expenditures. Again, equally-weighted capital expenditures are
essentially unchanged from the last year of the boom to the first year of the crisis. However, they fall
sharply in the first quarter of 2009. This one-quarter fall in 2009 is greater than the fall in either of the
previous recessions from the peak to the last quarter of the recession. As for the asset-weighted results,
12
Note that, for simplicity, we use actual assets in these computations rather than what the assets would have been
had equity issuance or debt issuance been different.
16
the drop in capital expenditures is accompanied by a very large drop in operating cash flows. In fact,
capital expenditures decrease proportionately less than cash flows.
Section 2.3. Lines of Credit
As discussed in our review of the literature, much attention has been paid to the role of credit lines
and credit line draw-downs during the financial crisis. Using the approach of Sufi (2009), we construct a
random sample of 300 firms as of the second quarter of 2007 to examine the economic importance of
credit line draw-downs for firms meeting our sampling criteria. We then obtain data on lines of credit
available and draw-downs from 10-Qs and 10-Ks. Table 2 shows data for the asset-weighted sample and
the equally-weighted sample.
Not surprisingly, given the existing literature, many firms have access to credit lines. Out of our 300
firms, 243 or 81% have credit lines in the second quarter of 2007. Columns 4-6 of Table 2 examine only
those firms in each quarter that have existing lines of credit. Column 4 shows that the percentage of firms
with credit lines having new draw-downs peaks in the fourth quarter of 2007 at 37.28%. The next column
shows the percentage of new draw-downs in each quarter as a percentage of the total line of credit. This
amount also peaks in the fourth quarter of 2007. Column 6 shows the ratio of total draw-downs
outstanding as a fraction of the total line of credit. This fraction increases the most in the last quarter of
2007; the increase in the last quarter of 2008 is only 1 percent. Our results for the average total drawdown outstanding to total line of credit are higher throughout the crisis compared to the estimates in
Ivashina and Scharfstein (2010b). In their sample, the percentage drawn increases by 3.81% (from
20.04% to 23.85%) in the third quarter of 2008 but falls in the fourth quarter to 23.30%. We do not
observe such an increase in the third quarter of 2008. However, our sample selection procedure is
different. Their sample includes only manufacturing firms that have revolving lines before the crisis and
still outstanding in the fourth quarter of 2008. Our sample makes no such requirements: It only requires
firms to be in existence before the crisis and is chosen randomly from more than 3,000 firms in
Compustat. These sample differences could well explain the differences in results.
17
In columns 7-10 of Table 3, we examine the importance of credit lines and credit line draw-downs
using asset-weighted averages and equally-weighted averages. Looking first at the asset-weighted sample,
we find that the ratio of total lines of credit to total assets is 9.5% at the start of the crisis. The aggregate
amount of new draw-downs per quarter divided by total assets is less than 1% of assets in all quarters
except one, the fourth quarter of 2007, when it is 1.28%. Yet, during that quarter, in untabulated results,
the ratio of cash holdings to total assets for our random sample falls. New draw-downs in the last two
quarters are much lower than in the fourth quarter of 2007. We turn next to the equally-weighted results.
In our equally-weighted sample, credit lines represent 15.0% of assets in the sample formation quarter,
which is not very different from the 16.5% in the Sufi (2009) sample or from the data collected by Gao
and Yun (2009) for manufacturing firms in 2008. The equally-weighted average of the new draw-downs
in the fourth quarter of 2007 is 1.72%. It is much lower in the fourth quarter of 2008, when it is 1.03%.
This percentage drawdown is similar to the percentage drawdown in the first quarter of 2008. It is also
similar to the new drawdowns to assets in the third quarter of 2006 (not tabulated) for the same firms,
which was 0.96%. This evidence on draw-downs seems consistent with both the credit channel hypothesis
and the risk hypothesis.
Section 3. Did credit contract more for firms more likely to be bank dependent?
Section 2 shows that equally-weighted net total debt issuance does not fall in the first year of the
crisis. It falls sharply after September 2008, however. We next explore whether net total debt issuance is
lower for firms that are more likely to be bank dependent, as is predicted by the credit supply shock
hypothesis. First, we consider raw data. Second, we use abnormal net total debt issuance from a
regression model.
In Panel A of Table 3, we assign firms to subgroups as of the end of the previous quarter. The first
column of Panel A of Table 3 shows the data for net total debt issuance for firms with investment-grade
debt. Strikingly, investment-rated firms borrow substantially more during the first year of the crisis than
during the last year of the boom. The increase is statistically significant. After September 2008, the net
18
debt issuance of investment-rated firms falls to 0.01% in the first quarter of 2009. This contrasts with the
lowest net total debt issuance for investment-rated firms in the sample, which is -0.23%. Surprisingly,
however, the net long-term debt issuance of investment-rated firms is not different in the first quarter of
2009 from the last quarter of the boom (not reported in the table). Turning to unrated firms, which are
more bank dependent as they do not have access to public debt, we find that their net total debt issuance is
not significantly different in the first year of the crisis from the last year of the credit boom. However, the
net debt issuance of unrated firms in the first quarter of 2009 is lower than in any other sample quarter.
Next, we divide firms into five quintiles of asset sizes using NYSE cut-offs. We expect the smallest
firms to be more bank dependent and reproduce results for only these firms. Small firms are typically
considered to be financially constrained (e.g., Almeida, Campello, and Weisbach (2004)) and their main
source of debt finance is loans from banks. We find that net total debt issuance of the smallest firms does
not fall in the first year of the crisis, but it reaches an extremely low value in the first quarter of 2009. Not
surprisingly, the net total debt issuance of the largest firms also does not fall during the first year of the
crisis (not reported in the table), but it is -0.01% in the first quarter of 2009. Over our entire sample
period, the four smallest size quintiles never experience a net total debt issuance rate lower than the one
they experience in the first quarter of 2009. However, the situation of the largest firms is different; they
experience lower net total debt issuance in 18 other quarters during the sample period. Since small firms
rely much more on bank loans than large firms, our evidence is consistent with the evidence in Becker
and Ivashina (2010) of a serious supply shock to bank debt in 2009.
We also construct a sample of firms that are financially constrained before the crisis using a definition
that is similar to the definition of Korajczyk and Levi (2003). We define a firm to be KL financially
constrained in quarter t if it (1) does not pay dividends, (2) does not have net equity repurchases, (3) does
not have a credit rating, and (4) has a Tobin’s q greater than one (defined as the market value of the assets
divided by the book value, where market value of assets is book value of assets minus book equity plus
market value of equity) in quarter t-1. With this definition of financially constrained firms, we isolate
firms that have investment opportunities that outpace their internally generated cash flow and that do not
19
have access to public debt markets. The net debt issuance of these firms is not different in the first year of
the crisis from the last year of the credit boom. Strikingly, these firms have higher net debt issuance in the
third quarter of 2008 than at the top of the credit boom. However, their net debt issuance falls sharply
after the third quarter of 2008. There is no quarter before the crisis when these firms have negative net
debt issuance, but in the first quarter of 2009 their net debt issuance is -0.10%.
The next column considers firms that have no debt at the end of the previous quarter. Presumably,
such firms should have unused debt capacity. These firms are able to borrow at the peak of the crisis.
They have the highest net debt issuance of any subgroup in the first quarter of 2009, at 0.36%. We then
investigate firms whose cash to assets ratio is in the top quintile of cash to assets ratios. These firms do
not have to borrow – they have extremely large amounts of cash by any standard. At the start of the crisis,
these firms have an average cash to assets ratio of 61%; their median cash ratio is the same. Their net debt
issuance is significantly lower during the crisis than in the last year of the boom. Further, their net debt
issuance is the second lowest of any subgroup in the first quarter of 2009. The last column examines
firms whose leverage ratio is in the top quintile of debt to asset ratios. These firms are highly levered, so
that a drop in firm value would increase the risk of their debt sharply. Their net debt issuance is
insignificantly different in the first year of the crisis, but it plummets afterwards. This is the group of
firms with the most negative net debt issuance of any group.
While net debt issuance attains extreme values in the first quarter of 2009 for firms that are more
likely to be bank dependent, so does operating cash flow (not reported). For almost all subgroups,
operating cash flow in the first quarter of 2009 is roughly 200 basis points less than in the third quarter of
2008. For many types of firms, the operating cash flow of the first quarter of 2009 is the lowest in the
sample. For instance, for unrated firms, operating cash flow in the first quarter of 2009 is 0.88%. The
lowest operating cash flow between 1983 and 2004 is 1.20% and the average is 2.49%. Such low cash
flows would make it harder for firms to raise funds even in the absence of a credit supply shock. It is
noteworthy that the mean KMV EDF, a widely used estimate of the probability of default, for the North
20
American CDX index (125 investment grade names) increased from 0.69% on September 15, 2008 (the
day of the Lehman bankruptcy) to 2.99% at the end of March 2009.
The raw data on net debt issuance does not allow us to understand whether the changes in net debt
issuance are unusual during the crisis (after controlling for the determinants of debt issuance) and whether
they are more so for firms that are more likely to be bank dependent. To estimate abnormal debt issuance,
we follow Fama and French (2008). We estimate their models from 1995 to 2009. However, they estimate
their models using annual data, and some of the variables they use are not reported quarterly.
Consequently, we modify some variable definitions to account for the limitations of quarterly data. We
regress net debt issuance of period t on the following variables computed for period t-1: the percentage
change in assets from the previous quarter (negative coefficient), operating cash flow to assets (positive
coefficient), dividends to assets (positive coefficient), a dummy variable equal to one if the firms pays
dividends (negative coefficient), lagged leverage (negative coefficient), an indicator variable equal to one
if book equity is negative (positive coefficient), the lagged log market capitalization (negative
coefficient), the market-to-book ratio (positive coefficient), lagged cash over assets (negative coefficient),
lagged debt in current liabilities over assets (negative coefficient), R&D over assets (positive coefficient),
capital expenditures over assets (positive coefficient), and the change in the stock price (negative
coefficient). To save space, we do not reproduce the regression coefficients on these regressors. All
coefficients are significant in all the regressions we estimate and these coefficients are similar in all the
regressions as well. We take into account seasonal effects through indicator variables. Only the fourth
quarter indicator variable is significant and its coefficient is always negative. We also use firm fixedeffects to capture unmodeled firm characteristics.
