Full text of DOI:10.1162/003355300554773

INVESTMENT-CASH FLOW SENSITIVITIES ARE USEFUL:
A COMMENT ON KAPLAN AND ZINGALES*
STEVEN M. FAZZARI
R. GLENN HUBBARD
BRUCE C. PETERSEN
A recent paper in this J ournal by Kaplan and Zingales reexamines a subset of
firms from work of Fazzari, Hubbard, and Petersen and criticizes the usefulness of
investment-cash flow sensitivities for detecting financing constraints. We show
that the Kaplan and Zingales theoretical model fails to capture the approach
employed in the literature and thus does not provide an effective critique.
Moreover, we describe why their empirical classification system is flawed in
identifying both whether firms are constrained and the relative degree of
constraints across firm groups. We conclude that their results do not support their
conclusions about the usefulness of investment-cash flow sensitivities.
In a recent paper in this J ournal Kaplan and Zingales [1997,
hereinafter KZ] argue that investment-cash flow sensitivities do
not provide useful evidence about the presence of financing
constraints. Because KZ use a subset of the same firms and the
same regressions as Fazzari, Hubbard, and Petersen [1988,
hereinafter FHP] and claim [page 176] that FHP ‘‘can legitimately
be considered the parent of all papers in this literature,’’ it is
appropriate that we respond. Based on a simple theoretical model,
KZ reach the provocative conclusion [page 211] that ‘‘the invest-
ment-cash flow sensitivity criterion as a measure of financial
constraints is not well-grounded in theory.’’ In Section I we show
that the KZ model does not capture the theoretical approach
employed in FHP and many subsequent studies. Most of the KZ
paper attempts to show that empirical investment-cash flow
sensitivities do not increase monotonically with the degree of
financing constraints within the 49 low-dividend firms from the
FHP sample. In Section II we explain why the KZ classification of
the degree of constraints is flawed in identifying both whether or
not firms are constrained (absolute constraints) as well as the
relative degree of constraints across firms. As a result, we argue in
Section III that there is no expected ex ante pattern for the
* We thank Michael Athey, Charles Calomiris, Robert Carpenter, Robert
Chirinko, Mark Gertler, Simon Gilchrist, Kevin Hassett, Charles Himmelberg,
Anil Kashyap, Ronald King, Wende Reeser, J oachim Winter, and two referees, one
of the editors (Andrei Shleifer), and participants in seminars at the London School
of Economics and the NBER Summer Institute Conference on Corporate Finance
for comments and suggestions.
2000 by the President and Fellows of Harvard College and the Massachusetts Institute of
Technology.
௠
The Quarterly J ournal of Economics, May 2000
695

696
QUARTERLY J OURNAL OF ECONOMICS
investment-cash flow sensitivities across the KZ categories, mak-
ing their empirical results uninformative about the usefulness of
investment-cash flow sensitivities.1
I. THE KZ MODEL AND TESTS OF FINANCING CONSTRAINTS
The one-period KZ model consists of a return on investment
F (I ), internal financing (W) with constant opportunity cost,
external financing (E ), and a premium for external funds C(E,k),
where k measures the cost wedge between internal and external
funds. KZ show that the investment-cash flow sensitivity is
C11
dW C11 Ϫ F11
dI
(1)
ϭ
,
where C11 is the slope of the supply curve for external finance and
F11 is the slope of the investment demand curve. KZ focus on firm
heterogeneity in dI/dW as measured by the level of W. To analyze
dI/dW at different levels of W they compute
F 211C211
d 2I
dW2
F111 C111
(2)
ϭ
Ϫ
.
3
4
(C11 Ϫ F11)3
F11
C11
KZ note that d 2I/dW2 is negative only if the term in brackets is
negative. They then point out that the bracketed term could be
positive if F111 Ͼ 0 or C111 Ͻ 0. This leads KZ to conclude that the
theoretical foundation of previous research is weak because dI/dW
may not fall as the degree of financing constraints declines (with
larger W).
Before we assess this conclusion, it is helpful to consider the
intuition (which does not appear in KZ) behind why d 2I/dW2 may
be positive. In Figure I investment is on the horizontal axis, F1 is
investment demand, WL or WH indicates the quantity of internal
financing (with constant marginal cost as indicated by the horizon-
tal line segment), and C1 is the supply of external funds. In the left
panel of Figure I, F111 ϭ 0 and C111 Ͻ 0 (i.e., linear demand and
concave supply). Investment is more sensitive to small internal
finance fluctuations (⌬W) at high internal finance (WH ) than at
low internal finance (WL) because a firm at WH uses less external
1. Extensive empirical research since FHP (surveyed by Hubbard [1998]) also
addresses many of the issues raised in KZ.

