Joint analysis of the non-linear debt–growth nexus and cash-flow sensitivity: New evidence from Italy

Available online 20 December 2012 We exploit a large panel of Italian manufacturing firms observed over the period 1998-2003 to jointly assess the impact of cash flow and leverage on co

Structural Change and Economic Dynamics 24 (2013) 34-44

Contents lists available at SciVerse ScienceDirect _ | STRUCTURAL

GEES AND ECONOMIC

“T= | YAS

a eh

SP

na

Joint analysis of the non-linear debt-growth nexus and cash-flow

sensitivity: New evidence from Italy

Massimo Molinari

Department of Economics and Management, University of Trento, via Inama 5, 38122 Trento, Italy

Article history: This paper conducts an empirical investigation of the finance-growth nexus at firm level Available online 20 December 2012 We exploit a large panel of Italian manufacturing firms observed over the period 1998-2003

to jointly assess the impact of cash flow and leverage on corporate growth measured in

cáo geneity and persistence by using a system GMM estimator fully developed by Blundell and

125 Bond (1998) We find an inverted U-shaped relationship between debt and growth: at low

levels, leverage exerts a positive influence on growth and yet there appears to be a nega- Keywords: tive relationship between growth and debt-exposure for fragile firms, i.e highly leveraged Firm growth ones This finding is consistent with the idea that debt initially enables firms to broaden Leverage their financing options and provides additional resources to growth Nevertheless, once

negatively affect firm expansion Highly leveraged firms are also endowed with relatively lower levels of internal cash flow and exhibit higher growth-cash flow sensitivity We con- clude that this latter result can be interpreted as evidence of the existence of financial constraints

© 2013 Elsevier B.V All rights reserved

cial constraints because internal cash-flow and debt are the

In recent years, empirical research on firm dynamics has

paid increasing attention to the finance—growth nexus, and

two streams of literature can be distinguished One the one

hand, several scholars have studied the effect of leverage

on firm performance; on the other hand, a large body of

works have focused on financial constraints and the inter-

pretation of the correlation between internal cash-flow and

investment or growth This study seeks to provide a bridge

between these two strands of research

We work with a large panel of small-medium Italian

manufacturing firms observed over the period 1997-2003

The vast majority of manufacturing firms in Italy are not

listed, and financial markets in Italy are less developed than

in the United States or other European countries Hence, the

E-mail address: massimo.molinari@unitn.it

0954-349X/$ - see front matter © 2013 Elsevier B.V All rights reserved

http://dx.doi.org/10.1016/J.strueco.2012.11.003

primary sources of financing for most firms

In this context, we estimate a dynamic growth model augmented by cash-flow and a measure of leverage, and

we rely on a system GMM estimator to overcome prob- lems of unobserved firm-specific effects, persistence and endogeneity of explanatory variables

Our findings suggest that there is an important connec- tion between cash-flow and debt-structure More in detail,

we find an inverted U-shaped relationship between debt and growth: initially, leverage exerts a positive effect of growth and low-leveraged firms are not constrained (or are so to a much less extent) by internal funds Conversely, there seems to be a negative relationship between growth and debt-exposure for fragile firms, i.e highly leveraged ones

Our interpretation is that debt initially allows firms

to expand their financing options and provides additional

resources to sustain investment plans and future growth

This is until negative liquidity and debt overhang effects

kick in and hamper firm growth In this respect, such firms

can be considered debt-constrained In addition, highly

leveraged firms also appear to be internally constrained

because they exhibit relatively lower levels of internal

funds Cash-flow, or the lack thereof, is more binding for

these firms, and this also explains why the sensitivity of

firm growth to cash-flow is found to increase with the level

of indebtedness

The remainder of the paper is organized as follows:

Section 2 provides a review of the main contributions on

cash-flow sensitivities and financial constraints and puts

forward the first research hypothesis Section 3 outlines

the possible links between leverage and growth and form-

ulates the second research hypothesis Section 4 presents

the data-set and Section 5 focuses on the empirical analysis

Section 6 summarizes the main results and concludes

2 Cash flow sensitivities and financial constraints

Carpenter and Petersen (2002) put forth a model of

firm growth based on the internal finance theory If there

were perfect capital markets and no bankruptcy costs, the

supply of finance would be a flat schedule In this world,

firms facing a temporary shortage of internal funds could

simply resort to external finance, so that their growth pro-

cess should not be hampered by the lack of cash-flow Yet,

in practice, the debt portion of the supply schedule will

be upward sloping The supply is flat as long as the firm

exploits its internal resources at a given shadow cost Once

such funds are entirely used up, the firm must finance

additional projects through debt The more indebted a

firm is, though, the more incentives it will have to select

risky projects, and this increases the probability of default

Providers of external funds want to hedge against the

moral-hazard risk attached to the use of debt and charge

higher interest rates This explains why the cost of finance

increases with the degree of leverage and hence the upward

sloping part of the supply of funds.! In this world, a nega-

tive shock to cash flow would not be replaced one for one by

debt and will hence have a negative impact on investment

The standard empirical approach to testing this hypoth-

esis is to estimate either Tobin’s-Q investment equation

or firms’ growth models augmented by cash-flow A pos-

itive and significant coefficient on cash-flow would then

be interpreted as evidence of the existence of financial

constraints A vast empirical literature has tested this

hypothesis by focusing on the impact of cash flow on firms’

investment decisions and growth.”