To capture changes during the financial crisis, we use a financial crisis indicator variable and a crisis
peak indicator variable. The financial crisis indicator variable equals one for each crisis quarter. The crisis
peak indicator variable equals one for each quarter after September 2008. To test the hypothesis that firms
of a given subgroup have a different change in net total debt issuance from the other firms, we create an
indicator variable for that subgroup and interact that indicator variable with the crisis indicator variables.
21
The interactions capture how the subgroup reacts differently to the crisis from the remainder of the
sample. For the regressions we report in Panel B of Table 3, we form the subgroups at the end of the
second quarter of 2006 as in Duchin, Ozbas, and Sensoy (2010), so there is no reason to believe that the
crisis itself affects a firm’s assignment into a subgroup. We also estimate regressions with quarterly
assignments to groups, however, and discuss the cases where the results differ meaningfully.
The first regression is for the whole sample. We see that net total debt issuance is higher by 0.11%
during the crisis, but the value of the crisis peak indicator variable is -0.96%, so that for the last two
quarters of the sample, net total debt issuance is abnormally low by -0.85%. We next investigate whether
abnormal net total debt issuance is significantly different for the unrated firms and the smallest firms.
Both interactions are positive for unrated firms, indicating that their net debt issuance drops by less than
for the other firms both during the crisis and during the last two quarters of our sample. The interactions
are insignificant for the smallest firms. The crisis interaction is not significant for the KL financially
constrained firms, but the crisis peak indicator variable is positive and significant (not reported). We
create a subgroup of KL financially constrained firms that have positive net debt issuance in the year
ending in June 2006. As discussed earlier, we call these highly constrained firms. These firms borrow less
during the crisis as the crisis indicator variable interaction has a value of -0.59%. However, the peak crisis
interaction is not significant for these firms and the coefficient is positive, so their net debt issuance does
not change more than for other firms after September 2008. It is striking, though, that these firms’ average
operating cash flow normalized by assets in the first year of the crisis is 66.20% higher than their average
operating cash flow normalized by assets in the last year of the credit boom. Firms with no debt at the end
of the second quarter of 2006 have a positive and significant crisis interaction as well as a positive and
significant peak crisis interaction, indicating that they have abnormal net debt issuance relative to other
firms. In the next regression, we show that firms with high holdings of cash borrow less during the crisis
than other firms, but they have positive abnormal net debt issuance after September 2008, so their net
debt issuance falls significantly less than for other firms.
22
In the last column, we turn to firms that are in the top quintile of leverage at the end of the second
quarter of 2006. These firms have abnormal net borrowing of -1% per quarter during the crisis and their
abnormal borrowing falls by an additional 1.28% per quarter after September 2008. In results not reported
in the table, we also consider firms that at the end of the second quarter of 2006 are in the top quintile of
the ratio of short-term debt to assets. These firms face the greatest rollover risk. They have abnormally
low net total debt issuance during the crisis and the crisis peak interaction is negative as well. However,
both the crisis interaction and the crisis peak interaction have absolute values that are lower than for the
firms in the top quintile of leverage.
A concern with our approach is that more bank-dependent firms could be subjected to different
shocks during the crisis that have nothing to do with these firms being bank dependent. As a result, the
patterns we uncover could be caused by these unmodeled shocks. One way to investigate this possibility
is to check whether our results still hold when we estimate our regressions within industries since shocks
would be more similar within industries. Consequently, we estimate all the models for firms producing or
selling durable consumer goods.13 Obviously, this approach restricts the sample as the firms producing or
selling durable consumer goods represent slightly more than 10% of the total number of firms in the
sample. Nevertheless, the patterns documented for the whole sample hold when we restrict firms to this
industry.
The regression evidence overall shows the net debt issuance of firms that would generally be
considered to be financially constrained in the literature (namely the smallest firms, unrated firms, and
firms satisfying the Korajczyk and Levi (2006) definition of financially constrained firms) does not fall
abnormally compared to the net debt issuance of other firms during the crisis. However, highly financially
constrained firms experience a significant reduction in their net debt issuance during the crisis relative to
other firms, though this finding does not hold at the peak of the crisis. The highly financially constrained
firms reduce their net debt issuance before September 2008 in a way that is consistent with the hypothesis
13
We define the durable consumer goods industry based on the Fama-French 49 industry classification and include
Consumer Goods, Apparel, and Retail.
23
that more bank-dependent firms were more affected. We will see, however, that the story of these firms is
more subtle. In any case, it is worth noting that the assets of the highly financially constrained firms
represent less 1% of the total assets of the firms in the sample at the end of June 2006. Finally, highly
levered firms and firms with a high short-term debt ratio borrow less during the crisis and even less at the
peak of the crisis.
Section 4. Is there a substitution effect towards equity?
In this section, we investigate net equity issuance. In the first column of Panel A of Table 4, we show
the evolution of net equity issuance for investment grade firms. These firms repurchase more than they
issue. The level of net equity issuance of these firms is not significantly different in the first year of the
crisis from the last year of the boom. However, net equity issuance increases dramatically after September
2008 for these firms as they decrease repurchases, reaching -0.14% in the first quarter of 2009. We
already saw that for the equally-weighted average, net equity issuance falls sharply during the crisis.
When we consider subgroups of firms that are more likely to be bank-dependent, the result is even
stronger, as seen in Panel A of Table 4. In particular, for the smallest firms, net equity issuance is 1.90%
per quarter in the last year of the boom, but 1.12% per quarter in the first year of the crisis. Net equity
issuance drops to 0.52% in the first quarter of 2009 for these firms, a net equity issuance rate which is the
lowest since the start of our sample period in 1983. A similar result holds for unrated firms. Net equity
issuance of KL constrained firms falls in the first year of the crisis by almost 80 basis points. Further,
their net equity issuance in the last quarter of 2008 is slightly more than one sixth of their net equity
issuance at the top of the credit boom. For firms with no debt, net equity issuance is 2.05% during the last
quarter of the credit boom. It is negative each quarter of 2008. Prior to 2008, net equity issuance of nodebt firms was negative in only one quarter, the third quarter of 2002. The high cash firms are heavy net
equity issuers in the year after selection as their net equity issuance averages 3.13% of assets per quarter,
but it drops to 1.59% of net assets in the first year of the crisis and falls to 0.19% in the fourth quarter of
24
2008 before increasing to 1.20% in the first quarter of 2009. Finally, the high leverage firms significantly
decrease their net equity issuance in the first year of the crisis.
It could be that fundamentals led to an even sharper decrease in net equity issuance than the one we
observe, i.e. there could have been substitution towards equity by firms that are more bank-dependent in
the sense that they would have abnormally high net equity issuance given the evolution of fundamentals.
To examine this possibility, we estimate regression models of net equity issuance. These models again
follow Fama and French (2008) and are the same as the ones used for net debt issuance. Our control
variables (and the sign of their coefficients from the regressions) are the values of the following variables
computed at the end of the previous quarter: the percentage change in assets from the previous quarter
(negative coefficient), operating cash flow to assets (negative coefficient), dividends to assets (not
significant), a dummy variable equal to one if the firms pays dividends (not significant), lagged leverage
(positive coefficient), an indicator variable equal to one if book equity is negative (not significant), the
lagged log market capitalization (negative coefficient), the market-to-book ratio (positive coefficient),
lagged cash over assets (negative coefficient), lagged debt in current liabilities over assets (positive
coefficient), R&D over assets (positive coefficient), capital expenditures over assets (positive coefficient),
and the change in the stock price (positive coefficient). We also include indicator variables for quarters to
capture seasonality. The indicator variable for the third quarter is significantly negative and the seasonal
variable for the fourth quarter is significantly positive. To save space, we do not reproduce the
coefficients on these variables in the table. All regression coefficients are similar in sign and magnitude in
each regression.
Table 4 reproduces estimates for the crisis indicator variables and interactions. We see from the first
regression that net equity issuance is abnormally low during the crisis, but that compared to the other
quarters of the crisis, it is higher after September 2008. This abnormally higher net equity issuance after
September 2008 could be evidence of substitution towards equity because of an impaired credit channel.
However, for that to be the case, we would want abnormally high net equity issuance for firms that
experience a greater abnormal decrease in net debt issuance.
25
As in the analysis of net debt issuance, we create subgroups based on the status of firms at the end of
the second quarter of 2006 to make sure that our classification is not affected by the crisis. The next two
regressions immediately put to rest the possibility that firms with greater abnormal decrease in net debt
issuance issue more equity. The unrated firms have even lower net equity issuance during the crisis and
post-September 2008 than the other firms in the sample. The same is true for the smallest firms. In results
not reproduced in the table, we investigate large firms and investment grade firms using the same
regression framework. We find that the abnormal net equity issuance increase is concentrated among
these firms and is explained by a sharp drop in equity repurchases after September 2008. When we turn
to the KL financially constrained firms (estimates not reproduced), their net equity issuance falls more
during the crisis and falls further after September 2008. Turning to the subset of highly financially
constrained firms, we find that these firms do not have a significant crisis interaction, but their net equity
issuance drops more after September 2008 than for other firms. This pattern is the opposite from the one
we would expect with a bank credit supply shock. Firms with no debt have a negative crisis interaction,
but do not have a negative peak crisis interaction. The net equity issuance of firms with high cash is
abnormally low throughout the crisis and further decreases abnormally after September 2008. Finally, the
crisis interaction is positive for the highly levered firms, but they abnormally reduce net equity issuance
after September 2008. Firms with high short-term debt (regression estimates not reproduced) do not have
net equity issuance that differs from the remainder of the sample.
When we estimate our regressions for durables only, we find that equity issuance falls during the
crisis, but is higher after September 2008. As for the whole sample, however, bank-dependent firms do
not experience this increase in equity issuance because it is driven by firms that decrease repurchases.