COMMENT ON KAPLAN AND ZINGALES
697
FIGURE I
Illustrations of the KZ Model
financing, and therefore the concavity of supply causes its C11 to
be larger (see equation (1)). Alternatively, consider F111 Ͼ 0 and
C111 ϭ 0 (i.e., convex demand and linear supply) as in the right
panel of Figure I. Again, investment is more sensitive to W at WH
than at WL because investment demand is more sensitive to the
cost of capital as W rises.
This focus in KZ on d 2I/dW2 does not provide an effective
critique of the literature (including the FHP theoretical approach)
because most studies do not use the level of W to classify firms.2
Instead, FHP and much of the literature classify firms according
to a priori criteria designed to give large differences in the slope of
the external financing schedule, C11, across groups. The obvious
testable implication of this approach, using equation (1), is that
constrained firms with a large C11 should have a larger dI/dW
than (relatively) unconstrained firms with a small (or zero) C11,
other things equal.3
The necessary condition for dI/dW to be larger for constrained
firms is
Unconstrained
11
Constrained
11
Unconstrained
11
(3)
CC11onstrained/C
Ͼ F
/F
.
2. In fact, KZ never reference any specific study, including FHP, to demon-
strate the relevance of d 2I/dW2.
3. To appreciate the intuition graphically, consider the effect of a small change
in W on two firms with linear demand curves. If the ‘‘constrained’’ firm faces
relatively steep and the ‘‘unconstrained’’ firm relatively flat supply, the result is
obvious. KZ implicitly assume away this possibility by positing that all firms face
the same C11 for a given level of E.

698
QUARTERLY J OURNAL OF ECONOMICS
While F11 may differ across firms, we can think of no reasons why
Constrained
11
Unconstrained
11
F
should be systematically greater than F
, and
KZ provide no reasons. Thus, as long as researchers separate
Constrained
11
Unconstrained
11
firms by a priori criteria that result in C
Ͼ C
in
the relevant range, the comparison of dI/dW across firm groups
has a solid theoretical foundation. We also point out that as
Unconstrained
11
C
approaches zero, as we argue below is the case in
many studies, (3) almost certainly holds. In addition, if (3) holds,
the issues that KZ raise about curvature and nonlinearity are not
likely to be relevant.4
The only remaining question is whether previous research
has effectively classified firms in ways that generate large differ-
ences in C11. Consider the model and discussion in FHP [pages
146–157 and Appendix A]. In the supply of funds schedule in FHP
Figure I, C11 equals zero for internal financing (as in KZ) and C11 is
greater than zero for external financing. One group of firms faces
C11 of zero at the margin because investment demand is less than
internal financing. In contrast, constrained firms exhaust inter-
nal funds and finance marginal investment with external funds,
and thus face a positive C11. Operationally, as implied by the
model, unconstrained firms are those with large dividend payouts,
and constrained firms are those with low or zero dividends.
Since FHP, many other researchers have devised different
approaches for separating firms into groups with low and high
C11.5 A common separating criterion is access to public debt.
Calomiris, Himmelberg, and Wachtel [1995] report that firms
with debt ratings are very different from firms without rated debt.
Firms that issue public debt, especially commercial paper, are far
larger on average, have much lower volatility of sales and income,
and therefore pose relatively little, possibly negligible, default
risk. The case can be made that firms with commercial paper or
high bond ratings face a C11 close to zero. Almost surely, firms that
issue public debt tend to have a substantially lower C11 than those
4. KZ also mention the possibility of ‘‘nonmonotonicity’’ with the wedge k as a
proxy for the degree of financing constraints. This approach is not relevant to the
FHP model, discussed in the next paragraph, because high-dividend firms, in
theory, face no wedge at the margin. In general, if researchers effectively split their
samples with criteria that generate large differences in k that lead to large
differences in C11, the condition in equation (3) is likely to be satisfied.
5. See Calomiris, Himmelberg, and Wachtel [1995]; Gilchrist and Himmel-
berg [1995, 1998]; Kashyap, Lamont, and Stein [1994]; and Whited [1992].
Hubbard [1998] provides many other references. Hoshi, Kashyap, and Scharfstein
[1991] use association with a large bank to identify firms with a relatively low C11
.
In addition, many studies split samples by firm size which is highly correlated with
both dividend payout and access to public debt.