1 In the original model outlined by Carpenter and Petersen, the firm

may turn to external equity at a flat rate once the cost of debt becomes

sufficiently high Yet, in the case of small business enterprises (SME), we

can safely neglect this portion of the supply because the vast majority of

firms are not publicly listed

2 See, among others, Carpenter and Petersen (2002) for a study on US

small (listed) firms; Audretsch and Elston (2002) use data on German

listed firms; Fagiolo and Luzzi (2006) rely on Italian data on manufac-

turing firms; Oliveira and Fortunato (2006a) exploit a panel of Spanish

manufacturing firms

This empirical approach has been criticized on sev- eral grounds A wide array of studies have documented that cash flow matters, but several authors have pointed out that the cash flow could be significant in an invest- ment or growth regression simply because of the bias that arises when cash flow is positively correlated with (not properly controlled for) investment opportunities.? Should this be the case, one could not interpret positive cash flow sensitivities as an evidence that financial con- straints are binding Nonetheless, as discussed in Carpenter and Guariglia (2008), one would expect this effect to be constant across all groups of firms, while the severity of capital markets imperfection may vary within the sample Hence, researchers have started to sort firms into groups characterized by an increasing likelihood of being finan- cially constrained The detection of higher sensitivity to cash flow for firms a priori categorized as constrained is believed to provide stronger support for the financial con- straint hypothesis

Several authors have proposed alternative criteria with which to identify such financially constrained firms.* It appears that different segmenting strategies could lead to different results, and a shared consensus has not yet been reached The view according to which ex ante constrained firms should exhibit higher sensitivity to cash flow first came under attack when Kaplan and Zingales published

an influential paper in 1997 as a response to Fazzari et al (1988) The former authors re-examined the samples of low-dividend firms used by Fazzari et al (1988) and re- classified firms according to some qualitative information provided by the management, together with some infor- mation on liquidity statements They showed that firms classified as constrained appeared to be less sensitive to cash flow than other firms This controversy has spurred

a number of follow-up publications and a lively and still unresolved debate, and researchers are still arguing on the- oretical and empirical grounds.°

As far as the theory is concerned, economists are debat- ing over the conditions under which one can expect to have

3 See Chirinko and Schaller (1995) or Carpenter and Guariglia (2008)

4 In their seminal paper, Fazzari et al (1988) split the sample according

to dividend payout, on the assumptions that a high dividends payout ratio signals the absence of financial constraints Among others, Gilchrist and Himmelberg (1995) use bond rating, whereas Devereux and Schiantarelli (1990) consider group membership Most of this information is only avail- able for listed firms, however Cabral and Mata (2003) proxy financial constraints with age, on the assumption that young firms are endowed with less cumulated wealth, or have access to a narrower set of external sources Gertler and Gilchrist (1994) use firms size to identify financial constraint, relying on the idea that smaller firms typically can have access

to fewer external finance options Hoshi et al (1991) work with Japanese firms and distinguish between firms on the basis of whether or not they belong to a keiretsu, and as such have a main-bank relationship Bond

et al (2003) use countries in which the firm is based as a sample-splitting criterion

” See, among others, Fazzari et al (2000), Kaplan and Zingales (2000) and Cleary (1999) To be noted is that this unresolved debate has stim- ulated some researchers to steer away from analyses based on cash flow sensitivities and opt for a more encompassing measure of finan- cial constraints See for instance, Musso and Schiavo (2008), who develop

a time-varying multivariate continuous index to identify the degree of financial constraints.

a monotonic positive relation between the degree of finan-

cial constraints and the cash flow sensitivity.®

Empirical studies are still arguing over the soundness of

the ex ante criteria used to identify constrained and non-

constrained firms More specifically, much of the dispute

between Kaplan and Zingales and Fazzari, Petersen and

Hubbard revolves around the notion and the operational

definition of financial constraints For example, Fazzari and

co-authors claim that what Kaplan and Zingales called

financially constrained firms are actually financially dis-

tressed and constrained firms The difference is important

because a financially distressed firm may be prevented by

creditors from making normal use of cash flow and that

would reduce the sensitivity of investment to cash flow In

this particular case, a financial constraint may not be cap-

tured by the cash flow coefficient in a regression analysis.’