It follows from the results in Panel B of Table 4 that for the subgroups which are more likely to be
bank dependent, net equity issuance falls before September 2008 even when net debt issuance does not.
Further, there is no evidence of a substitution effect towards equity from the firms that had the greatest
credit contraction.
26
Section 5. Did cash holdings fall?
In the presence of an unexpected credit supply contraction, we would expect firms to use their cash
balances to mitigate the adverse impact of the credit contraction. Paradoxically, for most types of firms,
cash ratios fall in the first year of the crisis when net debt issuance does not fall, but cash ratios increase
after September 2008 when net debt issuance falls sharply. In the first column of Panel A of Table 5, we
show the cash ratio for investment grade firms. The average ratio of cash to assets for these firms is not
significantly different in the last year of the boom from the first year of the crisis. However, the ratio
increases dramatically from 8.28% at the end of September 2008 to 9.47% at the end of March 2009.
Turning next to the smallest firms, we find in Panel A of Table 5 that their average ratio of cash to assets
falls from 27.85% to 26.97% from the last year of the credit boom to the first year of the credit crisis. The
drop is statistically significant. However, the cash ratio is 25.14% at the end of September 2008 and
increases to 25.72% at the end of the first quarter of 2009, a statistically significant increase. For large
firms (not reported), the drop in the first year of the crisis is insignificant, but the increase in the cash ratio
after September 2008 is significant and much larger, as the average cash ratio increases from 8.56% at the
end of the September to 9.80% at the end of March 2009. Unrated firms behave like the smallest firms,
while investment-grade firms behave like the largest firms. We also find a similar pattern for KL
financially constrained firms. For these firms, cash holdings average 34.95% in the last year of the boom,
34.10% in the first year of the crisis, and fall to 30.65% by the end of September 2008. However, they
increase to 34.17% by the end of January 2009. It should be noted, however, that for all other subgroups
discussed so far, there is no material difference in the change in the average cash ratio after September
2008 if we compare firms that belong to the same subgroup at the end of September 2008 and at the end
of March 2009 or follow the firms that belong to the subgroup in the third quarter of 2008 even if they do
not belong to that subgroup at the end of March 2009. This is not the case for the KL constrained firms in
Panel A. The change in the cash ratio of the KL constrained firms at the end of September 2008 is
essentially zero from then to the end of March 2008 while the average cash holdings of KL constrained
firms selected quarterly increase sharply. However, in neither case does the ratio of cash to assets fall.
27
An interesting group of firms is the group with no debt. We would expect the changes in credit
markets to have less impact on firms with no debt than on more typical firms. These firms have an
extremely high cash ratio of 42.07% at the end of the credit boom. The cash ratio of these firms falls to
38.32% at the end of the third quarter of 2008. However, these firms hoard cash after September 2008, as
their cash ratio increases to 40.27% by the end of our sample. We also want to examine a subset of firms
that did not draw down credit lines. For this purpose, we investigate firms that have no debt at the end of
the quarter (results not tabulated). Because these firms have no debt at the end of the quarter, they
typically have negative net debt issuance during the quarter. These firms experience a 1.19 percentage
point increase in the cash ratio after the events of September 2008. This increase is statistically significant
as well and represents a 15.38% increase in cash holdings. It follows that the increase in cash holdings is
not tied to credit line draw-downs.
Though we reproduce the cash ratio of firms with a cash ratio in the top quintile of cash ratios,
changes in the cash ratio for firms belonging to that quintile are not informative since firms are selected
based on their cash ratio. It is instructive to note, however, that the cash ratio of firms in the top quintile
of cash ratios is extremely high. In particular, firms in the top quintile of cash ratios have an average cash
ratio of 61.10% at the end of the credit boom. The firms that have a high cash ratio at the end of
September 2008 decrease it by 1.32% over the next two quarters. High leverage firms have a much lower
cash ratio. Their cash ratio stays remarkably stable throughout the crisis.
To understand better where the cash comes from for the firms that increase their cash holdings
sharply, we divide the sample into quintiles based on the change in cash in the last two quarters of the
sample period normalized by assets at the end of September 2008. We then compare the firms in the top
quintile of cash increases to the firms in the bottom quintile of cash increases. The results are striking.
The average operating cash flow to assets of the firms in the top quintile for the last two quarters is
5.46%; in contrast, it is -2.28% for the firms in the bottom quintile. The firms in the bottom quintile have
negative net debt issuance of -0.32% on average for these two quarters, while the firms in the top quintile
have positive net debt issuance of 0.47%. The firms in the top and bottom quintiles have similar capital
28
expenditures to assets, but the firms in the bottom quintile spend 3.14% of assets on R&D on average
across the two quarters, while the firms in the top quintile spend only 1.62%. We also compare the firm
characteristics for the fourth quarter of 2008 and the first quarter of 2009 to the last two identical quarters
of the boom, namely the fourth quarter of 2006 and the first quarter of 2007. Strikingly, capital
expenditures fall for firms in the bottom and top quintile, but they fall only by 0.24% for firms in the
bottom quintile and by 0.45% for firms in the top quintile – in other words, firms that hoard cash more
actually decrease capital expenditures more. Both types of firms have higher R&D expenditures to assets
in the last two quarters of the sample than in the fourth quarter of 2006 and the first quarter of 2007. Net
equity issuance falls sharply for firms in the bottom quintile, but is largely unchanged for firms in the top
quintile. The bottom line of this comparison is that firms that accumulate the most cash are firms that
have high operating cash flow, are still borrowing, and do not suffer a large drop in equity issuance.
We now turn to the estimation of a cash holdings model. Our model for the expected cash ratio is the
model used in Bates, Kahle, and Stulz (2009), but we estimate this model using quarterly data. This
model allows for a transaction demand for cash as well as a precautionary demand for cash. The
dependent variable is the cash ratio, and the following explanatory variables are used: the standard
deviation of cash flows at the industry level using two-digit SIC codes (positive coefficient), the marketto-book ratio (positive coefficient), the log of firm size (insignificant coefficient), the ratio of cash flow to
assets (positive coefficient), the ratio of net working capital to assets (negative coefficient), capital
expenditures to assets (negative coefficient), leverage (negative coefficient), R&D to assets (negative
coefficient), a dummy variable for missing data on R&D (insignificant), dividends to assets (positive
coefficient), acquisitions to assets (negative coefficient), net equity issuance to assets (positive
coefficient), and net debt issuance to assets (positive coefficient).14 The coefficient estimates are
consistent with the coefficient estimates in Bates, Kahle, and Stulz (2009). Because we use quarterly data,
we add to the model indicator variables for the second, third, and fourth quarters to accommodate
14
Not all firms report R&D on a quarterly basis. When a firm does not report R&D quarterly, we use the annual
R&D, divided by four, as an estimate of R&D in each quarter of that year. Results are similar if we set R&D as
missing for these firms.
29
seasonal effects. The indicator variables for the third and fourth quarters have negative significant
coefficients. As with the debt and equity issuance models, we allow for firm fixed effects and estimate the
model over the period 1995 to the end of the sample. We do not report the coefficients to save space.
We first estimate a regression for the whole sample. The cash ratio is significantly lower during the
crisis by 0.94%. However, the cash ratio increases abnormally after September 2008 by 0.59%. The
regression shows that abnormal cash holdings follow a u-shape pattern as they fall during the crisis but
increase after the events of September 2008. The next two regressions show that unrated and small firms
decrease their cash ratio more during the crisis than other firms. However, after September 2008, there is
no difference for the smallest or the unrated firms. Turning to KL financially constrained firms (not
tabulated), we find that the cash ratio of these firms does not evolve differently from the sample as a
whole. The subset of highly constrained firms actually has an abnormal increase in the cash ratio during
the crisis, but the abnormal increase after September 2008 in excess of the sample as a whole is not
statistically significant even though economically large. Firms with no debt experience an abnormal
decrease in cash during the crisis and a further abnormal decrease in cash after September 2008. Firms
with high cash holdings experience abnormal decreases in cash holdings in excess of the sample as well
and these decreases are economically large. Finally, firms with high leverage do not behave differently
from the other firms.
For robustness, we examine several other regressions. The firms with high short-term debt hold a
positive abnormal amount of cash during the crisis, but they do not increase cash holdings after
September 2008 abnormally. Though we do not reproduce the results, we also estimate the regressions
classifying firms each quarter. When we do so, the increase in cash holdings of firms with no debt and
firms with high leverage after September 2008 is not significantly different from the increase of the whole
sample. We also estimate the model for firms that have no debt at the end of a quarter, which are firms for
which credit line draw-downs are irrelevant. We find that these firms do not behave differently from the
firms that have no debt at the beginning of a quarter. Finally, firms in the durables industry exhibit the
30
same patterns as those documented for the sample as a whole. For these firms, cash falls during the crisis,
but increases after September 2008.
It follows from this discussion that firms reduce cash holdings abnormally before September 2008,
but after September 2008, most subgroups of firms either increase cash holdings abnormally or stop
consuming cash. The exceptions are firms with high cash holdings or no debt at the end of the second
quarter of 2006. However, the average cash ratio of firms that are most likely to be bank dependent does
not increase differently from the cash ratio of other firms after September 2008.
Section 6. Capital expenditures and the financial crisis
As shown in Panel A of Table 6, capital expenditures do not differ on average in the first year of the
crisis and the last year of the boom for any type of firm. Though we do not show the results in the table,
this is the case even for firms with high short-term debt holdings. However, after September 2008, capital
expenditures fall for all types of firms. The fall from September 2008 to the end of the sample period is
smallest for high cash firms in percent or percentage points, but it is important to note that these firms are
the firms that invest the least in the last year of the boom among the various subgroups we consider. The
fall is the largest for the KL financially constrained firms, but it is interesting to note that the ratio of
capital expenditures to assets of these firms has a higher value in the fourth quarter of 2008 than the
average quarterly value during the last year of the credit boom, so that all the decrease takes place in the
first quarter of 2009. The capital expenditure ratio of these firms in the third quarter of 2008 is actually
the highest since 2000, and their capital expenditure ratio in the second quarter of 2008 is the second
highest since 2000.