COMMENT ON KAPLAN AND ZINGALES
699
that do not. In contrast, many firms without public debt also have
little bank debt, consistent with the fact that many of them are
small, high-technology companies with little collateral and likely
pronounced moral hazard and adverse selection problems (see, for
example, Himmelberg and Petersen [1994]). Thus, a strong case
can be made that these firms face a high C11 for external financing.
Empirical evidence from most studies is consistent with
equation (1) in the sense that firms likely to have a priori high C11,
(e.g., firms with low dividends, no public debt, or small size)
almost always have a larger dI/dW than firms likely to have a low
C11. Furthermore, many studies cannot reject dI/dW equals zero
for control groups selected to have a low (or zero) C11 (e.g.,
Gilchrist and Himmelberg [1995, 1998]).6 Thus, the implications
of the theoretical approach in much previous research are sup-
ported by the evidence.
II. PROBLEMS WITH THE KZ EMPIRICAL CLASSIFICATION APPROACH
KZ employ managerial statements and quantitative mea-
sures from firms’ financial statements to sort the 49 FHP low-
dividend firms into one of five groups:7 Not Financially Con-
strained (NFC), Likely Not Financially Constrained (LNFC),
Possibly Financially Constrained (PFC), Likely Financially Con-
strained (LFC), and Financially Constrained (FC). This section
summarizes our concerns about the effectiveness of their ap-
proach for determining both absolute and relative constraints
across firms.
A. Reliance on Managers’ Statements and Regulation S-K
To justify use of managerial statements to identify the degree
of financing constraints, KZ [p. 180] rely on Securities and
Exchange Commission Regulation S-K which they claim ‘‘explic-
itly requires firms to disclose whether or not they are having
difficulty financing their investments.’’ It is not obvious, however,
that this regulation forces a firm to reveal financing constraints.
We contacted Robert Lipe, Academic Fellow in the Office of the
Chief Accountant of the SEC and asked whether a firm that is
6. See also Kashyap, Lamont, and Stein [1994]. Some Euler equations studies
cannot reject C11 equal to zero for control groups of firms [Gilchrist 1991; Hubbard,
Kashyap, and Whited 1995; Whited 1992].
7. KZ do not explain how these diverse criteria are specifically combined to
classify firms into the five groups.

700
QUARTERLY J OURNAL OF ECONOMICS
unable to undertake a new, positive NPV project due to financing
constraints would be obliged to reveal this information. Lipe
responded that this is not the case. Rather, he explained, Regula-
tion S-K requires the firm to reveal the inability to invest due to
financing constraints only when the firm fails to act on a previ-
ously announced investment commitment. As a result, we doubt
the relevance of self-serving managers’ statements as evidence of
the absence of financing constraints in most situations.
B. Problems with the Quantitative Classification Criteria
The classification criteria in KZ include cash stocks, unused
lines of credit, and leverage. They report summary measures for
these variables in Table III [KZ, pages 185–187] and argue that
they support the success of their relative ranking of the degree of
financing constraints and their finding that the firms face abso-
lute financing constraints (PFC, LFC, or FC) in only 15 percent of
the firm-years.
We begin by explaining why the summary statistics in KZ do
not support their surprising finding about the infrequency of
absolute constraints in the FHP sample. KZ suggest that both the
cash flow and the cash stock positions for NFC and LNFC
firm-years are so large relative to fixed investment that these
firms could not be financially constrained. Their numbers in Table
III, however, are misleading because they implicitly assume that
firms use sources of financing only for fixed investment when, in
fact, growing companies invest heavily in both inventories and
accounts receivable (see Fazzari and Petersen [1993, pages 330–
331]).
We recomputed the KZ figures with the proper comparison of
cash flow and cash stocks relative to total investment (fixed
investment plus the changes in inventories and accounts receiv-
able). These new statistics change some of the KZ conclusions. For
example, KZ [page 184] note that the median value of cash flow
less fixed investment is positive for NFC firm-years and write
‘‘[t]his suggests that NFC firms could have increased their invest-
ment without tapping external sources of capital.’’ In sharp
contrast, in our computations the median value of cash flow less
total investment is negative at the seventy-fifth percentile for
even the NFC and LNFC firms. Thus, most NFC and LNFC firms
exhaust all internal finance for investment purposes. Further-
more, while the median cash stock-fixed investment ratio for NFC
and LNFC firm-years is 0.66 (similar to the statistics in KZ Table