This paper contributes to this debate by focusing on

the degree of indebtedness as a useful variable with which

to discriminate between constrained and non-constrained

companies Debt and cash flow are the two main channels

of financing for most firms, and it makes sense to jointly

assess the direct effect of leverage on growth, evaluate the

degree to which leverage affects cash flow sensitivities, and

finally discuss why this can be interpreted in favor of the

financial constraint hypothesis

Several authors have previously associated the level of

indebtedness with the firm’s financing constraints status.®

We wish to revisit and build on the argument for this

choice The basic idea is that highly leveraged firms can

be considered (debt) constrained because they are closer

to the exhaustion of their maximum debt capacity Hence

they may not be able to raise additional external finance

(at least not at reasonable borrowing terms) when needed

For obvious reasons, an excessively high cumulated debt

increases the likelihood of default and therefore calls for a

higher risk premium that ultimately accounts for a steeper

supply schedule As explained, this would then make firms

more sensitive to cash flow

Other authors, such as Fazzari et al (2000), assert that

leverage is an ambiguous prior, and that cannot be used

to distinguish the degree of financing constraints In fact, a

firm may have low debt simply because it cannot obtain

external finance from lenders If this were the case, the

low debt (not the high) firm would be constrained This

argument implicitly assumes that the causality between

leverage and financing constraints goes the other way

round, i.e the observed level of leverage is the result of

financing constraints

Some observations on this critique are in order This lat-

ter argument is not necessarily in contradiction with the

6 Kaplan and Zingales (1997) make a first attempt with a simple static

one-period model, and Cleary et al (2007) develop a more complicated

dynamic framework in which costs and revenue effects counteract each

other In this model, the observed sign of cash flow sensitivity ultimately

depends on which effect dominates

7 For this reason we deliberately discard distressed firms, i.e firms with

negative net-worth By doing so, we lose 399 observations on the balanced

panel

8 See, for instance, D’Espallier and Guariglia (2009), Whited and Wu

(2006), Kaplan and Zingales (1997) and Whited (1992)

previous one In fact the former relies on the notion of cumulated, whereas the latter implicitly refers to newly- issued debt At any point in time, the observed leverage ratio mainly reflects cumulated debt, and it may well be that a firm does not gain access to additional loans precisely because it already has too much debt and hence is debt- constrained This is precisely our argument Alternatively,

it may be that the firm exhibits a persistently low cumu- lated debt and hence a low leverage ratio If this were the case though, there must be some other factors that account for this condition For example, size or age consideration may have a role It is well documented that small and young firms pay a considerably higher risk-premium (pos- sibly because they lack of collateral), and therefore their growth process is hampered by a limited access to external funds This would explain low debt and a constrained sta- tus, and it would invalidate the initial argument because this would bias the monotonic positive relation between debt and the severity of financial constraints Hence, we attempt to attenuate this possible bias by controlling first for age and then for size in our regression analysis Moreover, Fazzari et al (2000) do not offer an explicit alternative explanation for why highly leveraged firms would display higher sensitivity to cash flow In fact, if leverage does not necessarily pick up differences in the degree of financial constraint, at best no significant pattern should be detected when comparing cash flow sensitivi- ties across samples sorted by the leverage ratio If low debt firms are constrained (as Fazzari et al., 2000 suggest may

be the case), then one should find that these firms are the ones most responsive to cash flow fluctuations

Empirical analysis can be used to discriminate between these two competing views For the reasons put forward in this discussion, our first research hypothesis is that lever- age captures the degree of financial constraints and this is captured by highly indebted firms being more responsive

to cash flow shocks

3 Leverage and growth

In this section we introduce the literature studying the impact of leverage on firm’s performance, we then intro- duce the second hypothesis that will be tested in our empirical analysis in Section 5 There are several channels through which leverage can impact on growth One possi- bility is that debt overhang has a negative effect because it creates a conflict between shareholders and bondholders.? According to this view, managers are assumed to act in the best interest of shareholders A manager of a low- leveraged firm has a stronger incentive to take advantage of positive growth opportunities because most gains will be enjoyed by shareholders By contrast, a manager of a highly indebted firm may pass up on positive net-present value projects because most revenues would accrue to bondhold- ers For these reasons, excessive debt overhang is believed

to induce under-investment and has a negative impact on growth

° See Myers (1977)