Though we do not tabulate the results, we also investigate the ratio of R&D expenditures to assets.
There is no difference between R&D in the last year of the boom and the first year of the crisis for any
type of firms. Not surprisingly, the firms with high cash holdings have extremely high R&D expenditures
as their quarterly average in the first year of the crisis is 4.57%. For all types of firms, R&D in the fourth
quarter of 2008 is higher than R&D in the last quarter of the boom. However, R&D does fall in the first
31
quarter of 2009. The drop in R&D is smaller in magnitude to the drop in capital expenditures, except for
the high cash firms. For the high cash firms, capital expenditures drop by 0.17% while R&D expenditures
drop by 0.43%. It is noteworthy that in the year before the crisis, the high cash firms had quarterly capital
expenditures of only 0.87%. In contrast, their quarterly R&D expenditures were 4.31%. In other words,
capital expenditures are low for high cash firms compared to R&D expenditures.
The credit supply shock hypothesis predicts that capital expenditures should fall the most for the most
bank-dependent firms. To examine this hypothesis, we estimate regressions for capital expenditures in
Panel B of Table 6. We regress capital expenditures to assets in period t on four lags of market-to-book
and four lags of operating cash flow. The regression model is the same as the one in Duchin et al. (2010),
except that we use several lags because we use quarterly data, as within-quarter adjustments seem
implausible. The lagged variables have significantly positive coefficients. We also use indicator variables
for the last three quarters of the year. All these indicator variables have positive and significant
coefficients. For brevity, we do not reproduce the coefficient estimates.
For the sample as a whole, capital expenditures are abnormally low during the crisis by 0.11% per
quarter. This finding is surprising since capital expenditures are roughly the same for the sample in the
last year of the boom and the first year of the crisis. However, because the stock market does not fall early
in the crisis, the model would predict higher capital expenditures. Capital expenditures become
significantly lower by 0.20% after September 2008. Turning to the smallest firms and the unrated firms,
they invest more during the crisis and their investment does not evolve differently from other firms after
September 2008. Further, capital expenditures for KL financially constrained firms (not tabulated) and for
highly financially constrained firms in June 2006 do not evolve differently from the sample as a whole.
The post-September 2008 shock is not different for firms that have no debt, even though these firms
invest more during the crisis. However, firms with a high cash ratio have abnormally higher investment
throughout the crisis compared to other firms, so that they do not experience a decrease in investment
after September 2008. This latter finding is different from Duchin et al. (2010) who find that firms with
more cash invest more only in the first year of the crisis, but a possible explanation is that they did not
32
have access to complete data for the first quarter of 2009. Finally, firms with high leverage invest more
during the crisis, but they experience a decrease in investment after September 2008 that is twice the size
of the decrease of other firms in the sample. Firms with high short-term debt do not experience abnormal
decreases in investment either for the crisis as a whole or after September 2008.
As we did in the earlier sections, we also estimate our regressions for firms in the durables goods
industry only. The patterns uncovered in this section hold for those firms as well. Bank-dependent firms
do not decrease capital expenditures more after September 2008 and firms with high cash holdings do not
decrease capital expenditures abnormally throughout the crisis. The only result that differs is that, while
firms with no debt where the attribution is made quarterly do not decrease capital expenditures more than
other firms after September 2008, firms that have no debt in June 2006 do so.
In summary, capital expenditures are abnormally low during the crisis and drop sharply after
September 2008. However, firms that are financially constrained before the crisis do not reduce capital
expenditures more. In fact, financially constrained firms have abnormally higher capital expenditures
during the crisis than other firms, but the abnormal decrease in their capital expenditures after September
2008 is insignificantly different from the sample as a whole. In contrast, firms that have stronger balance
sheets before the crisis or before September 2008 experience a lower reduction in capital expenditures. In
particular, firms that have high cash holdings or not debt in June 2006 do not reduce capital expenditures
abnormally after September 2008, firms that have no debt or an investment rating decrease capital
expenditures less, but firms with high debt decrease capital expenditures drastically.
Section 7. Conclusion
We examine financial and investment policies of firms in the last year of the credit boom and from
the start of the 2007 financial crisis to its peak to understand what these policies tell us about the financial
crisis. The conventional view of the crisis focuses on the impact on industrial firms of a bank credit
supply shock resulting from bank losses in so-called toxic assets. However, the crisis also reduced
33
demand for goods and increased risk. We explore the impact on financial and investment policies of these
three shocks.
If the bank credit supply shock is the dominant shock, we expect more bank-dependent firms to have
lower net debt issuance, a greater decrease in cash holdings, an increase in net equity issuance, and a
greater drop in capital expenditures than less bank-dependent firms. Net debt issuance increases before
September 2008 for most subsamples in our study, including for firms that are expected to depend on
banks for their borrowing. Net debt issuance falls sharply after September 2008, but not more for the
bank-dependent subsamples of firms. We find that net equity issuance decreases during the crisis and that
this decrease occurs before the decrease in net debt issuance. There is no evidence that net equity issuance
decreases less for more bank-dependent firms. Further, while the ratio of cash to assets decreases for
bank-dependent firms before September 2008, it either increases or stays constant in all other subsamples,
except for the subsample of firms in the top quintile of cash ratios. These firms have an average cash to
assets ratio at the peak of the credit boom of 61%, which would allow them to pay for their average
capital expenditures and R&D expenditures in the last year of the boom for several years if their operating
cash flow were zero and they could not raise any money. There is no evidence that more bank-dependent
firms reduce capital expenditures more during the crisis. None of these findings is consistent with a
dominant role in financing and investment policies of an exogenous bank credit contraction.
In this paper, we focus on the economic importance of a bank credit contraction for non-financial
public firms. Therefore, we have nothing to say about the impact of reduced credit availability for
consumers. Such reduced credit availability would affect firms through the demand shock in our analysis.
We find asset-weighted as well as equally-weighted net debt issuance to be positive in every quarter in
2008; the flow-of-funds data has business credit growing in every quarter of 2008 as well. However, in
the flow-of-funds data, household debt falls in 2008 for the first time since 1977, the first year the data is
available. Further, we do not have data for private firms. This lack of data on private firms may explain
why new loans fall in 2008 compared to 2007 according to the flow-of-funds data, but they do not in our
data. According to Dealogic, there were leveraged buyouts totaling $375.1 billion in 2007, but in 2008
34
leveraged buyouts were only for $61.1 billion.15 The dramatic drop in leveraged buyouts represents a
sharp decrease in borrowing that undoubtedly explains part of the drop in the amount of new loans in
2008 but does not show up in our data as we focus on public companies.
In contrast to the prevalent narrative of the crisis which focuses on the dominant role of the bank
credit supply shock, we show that net equity issuance decreases materially early in the crisis, before net
debt issuance falls, and is unusually low throughout the crisis. The cumulative decline in financing cash
flows during the crisis from reduced net equity issuance exceeds the cumulative loss from decreased net
debt issuance for small and unrated firms, which are firms that we would expect to rely on banks for their
borrowing. The decrease in net equity issuance is consistent with a decrease in expected cash flows and a
loss of investment opportunities, and with an increase in the cost of equity as investors become more risk
averse. The cash hoarding we document after September 2008 for most types of firms (except those with
the highest ratio of cash to assets) is consistent with a greater demand for precautionary cash holdings
resulting from an increase in risk, including an increase in risks associated with the provision of capital to
firms. As firms face a loss in future expected cash flows, so that their net worth falls and their prospects
dim, we would expect them to find it more expensive and difficult to borrow, as emphasized by theories
of endogenous contraction of credit. In support of the latter theories, we find that capital expenditures
decrease less for firms that are financially stronger before the crisis and more for firms that are highly
levered before the crisis. Our findings strongly suggest a first-order impact on corporations of the
decrease in demand and the increase in risk that took place during the crisis.
15
See “Final 2008 data confirm: ‘Little but tumbleweeds’ in LBO arena,” Deal Journal, January 7, 2009, The Wall
Street Journal.
35
References
Almeida, Heitor, Murillo Campello, and Michael S. Weisbach, 2004, The cash flow sensitivity of cash,
Journal of Finance 59, 1777-1804.
Almeida, Heitor, Murillo Campello, Bruno Laranjeira and Scott Weisbenner, 2009, Corporate debt
maturity and the real effects of the 2007 financial crisis, working paper, University of Illinois, Ill.
Baker, Malcolm, 2009, Capital market-driven corporate finance, Annual Review of Financial Economics
1, 181-205.
Bates, Thomas, Kathleen Kahle, and René Stulz, 2009, Why do U.S. firms hold so much more cash than
they used to? Journal of Finance 64, 1985-2022.
Becker, Bo, and Victoria Ivashina, 2010, Cyclicality of credit supply: Firm level evidence, unpublished
paper, Harvard Business School, MA.
Bernanke, Ben S., 1983, Nonmonetary effects of the financial crisis in the propagation of the Great
Depression, The American Economic Review 73, 257-276.
Bernanke, Ben S., 2007, The financial accelerator and the credit channel, Federal Reserve Board,
Washington, D.C.
Bernanke, Ben, and Mark Gertler, 1989, Agency costs, net worth, and business fluctuations, American
Economic Review 79, 14-31.
Boyson, Nicole, Jean Helwege, and Jan Jindra, 2010, Crises, liquidity shocks, and fire sales at financial
institutions, working paper, University of South Carolina.
Brunnermeier, Markus K., 2009, Deciphering the liquidity and credit crunch 2007-2008, Journal of
Economic Perspectives 23, 77-100.
Caballero, Ricardo J., and Arvind Krishnamurthy, 2008, Collective risk management in a flight to quality
episode, Journal of Finance 63, 2195-2230.
Campbell, John Y., Stefano Giglio, and Christopher Polk, 2010, Hard times, working paper, National
Bureau of Economic Research, Cambridge, MA.
Campello, Murillo, Erasmo Giambona, John R. Graham, and Campbell R. Harvey, 2009, Liquidity
management and corporate investment during a financial crisis, working paper, University of Illinois,
Ill.