COMMENT ON KAPLAN AND ZINGALES
701
III) the median ratio of cash stocks to total investment is only
0.27.8
In our opinion, this cash stock ratio is too small to support the
interpretation in KZ of the absence of financing constraints.
Financially constrained firms will rationally maintain some buffer
stock of cash to protect against having to cancel or delay invest-
ment projects as well as to avoid the costs associated with
financial distress. It is well-known that cash flow is volatile in
manufacturing, frequently declining by 50 percent or more and
often becoming negative during a recession. Suppose, for example,
that cash flow declined to zero. Our computations indicate that
NFC and LNFC firms could maintain only about three months of
median total investment from cash stocks, and then only if these
stocks were (implausibly) driven to zero. We believe these statis-
tics are consistent with the view that these firms face absolute
financing constraints.
The cash stock, unused lines of credit, and leverage figures
are also unreliable measures of the relative degree of financing
constraints. Firms may have low debt because they cannot
convince lenders to provide them with credit, perhaps due to lack
of collateral, and low-debt firms may therefore face more severe
constraints. For example, small high-tech companies—much of
the FHP sample—tend to have little collateral value, and little
debt, possibly because their assets are intangible or firm-specific
(see, for example, Himmelberg and Petersen [1994]). In addition,
comparatively large cash positions or unused lines of credit may
indicate relatively severe constraints. As argued in recent papers
[Fazzari and Petersen 1993; Carpenter, Fazzari, and Petersen
1994; Calomiris, Himmelberg, and Wachtel 1995], it is costly for
constrained firms to adjust fixed investment when internal funds
fluctuate. Forward-looking firms will therefore partially protect
themselves with buffer stocks of cash or unused debt capacity. The
more financially constrained a firm is, the greater is its incentive
to accumulate liquid buffer stocks. Such a firm may be able to
invest more at the margin at a moment in time, but the firm is
nonetheless financially constrained. This dynamic perspective
contrasts with the static view of financing constraints employed
by KZ, which creates problems in their classification approach.
8. This statistic excludes observations for which total investment is less than
or equal to zero. KZ also point out that unused lines of credit are larger for NFC
and LNFC firms. We do not have these data, but the ratios of slack to investment
reported by KZ on page 188 would be similarly reduced by recognizing a broader
measure of investment.