A liquidity effect may work as a further impediment to

corporate growth.!° Typically, a firm committed to large

debt repayments is expected to fewer less internal liquid

resources to finance investment In addition, such firms

can find it harder to obtain additional loans and therefore

to grow In other words, when the existing debt reaches

the maximum debt-capacity of the firm, it becomes bind-

ing and limits the ability of the firm to raise more external

finance, de facto curtailing future growth

These negative effects are likely to be stronger for

unlisted firms In fact, unlisted firms lack the (exter-

nal) equity channel and as such their mode of operation

inevitably exhibits a higher dependence on debt The same

logic applies to a liquidity effect

That said, it is important to draw a further distinction

between listed and unlisted firms Listed firm choose the

desired level of debt to minimize agency conflicts The pre-

diction of a negative association between debt and growth

or investment is driven by the way managers respond to

minimize the risk related to the debt-overhang As pointed

out by Lang et al (1996), firms with high growth oppor-

tunities are likely to self-select into a low debt state in

order to take advantage of future investment opportunity

They essentially substitute debt for (external) equity Firms

with high investment opportunities choose low-debt states

in order to minimize the risk of under-investment (high

debt leads to under-investment) Another argument is dis-

cussed by Jensen (1986) and applies more to firms with

limited growth opportunities Managers are believed to

have a strong preference for increasing firm scale irrespec-

tively of the quality of the projects undertaken The extent

to which managers can undertake his suboptimal expan-

sion depends on the amount of free cash flow available

Higher leverage imposes more financial pressure on man-

agers and reduces the amount of funds that could be used

to finance poor investment projects Such funds are in fact

put aside to meet interest payments under the threat of

bankruptcy (see Nickell and Nicolitsas, 1999) Leverage is

tailored to reduce undesired poorly-planned investment

(especially in firms with low investment opportunities) and

induce managers to pursue efficiency and select only prof-

itable investment plans This ultimately should improve

firm’s performance and increase its value Even in this

case, though, the assumption is that firms enjoy substantial

degrees of freedom when setting their financial structure

In the case of listed firms, their investment capacity does

not depend (or not only) on the level of debt because they

can collect outside funds via the issue of new stocks, and

the choice of the most appropriate level of debt is made

to tame agency conflicts The empirical evidence seems

to support the theoretical predictions delivered by these

agency models: Lang et al (1996), Aiavazian et al (2005)

and Hovakimian (2009) have found a negative relationship

between leverage and investment/growth Yet, these stud-

ies are all based on samples of listed firms By contrast, the

extent to which unlisted firms can adjust their financial

structure in response to such conflicts is much more limited

as long as they do not have access to outside equity Also,

10 See Bernanke et al (1996) and Lang et al (1996)

for an unlisted firm, there is a much tighter link between its investment capacity-growth and debt finance Suppose that a firm with limited internal resources has a new profit- able investment opportunity It will respond to it by trying

to obtain the outside funds (i.e loans) necessary to real- ize the positive net-present-value investment Cooley and Quadrini (2001) recognize that “more debt allows them to expand the production scale and increase their expected profits” It is natural to think that fast growing firms will have a greater demand for finance, as acknowledged by Fazzari et al (2000) In similar vein, Guariglia et al (2008) associate leverage with the availability of more resources and hence growth As long as these firms do not fully pay off their debt at the end of each period, they will build up cumulated debt over time and their growth process will be for the most part debt-driven Hence we expect to observe

a positive association between growth and leverage Empirical evidence in support of this argument is start- ing to emerge from studies on small medium enterprises For instance, Guariglia et al (2008), working on a large sample of unlisted Chinese firms, find that leverage has a strong positive effect on asset growth Honjo and Harada (2006) use data on Japanese manufacturing firms and find that leverage has a positive effect on sales growth The same positive effect of leverage on sales growth has been detected by Huynh and Petrunia (2010) working

on Canadian manufacturing firms The present discussion shows that there are counteracting dynamics that can ulti- mately account for the observed relationship between debt and growth A positive association between leverage and growth should be expected at lower levels of debt up to the point where excessive exposure triggers some perverse dynamics (liquidity effect—debt overhang) that depress the firm performance For this reason, our second research hypothesis is that the relationship between growth and debt-exposure should follow an inverted U-shape

4 Data and descriptive statistics

We employ a dataset provided by the Italian Account Data Service Centrale dei Bilanci, CeBi.11 CeBi was founded

in the early 1980s and it is now recognized by the Italian banking system as an institutional provider of information

on firms, and its long-term reputation guarantees the reli- ability of the data Our sample comprises limited liability firms whose accounting books must by law be made pub- licly available at the Chambers of Commerce CeBi collects information, organizes it and performs preliminary clean- ing In particular, only balance sheets complying with the IV EEC directive are considered For these reasons, we expect the bias due to measurement errors to be relatively small and the information reported to be reliable

The database contains records of balance sheets and asset structure for firms operating in the manufactur- ing sector from 1998 to 2003 Firms have been grouped together according to their ATECO code of principal

1 All the data have been made available to us by the Research Office

“Pianificazione, Strategie e Studi” at Unicredit Bank under the mandatory condition of censorship of any individual information.