Campello, Murillo, John R. Graham, and Campbell R. Harvey, 2010, The real effects of financial
constraints: Evidence from a financial crisis, Journal of Financial Economics, forthcoming.
Chari, V.V., Lawrence J. Christiano, and Patrick J. Kehoe, 2008, Facts and Myths about the Financial
Crisis of 2008, working paper, Federal Reserve Bank of Minnesota.
DeAngelo, Harry, Linda DeAngelo, and René M. Stulz, 2010, Seasoned equity offerings, market timing,
and the corporate lifecycle, Journal of Financial Economics 95, 275-295.
36
Duchin, Ran, Oguzhan Ozbas, and Berk A. Sensoy, 2010, Costly external finance, corporate investment,
and the subprime mortgage financial crisis, Journal of Financial Economics, forthcoming.
Durand, Robert B., Dominic Lim, and J. Kenton Zumwalt, 2007, Fear and the Fama-French factors,
working paper, Colorado State University.
Erel, Isil, Brandon Julio, Woojin Kim, and Michael S. Weisbach, 2010, Macroeconomic conditions and
the structure of securities, working paper, The Ohio State University, OH.
Fama, Eugene F. and Kenneth R. French, 2005, Financing decisions: Who issues stock?, Journal of
Financial Economics 76, 549-582.
Fama, Eugene F. and Kenneth R. French, 2008, Opportunistic financing, working paper, University of
Chicago.
Faulkender, Michael and Mitchell A. Petersen, 2006, Does the source of capital affect capital structure?,
Review of Financial Studies 19, 45-79.
Gao, Pengjie, and Hayong Yun, 2009, Commercial paper, lines of credit, and the real effects of the
financial crisis of 2008: Firm-level evidence from the manufacturing industry, working paper,
University of Notre Dame, Notre Dame, IN.
Gertler, Mark and Simon Gilchrist, 1994, Monetary policy, business cycles, and the behavior of small
manufacturing firms, The Quarterly Journal of Economics 109(2), 309-40.
Gorton, Gary, 2010, Slapped by the invisible hand: The panic of 2007, Oxford University Press, Oxford.
Hamilton, James D., 2009, Causes and consequences of the oil shock of 2007-08, Brookings Papers on
Economic Activity, 215-259.
Ivashina, Victoria, and David Scharfstein, 2010a, Bank lending during the financial crisis of 2008,
Journal of Financial Economics 97, 319-338.
Ivashina, Victoria, and David Scharfstein, 2010b, Liquidity management in the financial crisis, working
paper, Harvard Business School.
Iyer, Rajkamal, Samuel Lopes, José-Luis Peydró, and Antoinette Schoar, 2010, Interbank liquidity crunch
and the firm credit crunch: Evidence from the 2007-2009 crisis, unpublished, MIT, Cambridge, MA.
Kashyap, Anil K., Owen A. Lamont and Jeremy C. Stein, 1994, Credit conditions and the cyclical
behavior of inventories, The Quarterly Journal of Economics 109(3), 565-92.
Kashyap, Anil K., Jeremy C. Stein and David W. Wilcox, 1993, Monetary policy and credit conditions:
Evidence from the composition of external finance, American Economic Review 83, 78-98.
Kiyotaki, Nobuhiro and John Moore, 1997, Credit cycles, Journal of Political Economy 105, 211-248.
Korajczyk, Robert A. and A. Levi, 2003, Capital structure choice: macroeconomic conditions and
financial constraints, Journal of Financial Economics 68, 75-109.
37
Kuppuswamy, Venkat, and Belen Villalonga, 2010, Does diversification create value in the presence of
external financial constraints? Evidence from the 2008-2009 financial crisis, Harvard Business
School working paper, Boston, MA.
Leamer, Edward E., 2007, Housing is the business cycle, NBER working paper, NBER, Cambridge, MA.
Leary, Mark T., 2009, Bank loan supply, lender choice, and corporate capital structure, Journal of Finance
64(3), 1143-1185.
Leary, Mark T., and Michael R. Roberts, 2005, Do firms rebalance their capital structure?, Journal of
Finance 60, 2575-2619.
Lemmon, Michael and Michael R. Roberts, 2010, The response of corporate financing and investment to
changes in the supply of credit, Journal of Financial and Quantitative Analysis 43, 555-587.
Miller, Merton H., 1963, The corporate income tax and corporate financial policies, Cowles Foundation
Paper 210b.
Minton, Bernadette A., and Catherine Schrand, 1999, The impact of cash flow volatility on discretionary
investment and the costs of debt and equity financing, Journal of Financial Economics 54, 423-460.
Myers, Stewart, 1977, Determinants of corporate borrowing, Journal of Financial Economics 5, 147-175.
Opler, Tim, Lee Pinkowitz, René M. Stulz, and Rohan Williamson, 1999, The determinants and
implications of corporate cash holdings, Journal of Financial Economics 52, 3-46.
Pastor, Lubos, and Pietro Veronesi, 2010, Uncertainty about government policy and stock prices, working
paper, University of Chicago.
Shleifer, Andrei, and Robert Vishny, 2010, Unstable banking, Journal of Financial Economics 97, 306318.
Sufi, Amir, 2009, Bank lines of credit in corporate finance: An empirical analysis, Review of Financial
Studies 22, 1057-1088.
38
Table 1: Financial Policies Statistics During the Credit Boom and the Financial Crisis
Panel A: Financial Policies Using Asset-weighted Averages
This table examines the time series of firm financial policy variables, on an asset-weighted basis, from the first
quarter of 2005 to the first quarter of 2009. We begin with quarterly data collected from the CRSP/Compustat
Merged (CCM) Fundamentals Quarterly database for 1980-2009. We delete observations with negative total assets
(ATQ), negative sales (SALEQ), negative cash and marketable securities (CHQ), cash and marketable securities
greater than total assets, and firms not incorporated in the U.S. We also eliminate all financial firms, which we
define as firms with SIC codes between 6000 and 6999, and utilities, which we define as firms with SIC codes
between 4900 and 4949. Other variable definitions are provided in the Appendix. To examine the significance of
the time series changes we use two approaches. In the seasonality-adjusted p-values, we compare the change of
interest to changes over identical quarterly calendar periods to account for seasonality. For the Newey-West pvalues, we use all two-quarter changes but rely on Newey-West t-statistics to account for overlap.
Quarter
Cash to
assets
Net longterm debt
issuance
Net total
debt
issuance
Net equity
issuance
Capex
Operating
cash flow
Avg 1983-2004
0.0675
0.0036
0.0065
-0.0009
0.0183
0.0451
Avg 1990-2004
0.0666
0.0036
0.0056
-0.0008
0.0171
0.0454
Min
0.0486
-0.0057
-0.0164
-0.0076
0.0097
0.0204
Max
0.1077
0.0144
0.0460
0.0061
0.0295
0.0608
Std. Dev.
0.0149
0.0034
0.0077
0.0021
0.0040
0.0076
0.0924
0.0069
0.0090
-0.0089
0.0146
0.0430
Crisis vs pre crisis
Avg (2007Q3-2008Q2)
Avg (2006Q3-2007Q2)
0.0956
0.0055
0.0073
-0.0084
0.0144
0.0454
-0.0032
0.0014
0.0017
-0.0005
0.0002
-0.0024
seasonality adjusted p-values
0.6484
0.5044
0.7192
0.8116
0.9365
0.6943
2009Q1
0.1018
0.0077
-0.0019
-0.0019
0.0111
0.0236
2008Q4
0.0981
0.0035
0.0009
-0.0028
0.0156
0.0484
2008Q3
0.0889
0.0061
0.0043
-0.0065
0.0155
0.0442
2007Q2
0.0949
0.0077
0.0087
-0.0110
0.0143
0.0458
Difference (2009Q1 - 2008Q3)
0.0130
0.0016
-0.0062
0.0046
-0.0044
-0.0206
Newey West test p-values
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
Seasonality adjusted p-values
0.0433
0.7082
0.2423
0.0758
0.0842
0.0274
Difference (2009Q1 - 2007Q2)
0.0070
0.0000
-0.0106
0.0091
-0.0032
-0.0222
Newey West test p-values
0.0000
0.6400
0.0000
0.0000
0.0000
0.0000
Seasonality adjusted p-values
0.4982
0.9995
0.1844
0.0009
0.3788
0.0234
Difference
39
Table 1, Panel B: Financial Policies Using Equally-Weighted Averages
This table examines the means of the time series of firm financial policy variables, on an equal-weighted basis, from
the first quarter of 2005 to the first quarter of 2009. We begin with quarterly data collected from the
CRSP/Compustat Merged (CCM) Fundamentals Quarterly database for 1980-2009. We delete observations with
negative total assets (ATQ), negative sales (SALEQ), negative cash and marketable securities (CHQ), cash and
marketable securities greater than total assets, and firms not incorporated in the U.S. We also eliminate all financial
firms, which we define as firms with SIC codes between 6000 and 6999, and utilities, which we define as firms with
SIC codes between 4900 and 4949. Other variable definitions are provided in the Appendix. We report p-values for
paired t-tests using the firms that exist in both quarters when examining the significance of the difference between
the quarters of interest.
Quarter
Cash to
assets
Net longterm debt
issuance
Net total
debt
issuance
Net equity
issuance
Capex
Operating
cash flow
Avg 1983-2004
0.1605
0.0049
0.0086
0.0125
0.0184
0.0272
Avg 1990-2004
0.1721
0.0044
0.0074
0.0138
0.0172
0.0288
Min
0.1215
-0.0020
-0.0013
0.0027
0.0105
0.0120
Max
0.2316
0.0144
0.0197
0.0354
0.0275
0.0426
Std. Dev.