702
QUARTERLY J OURNAL OF ECONOMICS
C. The Absence of Heterogeneity in the KZ Classification
One striking finding in KZ is that only 19 of 719 observations
(2.6 percent) are FC and another 34 observations (4.8 percent) are
LFC. Given so few FC and LFC observations, how do KZ obtain
enough FC firms for their regressions? KZ placed firms in the FC
category if they had just a single year (out of 15) with an FC or
LFC rating. In the FC category, 14 of the 22 firms had an FC or
LFC rating only one or two times, while six firms had FC or LFC
ratings in just three or four of the fifteen years. For this reason,
the difference in cash flow coefficients across the KZ regressions
may have little to do with their relative ranking of financing
constraints.
III. THE KZ REGRESSION RESULTS
KZ find that the investment of NFC and LNFC firms displays
a greater sensitivity to cash flow than FC firms. Space does not
permit a detailed discussion of this pattern of results. One
possibility is that the FC firms include some years of financial
distress. KZ describe firms in FC years as having ‘‘liquidity
problems,’’ which is not surprising given that their criteria for
receiving the FC classification include violation of debt covenants
and renegotiation of debt payments [page 182]. The KZ summary
statistics in Table III also strongly suggest that the FC firm-years
are periods of financial distress.9 During years of financial dis-
tress, firms, possibly at the insistence of their creditors, are likely
to use cash flow to enhance liquidity and avoid bankruptcy
resulting in little change in fixed investment as measured in
Compustat. A broader measure of investment, however, is likely to
respond much more to cash flow for such firms.10
Financial distress is one possible explanation for the low cash
flow coefficient of the FC firms. Regardless of how one explains the
9. The mean cash flow-net plant ratio for these observations is Ϫ0.047 and the
mean of interest coverage is only 1.650. While KZ recognize the possibility of
financial distress in FC observations [page 208], the defense they offer is not
convincing. They note that firms increase rather than repay debt in the PFC, LFC,
and FC years. This observation, however, may be due to creditors permitting
illiquid, but growing, firms to rebuild liquidity.
10. Financially distressed firms (with low or negative cash flow) often
disinvest assets with low adjustment costs such as working capital (see Fazzari
and Petersen [1993]). In addition, such firms likely sell off existing fixed assets.
Neither of these responses are included in the Compustat measure of fixed
investment and ignoring them causes a downward bias in the cash flow coefficient,
especially at times of financial distress.

COMMENT ON KAPLAN AND ZINGALES
703
pattern of results in KZ, however, we argue that this pattern is not
informative. As discussed in the previous section, the firms in the
NFC and LNFC categories likely are financially constrained and
the relative degree of constraints across the KZ categories is far
from clear. If there is no clear a priori difference in financing
constraints across the firm groups in KZ, their strategy does not
meet the criterion (summarized by equation (3)) necessary for
meaningful tests of financing constraints with firm heterogeneity.
Finally, KZ [page 196] present additional tests with group-
ings based on ‘‘quantitative/objective data.’’ The only one of these
tests consistent with their main findings shows that firms with
high interest coverage have higher cash flow coefficients than
firms with low coverage. KZ imply that the pattern should be the
opposite, but this need not be the case. As we discussed earlier,
either low levels of debt or high interest coverage may indicate an
inability to obtain debt financing, possibly signaling relatively
severe financial constraints. KZ [page 211] themselves note that
some studies use high leverage as an indicator of more severe
financing constraints, while other studies argue the opposite.
Thus, these tests do little to bolster the KZ conclusions.11
IV. CONCLUSION
KZ argue that investment-cash flow sensitivities do not
provide useful evidence about the presence of financing con-
straints. We believe that this conclusion does not follow from their
analysis for two reasons. First, their theoretical model fails to
capture the approach of most previous research, making their
theoretical analysis irrelevant as a criticism of FHP and most
subsequent research. Second, the KZ empirical findings are
difficult to interpret. The 49 low-dividend FHP firms are a poor
choice for such a study because they are relatively homogeneous
for purposes of testing for capital-market imperfections, making it
extremely difficult to classify these firms finely by degree of
financing constraints. Furthermore, some of the KZ classification
11. Two new studies are relevant to the KZ results. In a sample of large,
dividend-paying firms, Cleary [1999] argues that the ‘‘most financially con-
strained’’ firms have the lowest investment-cash flow sensitivity. These FC firms,
however, appear to be financially distressed. Their mean net income is Ϫ4.8
percent of sales compared with 9.6 percent for NFC firms. Mean sales growth for
FC firms is Ϫ2.3 percent versus 23.5 percent for the NFC firms. Winter [1998],
using the KZ sample, includes the KZ indicator of financial constraint status in
regressions for investment and firm exit. He finds that the KZ indicator is either
statistically insignificant or, when significant, has the wrong sign.