activity: manufacturing sector includes firms that have the

two-digits ATECO code from 15 to 36.12

The dataset to which we have access contains firms with

total sales larger than 1 million euros and with at least

two employees As pointed out in previous works based

on the same dataset (see Bottazzi et al., 2008), one should

try to focus on entities characterized by a minimum level

of organizational structure and leave self-employment-

related phenomena out of the picture Moreover, Bottazzi

et al (2008) explain how firms with only one employee

appear to be very different from the rest of the population

in terms of structure of production The 1 million euros cut-

off is another restriction imposed to identify true firms,

and this is a fairly standard threshold.!? In addition to

these filters, we discarded those firms with a growth rate

of employees larger than 200 percent and lower than —100

percent, in order to eliminate outliers due to badly reported

data and minimize the probability of including anomalous

growth rates possibly due to mergers

Firms that did not have complete records on our main

regression variables were also dropped.'* After applying

these filters, we are left with a sample of 9274 firms sur-

viving from 1998 to 2003 continuously observed for six

years

Note that as we are dealing with ratios, we need not take

explicitly into account the issue of inflation Given the short

time dimension, this is likely not to be a major problem

anyway

For the purpose of the analysis, we focused on two key-

variables In first place we considered cash flow (CF) as a

flow measure of firm liquidity

Cash flow is defined as the sum of net profits, amortiza-

tion and depreciation and proxies the firm’s self-financing

capacity

As in Fagiolo and Luzzi (2006), we use cash flow scaled

by firm sales (SCF) This seems to be a better proxy for

firm liquidity since it is less sensitive than cash flow (CF)

to shifts in firm size We verify that the correlation coeffi-

cient between firm cash flow and size is large Once cash

flow is scaled by firm size, the correlation coefficient falls

to almost zero This explains why we have decided to adopt

scaled cash flow instead of cash flow s a measure of firm

liquidity

We employ the leverage ratio (LEV) as a stock measure

of indebtedness Following the empirical literature, LEV is

computed as total (non-equity) liabilities/total assets.!5

12 The ATECO classification is released by the Italian Statistical Office and

by and large is equivalent to the European NACE 1.1 taxonomy

13 Among others, recent studies using the same filter are Fagiolo and

Luzzi (2006) and Bottazzi et al (2011)

14 The original dataset containing information on financial and real vari-

ables consists of a sample of about 146.000 firm-year observations The

sample drastically shrinks to 96,883 observation when firms with less

than 2 employees and less than 1 million sales are discarded Once suspi-

cious growth rates are removed, we only keep observations with complete

records, and the final unbalanced filtered sample comprises, on average,

slightly more than 13.300 firms per year We compared the descriptive

statistics of the unbalanced and balanced panel and the two samples

appeared to be very similar Therefore we regard survival bias as a minor

problem here

15 A non-exhaustive list of papers that rely on this measure would cer-

tainly include Lang et al (1996), Heshmati (2001), Oliveira and Fortunato

The descriptive statistics in Table Al in appendix show

that scaled cash flow fluctuates around 0.07.!° It is worth

noting a negative trend from 1999 onwards with cash flow falling more markedly in 2003 together with real GDP growth.!” Cash flow is a largely pro-cyclical variable and, not surprisingly, its evolution over time follows those of average growth rates The worsening of cash-flow avail- ability towards the end of the time-span is confirmed by the percentage of firms exhibiting negative cash-flow In fact this is around 3 percent in the first years but reaches

7 percent in 2003 The average for the leverage ratio is around 75 percent.!® This is substantially higher than is observed in countries such as the US Berger and Udell (1998) report that the average value for US small medium enterprises is slightly higher than 50 percent Michaelas

et al (1999) find a leverage ratio of slightly more than 40 percent in the UK for small business enterprises operat- ing in all sectors Nonetheless, Italy is not the only country whose firms exhibit a high reliance on debt.!9 For instance, similar levels of leverage are found by Honjo and Harada (2006) working on Japanese manufacturing firms, Huynh and Petrunia (2010) using Canadian data on manufactur- ing firms, Guariglia et al (2008) on Chinese (state-owned) firms, Heshmati (2001) on Swedish data, and Oliveira and Fortunato (2006b) on Spanish firms A negative trend is rather evident in the evolution of the leverage ratio This is likely to be due to the fact that over the period 1998-2003 the tax-advantage of debt reduced in Italy as new limits on the amount of passive interests deductible were introduced

with a new system based on DIT — Irap.2°

5 Empirical analysis Following Guariglia et al (2008), we initially estimated a dynamic growth model, augmented for cash-flow Firm growth rate was computed as:

model reads as follows:

GROWTH; ; = 8 GROWTH; ;¢-1 + Bo SCFi.t—1

(2006b, 2008), Guney et al (2007), Guariglia et al (2008), Aiavazian et al (2005), Honjo and Harada (2006) and Huynh and Petrunia (2010)