0.0319
0.0034
0.0050
0.0055
0.0035
0.0070
Avg (2007Q3-2008Q2)
0.2105
0.0066
0.0086
0.0050
0.0148
0.0318
Avg (2006Q3-2007Q2)
0.2198
0.0065
0.0082
0.0105
0.0147
0.0317
-0.0092
0.0001
0.0004
-0.0055
0.0001
0.0001
Ttest pre vs post crisis
0.0034
0.8549
0.6294
0.0001
0.7855
0.9726
2009Q1
0.2015
-0.0015
-0.0045
0.0030
0.0097
0.0131
2008Q4
0.2008
-0.0002
-0.0004
-0.0000
0.0131
0.0305
2008Q3
0.1957
0.0045
0.0081
0.0021
0.0144
0.0320
2007Q2
0.2193
0.0079
0.0105
0.0142
0.0150
0.0301
Diff. (2009Q1 - 2008Q3)
0.0058
-0.0060
-0.0126
0.0010
-0.0046
-0.0189
Paired Diff.
0.0070
-0.0057
-0.0124
0.0013
-0.0049
-0.0203
Ttest 2008Q3=2009Q1
0.0001
0.0001
0.0001
0.1426
0.0001
0.0001
Diff. (2009Q1 - 2007Q2)
-0.0178
-0.0094
-0.0150
-0.0112
-0.0053
-0.0171
Paired Diff.
-0.0196
-0.0096
-0.0163
-0.0109
-0.0061
-0.0202
0.0001
0.0001
0.0001
0.0001
0.0001
0.0001
Crisis vs pre crisis
Difference
Ttest 2007Q2=2009Q1
40
Table 2: Summary Statistics on Lines of Credit
This table examines data on lines of credit for a random sample of 300 firms chosen as of the second quarter of 2007. We follow the approach in
Sufi (2009) in sampling the firms. % new drawdown is the percentage of firms in that quarter that draw down a line of credit. Drawdown to loc is
the ratio of the amount drawn down to the firm’s total line of credit. Loc/assets is the ratio of lines of credit to total assets. Drawdown/assets is
the ratio of new draw-downs in that quarter to total assets.
Quarter
2007Q1
2007Q2
2007Q3
2007Q4
2008Q1
2008Q2
2008Q3
2008Q4
2009Q1
Nobs.
283
300
293
279
277
265
262
252
246
% Nobs
with loc
0.806
0.810
0.823
0.817
0.830
0.834
0.828
0.833
0.829
For obs. with loc > 0:
new
% new
drawdown
drawdown
to loc
0.2719
0.0387
0.3210
0.0428
0.2780
0.0409
0.3728
0.0784
0.3478
0.0660
0.2760
0.0311
0.3134
0.0491
0.3238
0.0527
0.2892
0.0285
Aggregate sample
new
loc/
drawdown/
assets
assets
0.0916
0.0037
0.0953
0.0031
0.0945
0.0032
0.0973
0.0128
0.0951
0.0050
0.0883
0.0019
0.0913
0.0057
0.0997
0.0048
0.1003
0.0016
total
drawdown
to loc
0.2533
0.2537
0.2533
0.2842
0.3078
0.2943
0.3015
0.3117
0.3273
41
Equal-weighted sample
new
loc/
drawdown/
assets
assets
0.1456
0.0066
0.1502
0.0093
0.1463
0.0079
0.1583
0.0172
0.1556
0.0109
0.1573
0.0056
0.1601
0.0086
0.1754
0.0103
0.1651
0.0066
Table 3: Net total debt issuance
Panel A examines time series data for subgroups determined at the end of each previous quarter. We divide firms quarterly into firms with an investment grade
rating, a speculative rating, and no rating using the S&P long-term rating (splticrm) available on Compustat. Size quintiles are formed quarterly by dividing all
NYSE firms into five quintiles based on assets; we then assign the non-NYSE firms to these quintiles. We define a firm to be KL financially constrained if it (1)
does not pay dividends, (2) does not have net equity repurchases, (3) does not have a credit rating, and (4) has a Tobin’s q greater than one (defined as the market
value of the assets divided by the book value, where market value of assets is book value minus book equity plus market value of equity). No-debt firms are
firms whose leverage ratio is zero at the end of the previous quarter. Hicash (hilev) firms are firms that are in the top quintile of cash to assets (leverage) at the
end of the previous quarter. Panel B shows regression estimates. The crisis indicator variable takes value 1 in each quarter after Q2 of 2007. The crisis peak
indicator variable takes value 1 for the last two quarters of the sample. Except in the first regression, we interact the crisis indicator variable and the crisis peak
indicator variable with a subgroup indicator variable and include the subgroup indicator variable in the regression as well. The subgroups are formed at the end of
June 2006 and are (1) investment-rated firms, (2) unrated firms, (3) small firms, namely firms whose asset size is below the top of the bottom quintile of NYSE
firms, (4) highly constrained firms, which are KL constrained firms in June 2006 that had positive net issuance in the four quarters ending in June 2006, (5) no
debt firms, (6) high cash firms, and (7) high leverage firms. The regressions use control variables that are not reported. These variables are listed in the text and
their definition is provided in the Appendix.
42
Panel A: Net Total Debt Issuance
Investment
Grade
Unrated
Smallest
KL
Constrained
No Debt
High Cash
High Lev
0.0074
-0.0023
0.0204
0.0055
0.0082
-0.0015
0.0171
0.0049
0.0092
-0.0011
0.0194
0.0052
0.0134
0.0019
0.0279
0.0061
0.0102
0.0039
0.0175
0.0033
0.0063
-0.0003
0.0196
0.0039
0.0012
-0.0175
0.0203
0.0080
Crisis vs pre crisis
Avg (2007Q3-2008Q2)
Avg (2006Q3-2007Q2)
Difference
Ttest
0.0109
0.0060
0.0049
0.0180
0.0083
0.0069
0.0014
0.1540
0.0083
0.0083
0.0000
0.9997
0.0097
0.0075
0.0023
0.2804
0.0073
0.0080
-0.0008
0.5592
0.0020
0.0058
-0.0038
0.0182
0.0069
0.0088
-0.0019
0.4442
2009Q1
2008Q4
2008Q3
2007Q2
0.0001
0.0029
0.0065
0.0069
-0.0035
-0.0005
0.0087
0.0089
-0.0025
-0.0007
0.0086
0.0104
-0.0010
0.0026
0.0109
0.0082
0.0036
0.0071
0.0065
0.0082
-0.0058
-0.0009
0.0000
0.0076
-0.0175
-0.0153
0.0001
0.0168
Diff. (2009Q1 - 2008Q3)
Paired Diff. (2009Q1 - 2008Q3)
Ttest 2008Q3=2009Q1*
-0.0064
-0.0067
0.0545
-0.0122
-0.0117
0.0001
-0.0111
-0.0108
0.0001
-0.0120
-0.0150
0.0001
-0.0029
-0.0050
0.0086
-0.0059
-0.0030
0.2645
-0.0176
-0.0169
0.0001
Diff. (2009Q1 - 2007Q2)
Paired Diff. (2009Q1 - 2007Q2)
Ttest 2007Q2=2009Q1
-0.0068
-0.0068
0.0079
-0.0124
-0.0134
0.0001
-0.0129
-0.0143
0.0001
-0.0093
-0.0100
0.0019
-0.0046
-0.0073
0.0060
-0.0135
-0.0107
0.0028
-0.0343
-0.0327
0.0001
Avg 1983-2004
Min
Max
Std. Dev.
43
Panel B: Net Total Debt Issuance Regressions (splits as of 2006Q2) Whole
sample
crisis
Unrated
Smallest
Highly
Constrained
No Debt
Hicash
Hilev
0.0274***
(0.000)
Yes
-0.0018*
(0.081)
0.0005
(0.458)
0.0039***
(0.002)
-0.0126***
(0.000)
0.0043**
(0.015)
0.0273***
(0.000)
Yes
0.0003
(0.733)
0.0010
(0.194)
0.0008
(0.497)
-0.0110***
(0.000)
0.0025
(0.139)
0.0274***
(0.000)
Yes
0.0012**
(0.040)
0.0066***
(0.005)
-0.0059**
(0.026)
-0.0098***
(0.000)
0.0031
(0.387)
0.0277***
(0.000)
Yes
0.0006
(0.351)
0.0007
(0.440)
0.0025*
(0.058)
-0.0115***
(0.000)
0.0097***
(0.000)
0.0272***
(0.000)
Yes
0.0012*
(0.069)
0.0045***
(0.001)
-0.0027*
(0.084)
-0.0112***
(0.000)
0.0100***
(0.000)
0.0275***
(0.000)
Yes
0.0027***
(0.000)
0.0030*
(0.087)
-0.0101***
(0.000)
-0.0072***
(0.000)
-0.0128***
(0.000)
0.0276***
(0.000)
Yes
182,604
7,984
0.044
180,756
7,633
0.044
180,756
7,633
0.043
182,604
7,984
0.044
180,756
7,633
0.044
180,756
7,633
0.044
180,756
7,633
0.044
0.0011*
(0.051)
split dummy
crisis X split dummy
post-Sept 2008
-0.0096***
(0.000)
post-Sept X split dummy
Constant
Fixed Effects
Observations
Number of gvkey
Adjusted R-squared
Robust pvalues in parentheses
*** p<0.01, ** p<0.05, * p<0.1
44
Table 4: Net equity issuance
Panel A examines time series data for subgroups determined at the end of each previous quarter. We divide firms quarterly into firms with an investment grade
rating, a speculative rating, and no rating using the S&P long-term rating (splticrm) available on Compustat. Size quintiles are formed quarterly by dividing all
NYSE firms into five quintiles based on assets; we then assign the non-NYSE firms to these quintiles. We define a firm to be KL financially constrained if it (1)
does not pay dividends, (2) does not have net equity repurchases, (3) does not have a credit rating, and (4) has a Tobin’s q greater than one (defined as the market
value of the assets divided by the book value, where market value of assets is book value minus book equity plus market value of equity). No-debt firms are
firms whose leverage ratio is zero at the end of the previous quarter. Hicash (hilev) firms are firms that are in the top quintile of cash to assets (leverage) at the
end of the previous quarter. Panel B shows regression estimates. The crisis indicator variable takes value 1 in each quarter after Q2 of 2007. The crisis peak
indicator variable takes value 1 for the last two quarters of the sample. Except in the first regression, we interact the crisis indicator variable and the crisis peak
indicator variable with a subgroup indicator variable and include the subgroup indicator variable in the regression as well. The subgroups are formed at the end of
June 2006 and are (1) investment-rated firms, (2) unrated firms, (3) small firms, namely firms whose asset size is below the top of the bottom quintile of NYSE
firms, (4) highly constrained firms, which are KL constrained firms in June 2006 that had positive net issuance in the four quarters ending in June 2006, (5) no
debt firms, (6) high cash firms, and (7) high leverage firms. The regressions use control variables that are not reported. These variables are listed in the text and
their definition is provided in the Appendix.