704
QUARTERLY J OURNAL OF ECONOMICS
criteria (e.g., stock of cash and degree of leverage), may indicate
high or low levels of constraints. We therefore believe their finding
of nonmonotonic investment-cash flow sensitivities is not
informative.
While the sweeping critical conclusions in KZ do not follow
from their results, we believe their paper makes a contribution.
Empirical work in this area has not always clearly identified the
theoretical model under investigation. While FHP provided a
model of investment behavior that described the criteria for
separating firms into ‘‘constrained’’ and ‘‘unconstrained’’ catego-
ries, not all papers have done so. In addition, while commonly
used separating criteria have a solid theoretical foundation, not
all criteria are as defensible. KZ (and we hope this comment) will
lead future researchers to clearly state their model and to
carefully choose the criteria used for defining constrained and
unconstrained groupings.
WASHINGTON UNIVERSITY AND J EROME LEVY ECONOMICS INSTITUTE
COLUMBIA UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH
WASHINGTON UNIVERSITY
REFERENCES
Calomiris, Charles W., Charles P. Himmelberg, and Paul Wachtel, ‘‘Commercial
Paper and Corporate Finance: A Microeconomic Perspective,’’ Carnegie-
Rochester Conference Series on Public Policy, XLI (1995), 203–250.
Carpenter, Robert E., Steven M. Fazzari, and Bruce C. Petersen, ‘‘Inventory
Investment, Internal-Finance Fluctuations, and the Business Cycle,’’ Brook-
ings Papers on Economic Activity (1994:2), 75–138.
Cleary, Sean, ‘‘The Relationship between Firm Investment and Financial Status,’’
J ournal of Finance, LIV (1999), 673–692.
Fazzari, Steven M., R. Glenn Hubbard, and Bruce C. Petersen, ‘‘Financing
Constraints and Corporate Investment,’’ Brookings Papers on Economic
Activity (1988:1), 141–195.
Fazzari, Steven M., and Bruce C. Petersen, ‘‘Working Capital and Fixed Invest-
ment: New Evidence on Finance Constraints,’’ RAND J ournal of Economics,
XXIV (1993), 328–342.
Gilchrist, Simon, ‘‘An Empirical Analysis of Corporate Investment and Financing
Hierarchies Using Firm-Level Panel Data,’’ mimeograph, Board of Governors
of the Federal Reserve System, 1991.
Gilchrist, Simon, and Charles P. Himmelberg, ‘‘Evidence on the Role of Cash Flow
for Investment,’’ J ournal of Monetary Economics, XXXVI (1995), 541–572.
Gilchrist, Simon, and Charles P. Himmelberg, ‘‘Investment, Fundamentals, and
Finance,’’ NBER Macroeconomics Annual, XIII (Cambridge, MA: MIT Press,
1998).
Himmelberg, Charles P., and Bruce C. Petersen, ‘‘R&D and Internal Finance: A
Panel Study of Small Firms in High-Tech Industries,’’ Review of Economics
and Statistics, LVI (1994), 38–51.
Hoshi, Takeo, Anil K. Kashyap, and David Scharfstein, ‘‘Corporate Structure,
Liquidity, and Investment: Evidence from J apanese Panel Data,’’ Quarterly
J ournal of Economics, CVI (1991), 33–60.
Hubbard, R. Glenn, ‘‘Capital-Market Imperfections and Investment,’’ J ournal of
Economic Literature, XXXV (March 1998), 193–225.

COMMENT ON KAPLAN AND ZINGALES
705
Hubbard, R. Glenn, Anil K. Kashyap, and Toni M. Whited, ‘‘Internal Finance and
Firm Investment,’’ J ournal of Money, Credit and Banking, XXVII (1995),
683–701.
Kaplan, Steven N., and Luigi Zingales, ‘‘Do Investment-Cash Flow Sensitivities
Provide Useful Measures of Financing Constraints?’’ Quarterly J ournal of
Economics, CXII (1997), 169–215.
Kashyap, Anil K, Owen A. Lamont, and J eremy C. Stein, ‘‘Credit Conditions and
the Cyclical Behavior of Inventories,’’ Quarterly J ournal of Economics, CIX
(1994), 565–592.
Whited, Toni M., ‘‘Debt, Liquidity Constraints, and Corporate Investment: Evi-
dence from Panel Data,’’ J ournal of Finance, XLVII (1992), 1425–1460.
Winter, J oachim K., ‘‘Does Firms’ Financial Status Affect Plant-Level Investment
and Exit Decisions?’’ mimeograph, University of Mannheim, 1998.