16 This level of scaled cash-flow is quite in line with previous evidence from other European countries See Fagiolo and Luzzi (2006), Carpenter and Petersen (2002), Oliveira and Fortunato (2006a), Michaelas et al (1999) and Hernando and Martinez-Carrascal (2003)

17 Real GDP growth constantly fell from 3.7 in 2000 to 0 in 2003 Source: http://www.imforg

18 This finding is in line with previous works on the financial structure

of non-financial firms in Italy For instance, see Di Majo et al (2005), Rajan and Zingales (1995) or Guiso (2003)

13 Tt is well known that in Italy it is very hard to raise external capital, so that firms are forced to resort to debt finance Venture Capital investment

in Italy in 2000 was 7 percent of GDP while in the US it was up to 78 per- cent (Bottazzi and Da Rin, 2002) In a recent research study performed by Capitalia, only 0.7 percent of firms had a capital structure characterized

by a shared ownership Furthermore, venture capital amounts to only 2 percent of such a tiny subset L’Associazione Italiana degli Investitori Isti- tuzionali nel Capitale di Rischio (Aifi) reports that venture capital operation were remarkably few and even decreasing in the period considered, from

646 in 2000 to 301 in 2002

20 See Bontempi et al (2004) for a detailed review of the tax-reform.

where the error term 7;; may contain both unobservable

time-invariant individual effects (œ;), and an idiosyncratic

error I.1.D (€;;), that is: nj; =a; + €¡; Following earlier stud-

ies, such as Oliveira and Fortunato (2006a), Fagiolo and

Luzzi (2006) or Evans (1987), we first concentrate on

growth in terms of firm employment Employment is a con-

venient choice because data are readily available in real

terms More importantly, it is certainly a variable of inter-

est because it is the main target of labor market policies

Employment can be seen as an input in the production

process, and we complement the analysis by also checking

the robustness of our findings with respect to an output

measure of growth To this end, as in Huynh and Petrunia

(2010), we use sales growth as an alternative indicator of

firm performance.?!

The econometric specification in all of our regres-

sions features the variable CONTROLS, which includes time

dummies to control for business cycle and macroeconomic

effects, sectoral dummies to control for industry effects and

firm localization.22 For the reason discussed in Section 2,

we also add to our set of controls some firm-specific char-

acteristics such as age and size.”3

Following, the literature, our preferred estimation

method consists of the system GMM system.”4 The choice

can be justified on three grounds: first, these models con-

trol for individual fixed effects by estimating the equation

both in levels and in first differences Second, the GMM sys-

tem is also known to yield consistent coefficient estimates

in the presence of endogenous regressors.*° More specif-

ically, endogeneity is tackled by instrumenting regressors

with their own lagged levels for the first-differenced equa-

tion and with their own lagged first differences for the

equation in levels Finally, the GMM system should be

preferred to the difference GMM because our explanatory

variables are heavily persistent and show high levels of

serial correlation In this case, lagged variables in levels

are poor instruments of the variables in first differences

Adding the levels-equation and using lagged differences

of the explanatory variables as instruments significantly

improve the efficiency of the model by reducing the weak-

instrument bias

We check the validity of instruments using a Hansen J-

test for over-identifying restrictions and the test for serial

21 Sales growth is log difference of total sales measured at constant

prices

22 Sectoral dummies are defined as the first two digits in ATECO classifi-

cation The ATECO sector classification mirrors to a great extent the NACE

one Firms localization is captured by adopting a set of dummy variables,

corresponding to geographical macro areas North-East, North-West, Cen-

ter and South of Italy

23 We initially only control for age The results are reported in Table 1

They do not change qualitatively when size is added as control The esti-

mates are set out in Table A3 (columns I and II) in the appendix

24 The estimator was fully developed by Blundell and Bond (1998) For

an application of this econometric method on firm-level data, see Oliveira

and Fortunato (2006a), D’Espallier and Guariglia (2009) and Coad (2007)

25 Endogeneity can arise from reverse causation For instance, one could

argue that cash-flow is clearly endogenous because high-growth firms

will generate high profits and therefore high cash-flow GMM estimators

account for this potential endogeneity bias and have been regularly used

in the financing constraints literature See, for example, Bond et al (2003)

and D’Espallier and Guariglia (2009)

correlation First-order and second-order serial correlation

in the first-differenced residuals is tested using m1 and m2 statistics The validity of the dynamic model is ver- ified if there is no second-order serial correlation in the differenced residuals (meaning that the €;; are serially uncorrelated) and the J-test accepts the null of instruments validity Both conditions appear to be valid in our model The estimates of the baseline model are presented in

Table 1 column 1.76

In line with similar studies (Fagiolo and Luzzi, 2006; Oliveira and Fortunato, 2006a), we find that cash flow is positively correlated with firm growth