45
Panel A: Net Equity Issuance (splits calculated quarterly)
Investment
Grade
Unrated
Smallest
KL
Constrained
No Debt
High Cash
High Lev
Avg 1983-2004
Min
Max
Std. Dev.
-0.0019
-0.0090
0.0042
0.0024
0.0148
0.0041
0.0429
0.0067
0.0180
0.0055
0.0487
0.0075
0.0254
0.0096
0.0686
0.0097
0.0150
-0.0003
0.0552
0.0088
0.0215
0.0055
0.0908
0.0134
0.0133
0.0040
0.0261
0.0045
Crisis vs pre crisis
Avg (2007Q3-2008Q2)
Avg (2006Q3-2007Q2)
Difference
Ttest
-0.0106
-0.0097
-0.0009
0.2234
0.0080
0.0152
-0.0071
0.0001
0.0112
0.0190
-0.0079
0.0001
0.0169
0.0247
-0.0078
0.0028
0.0045
0.0143
-0.0098
0.0001
0.0159
0.0313
-0.0154
0.0001
0.0081
0.0116
-0.0035
0.0571
2009Q1
2008Q4
2008Q3
2007Q2
-0.0014
-0.0048
-0.0084
-0.0106
0.0042
0.0005
0.0040
0.0204
0.0052
0.0016
0.0054
0.0254
0.0150
0.0054
0.0105
0.0306
0.0031
-0.0054
-0.0007
0.0205
0.0120
0.0019
0.0091
0.0415
0.0056
0.0061
0.0049
0.0125
Diff. (2009Q1 - 2008Q3)
Paired Diff. (2009Q1 - 2008Q3)
Ttest 2008Q3=2009Q1*
0.0069
0.0072
0.0001
0.0002
0.0006
0.6170
-0.0002
0.0000
0.9880
0.0045
-0.0029
0.3411
0.0038
0.0037
0.1136
0.0029
0.0058
0.1131
0.0007
0.0005
0.8290
Diff. (2009Q1 - 2007Q2)
Paired Diff. (2009Q1 - 2007Q2)
Ttest 2007Q2=2009Q1
0.0092
0.0093
0.0001
-0.0162
-0.0162
0.0001
-0.0202
-0.0214
0.0001
-0.0156
-0.0285
0.0001
-0.0174
-0.0177
0.0004
-0.0296
-0.0297
0.0001
-0.0069
-0.0068
0.0641
46
Panel B: Net Equity Issuance Regressions (splits as of 2006Q2)
Whole
sample
crisis
-0.0026***
(0.000)
split dummy
crisis X split dummy
post-Sept 2008
0.0013**
(0.033)
Post-Sept X split dummy
Constant
Fixed Effects
Observations
Number of gvkey
Adjusted R-squared
0.0126***
(0.000)
Yes
184,186
8,077
0.071
Unrated
Smallest
0.0004
(0.550)
-0.0007
(0.306)
-0.0039***
(0.000)
0.0056***
(0.000)
-0.0054***
(0.000)
0.0124***
(0.000)
Yes
-0.0008
(0.148)
-0.0005
(0.525)
-0.0031***
(0.002)
0.0071***
(0.000)
-0.0090***
(0.000)
0.0126***
(0.000)
Yes
182,331
7,722
0.072
182,331
7,722
0.072
Highly
Constrained
-0.0025***
(0.000)
-0.0018
(0.363)
0.0009
(0.716)
0.0017***
(0.007)
-0.0051**
(0.017)
0.0125***
(0.000)
Yes
184,186
8,077
0.071
Robust pvalues in parentheses
*** p<0.01, ** p<0.05, * p<0.1
47
No Debt
Hicash
Hilev
-0.0018***
(0.000)
-0.0024*
(0.066)
-0.0029*
(0.063)
0.0021***
(0.001)
-0.0013
(0.376)
0.0124***
(0.000)
Yes
-0.0015***
(0.001)
0.0023
(0.224)
-0.0100***
(0.000)
0.0024***
(0.000)
-0.0043**
(0.042)
0.0127***
(0.000)
Yes
-0.0030***
(0.000)
-0.0015
(0.200)
0.0025*
(0.063)
0.0023***
(0.000)
-0.0027*
(0.061)
0.0123***
(0.000)
Yes
182,331
7,722
0.072
182,331
7,722
0.072
182,331
7,722
0.071
Table 5: Cash to Assets
Panel A examines time series data for subgroups determined at the end of each previous quarter. We divide firms quarterly into firms with an investment grade
rating, a speculative rating, and no rating using the S&P long-term rating (splticrm) available on Compustat. Size quintiles are formed quarterly by dividing all
NYSE firms into five quintiles based on assets; we then assign the non-NYSE firms to these quintiles. We define a firm to be KL financially constrained if it (1)
does not pay dividends, (2) does not have net equity repurchases, (3) does not have a credit rating, and (4) has a Tobin’s q greater than one (defined as the market
value of the assets divided by the book value, where market value of assets is book value minus book equity plus market value of equity). No-debt firms are
firms whose leverage ratio is zero at the end of the previous quarter. Hicash (hilev) firms are firms that are in the top quintile of cash to assets (leverage) at the
end of the previous quarter. Panel B shows regression estimates. The crisis indicator variable takes value 1 in each quarter after Q2 of 2007. The crisis peak
indicator variable takes value 1 for the last two quarters of the sample. Except in the first regression, we interact the crisis indicator variable and the crisis peak
indicator variable with a subgroup indicator variable and include the subgroup indicator variable in the regression as well. The subgroups are formed at the end of
June 2006 and are (1) investment-rated firms, (2) unrated firms, (3) small firms, namely firms whose asset size is below the top of the bottom quintile of NYSE
firms, (4) highly constrained firms, which are KL constrained firms in June 2006 that had positive net issuance in the four quarters ending in June 2006, (5) no
debt firms, (6) high cash firms, and (7) high leverage firms. The regressions use control variables that are not reported. These variables are listed in the text and
their definition is provided in the Appendix.
48
Panel A: Cash to Assets (splits calculated quarterly)
Investment
Grade
Unrated
Smallest
KL
Constrained
No Debt
High Cash
High Lev
0.0594
0.0450
0.0924
0.0110
0.1801
0.1215
0.2755
0.0441
0.1946
0.1430
0.2857
0.0412
0.2306
0.1599
0.3686
0.0610
0.3734
0.3264
0.4450
0.0337
0.4700
0.3621
0.6160
0.0850
0.0721
0.0499
0.1284
0.0179
0.0810
0.0821
-0.0011
0.7722
0.2562
0.2671
-0.0109
0.0147
0.2697
0.2785
-0.0087
0.0833
0.3410
0.3495
-0.0085
0.4133
0.4020
0.4207
-0.0187
0.1243
0.5904
0.6077
-0.0173
0.2079
0.1161
0.1156
0.0005
0.9224
2009Q1
2008Q4
2008Q3
2007Q2
0.0947
0.0868
0.0828
0.0804
0.2414
0.2417
0.2370
0.2677
0.2572
0.2581
0.2514
0.2785
0.3417
0.3058
0.3065
0.3379
0.4027
0.3936
0.3832
0.4195
0.5586
0.5612
0.5628
0.6110
0.1117
0.1181
0.1151
0.1130
Diff. (2009Q1 - 2008Q3)
Paired Diff. (2009Q1 - 2008Q3)
Ttest 2008Q3=2009Q1*
0.0119
0.0105
0.0001
0.0044
0.0060
0.0006
0.0058
0.0050
0.0139
0.0352
0.0020
0.6142
0.0195
0.0126
0.0022
-0.0042
-0.0132
0.0050
-0.0034
0.0024
0.3919
Diff. (2009Q1 - 2007Q2)
Paired Diff. (2009Q1 - 2007Q2)
Ttest 2007Q2=2009Q1
0.0143
0.0128
0.0021
-0.0263
-0.0292
0.0001
-0.0213
-0.0316
0.0001
0.0038
-0.0444
0.0001
-0.0167
-0.0501
0.0001
-0.0524
-0.1104
0.0001
-0.0013
-0.0057
0.2300
Avg 1983-2004
Min
Max
Std. Dev.
Crisis vs pre crisis
Avg (2007Q3-2008Q2)
Avg (2006Q3-2007Q2)
Difference
Ttest
49
Panel B: Cash to Assets Regressions (splits as of 2006Q2)
Whole
sample
crisis
-0.0094***
(0.000)
split dummy
crisis X split dummy
post-Sept 2008
0.0059***
(0.003)
Post-Sept X split dummy
Constant
Fixed Effects
Observations
Number of gvkey
Adjusted R-squared
0.2558***
(0.000)
Yes
182,301
7,907
0.144
Unrated
Smallest
0.0035
(0.158)
-0.0041
(0.110)
-0.0148***
(0.000)
0.0102***
(0.000)
-0.0016
(0.651)
0.2543***
(0.000)
Yes
-0.0026
(0.290)
-0.0037
(0.205)
-0.0083**
(0.015)
0.0119***
(0.000)
-0.0047
(0.198)
0.2561***
(0.000)
Yes
180,184
7,544
0.144
180,184
7,544
0.144
Highly
Constrained
-0.0094***
(0.000)
-0.0160**
(0.029)
0.0117**
(0.037)
0.0053**
(0.012)
0.0093
(0.141)
0.2536***
(0.000)
Yes
182,301
7,907
0.145
Robust pvalues in parentheses
*** p<0.01, ** p<0.05, * p<0.1
50
No Debt
Hicash
Hilev
-0.0076***
(0.000)
0.0125**
(0.031)
-0.0172***
(0.001)
0.0112***
(0.000)
-0.0110*
(0.076)
0.2590***
(0.000)
Yes
-0.0081***
(0.000)
0.0706***
(0.000)
-0.0552***
(0.000)
0.0133***
(0.000)
-0.0301***
(0.000)
0.2655***
(0.000)
Yes
-0.0118***
(0.000)
0.0143***
(0.000)
0.0035
(0.354)
0.0099***
(0.000)
-0.0049
(0.236)
0.2602***
(0.000)
Yes
180,184
7,544
0.144
180,184
7,544
0.149
180,184
7,544
0.144
Table 6: Capital Expenditures
Panel A examines time series data for subgroups determined at the end of each previous quarter. We divide firms quarterly into firms with an investment grade
rating, a speculative rating, and no rating using the S&P long-term rating (splticrm) available on Compustat. Size quintiles are formed quarterly by dividing all
NYSE firms into five quintiles based on assets; we then assign the non-NYSE firms to these quintiles. We define a firm to be KL financially constrained if it (1)
does not pay dividends, (2) does not have net equity repurchases, (3) does not have a credit rating, and (4) has a Tobin’s q greater than one (defined as the market
value of the assets divided by the book value, where market value of assets is book value minus book equity plus market value of equity). No-debt firms are
firms whose leverage ratio is zero at the end of the previous quarter. Hicash (hilev) firms are firms that are in the top quintile of cash to assets (leverage) at the
end of the previous quarter. Panel B shows regression estimates. The crisis indicator variable takes value 1 in each quarter after Q2 of 2007. The crisis peak
indicator variable takes value 1 for the last two quarters of the sample. Except in the first regression, we interact the crisis indicator variable and the crisis peak
indicator variable with a subgroup indicator variable and include the subgroup indicator variable in the regression as well. The subgroups are formed at the end of
June 2006 and are (1) investment-rated firms, (2) unrated firms, (3) small firms, namely firms whose asset size is below the top of the bottom quintile of NYSE
firms, (4) highly constrained firms, which are KL constrained firms in June 2006 that had positive net issuance in the four quarters ending in June 2006, (5) no
debt firms, (6) high cash firms, and (7) high leverage firms. The regressions use control variables that are not reported. These variables are listed in the text and
their definition is provided in the Appendix.
51
Panel A: Capital Expenditures
Investment
Grade
Unrated
Smallest
KL
Constrained
No Debt
High Cash
High Lev
0.0189
0.0107
0.0276
0.0038
0.0180
0.0104
0.0275
0.0035
0.0181
0.0101
0.0281
0.0037
0.0205
0.0107
0.0339
0.0050
0.0163
0.0086
0.0268
0.0038
0.0169
0.0065
0.0281
0.0048
0.0173
0.0107
0.0241
0.0033
0.0138
0.0139
-0.0001
0.8942
0.0141
0.0141
0.0000
0.9289
0.0141
0.0140
0.0002
0.7639
0.0151
0.0144
0.0006
0.4779
0.0114
0.0112
0.0001
0.8522
0.0095
0.0087
0.0008
0.3094
0.0168
0.0165
0.0003
0.6963
0.0100
0.0139
0.0134
0.0137
0.0091
0.0123
0.0136
0.0142
0.0085
0.0117
0.0132
0.0141
0.0097
0.0151
0.0171
0.0152
0.0077
0.0105
0.0118
0.0119
0.0065
0.0074
0.0085
0.0091
0.0103
0.0143
0.0151
0.0176
Diff. (2009Q1 - 2008Q3)
Paired Diff. (2009Q1 - 2008Q3)
Ttest 2008Q3=2009Q1*
-0.0034
-0.0036
0.0001
-0.0045
-0.0048
0.0001
-0.0047
-0.0049
0.0001
-0.0074
-0.0071
0.0001
-0.0041
-0.0039
0.0001
-0.0020
-0.0028
0.0001
-0.0048
-0.0052
0.0001
Diff. (2009Q1 - 2007Q2)
Paired Diff. (2009Q1 - 2007Q2)
Ttest 2007Q2=2009Q1
-0.0037
-0.0042
0.0001
-0.0051
-0.0059
0.0001
-0.0056
-0.0063
0.0001
-0.0055
-0.0062
0.0001
-0.0042
-0.0049
0.0001
-0.0026
-0.0034
0.0001
-0.0073
-0.0081
0.0001
Avg 1983-2004
Min
Max
Std. Dev.
Crisis vs pre crisis
Avg (2007Q3-2008Q2)
Avg (2006Q3-2007Q2)
Difference
Ttest
2009Q1
2008Q4
2008Q3
2007Q2
52
Panel B: Capital Expenditure Regressions
Whole
sample
crisis
-0.0011***
(0.000)
split dummy
crisis X split dummy
post-Sept 2008
-0.0020***
(0.000)
Post-Sept X split dummy
Constant
Fixed Effects
Observations
Number of gvkey
Adjusted R-squared
0.0078***
(0.000)
Yes
173,594
7,618
0.046
Unrated
Smallest
-0.0010***
(0.006)
-0.0013***
(0.000)
0.0008*
(0.069)
-0.0024***
(0.000)
0.0006
(0.221)
0.0079***
(0.000)
Yes
-0.0013***
(0.000)
-0.0013***
(0.000)
0.0013***
(0.002)
-0.0018***
(0.000)
-0.0003
(0.491)
0.0078***
(0.000)
Yes
172,445
7,312
0.047
172,445
7,312
0.047
Highly
Constrained
-0.0011***
(0.000)
-0.0007
(0.369)
0.0008
(0.300)
-0.0021***
(0.000)
0.0009
(0.332)
0.0078***
(0.000)
Yes
173,594
7,618
0.046
Robust pvalues in parentheses
*** p<0.01, ** p<0.05, * p<0.1
53
No Debt
Hicash
Hilev
-0.0014***
(0.000)
-0.0012***
(0.001)
0.0022***
(0.000)
-0.0021***
(0.000)
0.0006
(0.371)
0.0078***
(0.000)
Yes
-0.0014***
(0.000)
-0.0013***
(0.003)
0.0026***
(0.000)
-0.0024***
(0.000)
0.0024***
(0.000)
0.0078***
(0.000)
Yes
-0.0012***
(0.000)
-0.0018***
(0.001)
0.0018***
(0.004)
-0.0017***
(0.000)
-0.0016**
(0.012)
0.0078***
(0.000)
Yes
172,445
7,312
0.047
172,445
7,312
0.047
172,445
7,312
0.047
Appendix
All variables are quarterly, unless otherwise noted. For variables reported on a year-to-date basis, the
quarterly value is calculated by subtracting the lagged value from the current value; in the first quarter of
a fiscal year, the lagged value is set equal to zero. Variables names preceded by “lag” are the value of
that variable in the previous quarter; variable names preceded by lag2 are the value of that variable two
quarters prior. Variable name
Description
Acquisitions
acquisitions (acqy) divided by assets
Avgtint
average of monthly intermediate term treasuries during the quarter
Capex
capital expenditures (capxy) / lagged assets
Cash
cash and marketable securities (cheq) divided by assets
Dividend dummy
dummy variable equal to one if firm paid dividends
Dividends
total cash dividends (dvy) minus preferred dividends (dvpq) paid during the
quarter, divided by lagged assets
Dqsize
assets minus lagged assets, divided by lagged assets
HYspread
spread of Merrill Lynch US High Yield 100 Index over intermediate term
treasuries, averaged over each month of the quarter
Leverage
long-term debt (dlttq) plus debt in current liabilities (dlcq), divided by assets
(atq)
logMC
log (market value of equity)
MB
market-to-book calculated as book value of assets (atq) minus book value of
common equity (ceqq) plus the market value of common equity (cshoq*prccq)
Mkt lev
long-term debt (dlttq) plus debt in current liabilities (dlcq), divided by long-term
debt (dlttq) plus debt in current liabilities (dlcq) plus the market value of
common equity
Net equity
Issuance
equity issuance (sstky) minus aggregate equity repurchase (prstkcy), divided by
lagged assets
Net LT debt
Issuance
long-term debt issuance (dltisy) minus long-term debt retirement (dltry) divided
by lagged assets
Net total debt
issuance
change in long-term debt (dlttq) and debt in current liabilities (dlcq) during the
quarter, divided by lagged assets
NgBE
dummy variable equal to one if book equity (ceqq) is less than 0
NWC
working capital (wcapq) minus cash, divided by assets
54
Variable name
Description
OCF (operating
cash flow)
Operating cash flow calculated following Minton and Schrand (1999) as sales
(saleq) less cost of goods sold (cogsq) less selling, general and administrative
expenses (xsgaq) less the change in working capital for the period, divided by
total assets (atq). Working capital is current assets other than cash and short-term
investments less current liabilities and is calculated as the sum of the nonmissing amounts for accounts receivable (rectq), inventory (invtq), and other
current assets (acoq) less the sum of the non-missing amounts for accounts
payable (apq), income taxes payable (txpq), and other current liabilities (lcoq). If
all components of working capital are missing in either the current quarter or the
previous quarter, working capital and operating cash flow are both set equal to
missing. Quarterly selling, general and administrative expenses exclude onequarter of annual research and development costs (xrd) and advertising expenses
(xad) when those data items are available.
PPE
PPE (ppentq) divided by assets
R&D
R&D (xrdq) / assets. If R&D is only reported annually, then quarterly R&D is
set equal to one-fourth of annual R&D. If R&D is missing, it is set equal to 0.
Rated
dummy variable equal to one if the S&P long-term rating (splticrm) available on
Compustat is investment grade or speculative
Rdmiss
dummy variable equal to one if R&D is missing in Compustat
Sigma12
the median of the standard deviations of cash flow/assets over past 12 quarters
for firms in the same industry, as defined by two-digit SIC code.
Size
log (book value of assets in 2009 dollars)
STDebt
change in debt in current liabilities (dlcq), divided by lagged assets
55