The point estimate (0.183 in column! in Table 1) implies that the elasticity evaluated at sample means is around 0.88, suggesting that a 10 percent increase in cash-flow brings about a 8.8 percent increase in employment growth

In order to test our hypotheses, we proceed as fol- lows: first we estimate the same model adding LEV as explanatory variable As we believe in a inverted U-shape relationship between employment growth and leverage,

we add a quadratic term LEV2, to capture possible non-

linearities.27

GROWTH; + = By GROWTH; ¢_1 + BoSCF; t-4 + B3LEV; +4

+aLEVỆ,_¡ + ÿ CONTROLS,_ + 1c — (2)

Our Họ, consistently with the hypothesis outlined at the end of Section 3, ¡is that Ø64>0 and Ø6s<0, and the non- linear effect may be so strong that the marginal effect switches sign Table 1 column II reports the estimates for this model-specification The results confirm that the effect

of debt-exposure is non-linear over the support of LEV In more detail, we see that LEV initially engenders a positive effect for firms but such influence turns negative for other highly indebted firms We can compute the marginal effect

OEMP _GRI/OLEV, and this becomes negative for values of LEV larger than 0.642.2° Note that only about 25 percent of

firms exhibit a leverage ratio below this threshold In other words, most firms are to some extent debt-constrained in that their current level of debt proves to be detrimental to growth

In the previous section, we discussed the possible interactions among cash-flow, leverage and investment opportunities, and we made the point that an omitted vari- able bias could arise simply because we do not control

26 In Table A2 in appendix, one can compare the GMM-system estimates with the pooled OLS and fixed-effect estimates In dynamic panels with endogenous variables, it has been shown that OLS and fixed-effects yield biased (upward and downward, respectively) estimators The GMM esti- mate of the coefficient on the lagged dependent variable is found to be

in between the OLS and the fixed-effects estimates, and this suggests no serious finite sample bias A coefficient attached to lagged growth close to

or higher than the FE estimates could also raise doubts as to the strength

of instruments (see Carpenter and Guariglia, 2008) This does not seem to

be the case here

27 This specification follows Guney et al (2007), who examine the non- linear effect of leverage on cash-holdings

28 This threshold has been computed using the point estimates displayed

in column II of Table 1

Table 1

Main regression table part

EMP _GRit-1 —0.041”” (0.013) —0.041”” (0.013) —0.042””(0.013) —0.041”” (0.013) —0.041”” (0.013)

SChịt | 0.183" (0.058) 0.17777(0.059) 0.174” (0.059) 0.107” (0.045)

(SCF x HighLEV);¿_¡

(SCF x LowLEV);¿_¡

LEV; 4-1 0.606" (0.191) 0.620” (0.1 90)

interacted with

ind dummies

Number of firms 36,627 36,627 36,627

Hansen Jj-test [0.501] [0.280] [0.925]

(P-value)

m2(P-value) [0.795] [0.751] [0.748]

0.374” (0.079)

~0.574"" (0.150) —0.460””(0.122) -—0.4717” (0.165) -0.563” (0.165)

Notes: All results in this table are obtained using a system GMM estimator The figures reported in parentheses are robust standard errors Instruments

in all columns are lags two or three of all regressors except age in the differenced equation, and lag one or two of the difference of all regressors except age in the levels equation In column V, we report the estimates obtained by adding further lags as instruments Lagged age is assumed to be exogenous Firm’s age, time, sectoral and geographical dummies are used as controls The m1 and m2 statistics are tests for first and second-order serial correlation

in the first-differenced residuals and they are asymptotically distributed as N(O, 1) under the null of no serial correlation; P-values are reported in square brackets The Hansen’s J statistics, under the null of instrument validity, is asymptotically distributed as a x? with degrees of freedom equal to the number

of instruments less the number of parameters J is employed to test the validity of overidentifying restrictions; P-values are reported in square brackets

*P<0.1

7” P<0.05

7” P<0.01

for future investment opportunity The positive coefficient

attached to cash-flow could simply pick up this bias, and

hence it would not be informative about the extent of finan-

cial constraints

Following Carpenter and Guariglia (2008) and Guariglia

et al (2008), we attempt to control for investment opportu-

nities as follows We use the time dummies interacted with

industry dummies as our proxy These dummies account

for all time-varying demand shocks at the industry level

One can think of this as an indirect way to control for invest-

ment opportunities, or more in general demand factors

In Table 1 column III we show the results on Eq (2)

augmented to control for investment opportunities The

coefficient attached to cash flow remains positive and

significant, even after accounting for investment opportu-

nities, and the size of the coefficient is very similar We

conclude that the investment opportunities are not a great

source of bias The same applies to the leverage ratio

As discussed at the end of Section 2, we now want to

assess the extent to which cash flow sensitivity depends

on the level of leverage

To this end, we also add two interaction terms

extent to which being highly leveraged makes the firm

more sensitive to internal cash-flow.?9

Our third and final specification takes the following

form:

22 More specifically, LowLEV is a dummy variable that takes value 1 if

leverage is below the median and 0 otherwise HighLEV takes value 1 if

leverage is above the median and zero otherwise This empirical approach

is similar to that used by Carpenter and Guariglia (2008) and others

+ B3 (SCF x LowLEV); +1 + Ba LEV;_1 + Bs LEV? 4

+ € CONTROLS; ¢-1 + Nit (3)

Our Hg is that $2 > 63, so that the marginal effect of cash- flow on growth increases with leverage Table 1 column IV reports the estimated coefficients and indeed shows that the more firms are leveraged, the more growth is sensitive

to cash-flow.39

We also performed a number of robustness checks and the results are presented in Table A3 in appendix First, we re-estimate the model adding size to the set of controls and we do not find any significant difference.?! Second, we check how cash flow sensitivity changes with leverage by stratifying the sample in three subgroups We construct three dummy variables as follows: LEVq; takes value 1 if firm’s leverage is in the bottom 25 percent of the distribu- tion LEVqy_);7 is equal to 1 when firm’s leverage falls into the second or third quartile of the distribution, and finally LEVqyy is set equal to 1 for firm being in the top quartile of the leverage distribution As before, we let these dummies interact with cash flow and the estimates are reported in column III of Table A3 Results are consistent and provide even stronger evidence that cash flow sensitivities increase with leverage As will be seen, the firms in the bottom

30 In column V of Table 1, we also report the estimates obtained with a larger set of instruments The results are qualitatively the same

31 Size at time t—1 is defined as log(SALES);.-1

Low Leverage —-—-— Medium Leverage .rem High Leverage

Fig 1 Cash-flow distribution conditional to leverage

quartile of the leverage distribution are not sensitive to

cash flow at all, and such firms are also those for which

leverage appears to exert a positive effect on growth Lastly,

we also check the estimates using the unbalanced panel

The results, displayed in columns IV and V in Table A3,

remain consistent.32

Highly indebted firms are not only more sensitive to

cash-flow, but they are also endowed with less internal

funds, as documented in Fig 1.37

It is interesting to compare these results with those

obtained using sales growth as dependent variable

Table 1 columns VI and VII show that the previous

findings carry over when we measure growth with

respect to sales Two remarks are in order First, we

see that sales growth is somewhat less sensitive to

cash flow than employment growth Second, the critical

threshold above which the marginal effect of lever-

age becomes negative is found at an even lower value

(0.53)

6 Conclusions

In this paper, we have provided a unified framework to

jointly assess the debate on cash-flow sensitivities and the

debt-growth nexus at firm level More in detail, two main

hypotheses have been discussed and verified empirically

First, we believe that the degree of financial constraints

increases with leverage and therefore more indebted firms

should exhibit higher growth-cash flow sensitivity Second,

we have examined the channels through which leverage

can impact on growth with particular emphasis on the role

of leverage for unlisted small medium enterprises In this

32 By unbalanced panel we mean a sample in which the number of time

periods available is not necessarily the same for all firms and the points

in time to which the observations correspond may also be different We

also restrict the sample to firms that are continuously observed with no

gaps for at least 4 contiguous observations

33 In line with our robustness check, low-leverage firms are those in the

first quartile of the leverage distribution, medium leverage firms are those

in the second, and third quartile and high-leverage firms are those in the

top quartile

world, leverage represents the primary means of exter- nal financing and as such it should favor firm’s expansion and ultimately promote growth Nonetheless, there may

be a critical level above which excessive debt exposure can trigger negative dynamics induced by debt overhang and liquidity effects that act as a drag on growth Thus

we maintain that a non-linear inverted U-shape relation- ship is likely to be detected between growth and leverage Using a large sample of Italian manufacturing firms tracked over the period 1998-2007, we have found robust evidence that there is a non-linear impact of debt on growth As hypothesized, leverage initially exerts a positive effect of growth Our analysis shows that only about 20-25 per- cent of firms have a level of indebtedness that turns out

to be growth-enhancing The estimates suggest that a tip- ping point for leverage does exist, and it stands at around 0.65 and 0.53 for the employment growth regression and sales growth regression respectively Above this threshold, debt constrains growth, and this is found to be the case for the vast majority of firms in our sample We think of these firms as being debt-constrained Overall, one can charac- terize fragile firms as heavily indebted and endowed with relatively little cash-flow This lack of cash-flow is more likely to be binding and explains why such firms appear to

be more sensitive to internal funds Our results are robust

to measuring growth in terms of both sales and number

of employees, to controlling for firm’s specific character- istics and investment opportunities, and to using either

a balanced or an unbalanced panel The message is that Italian firms are too indebted, and the policy implication

is that efforts should be made in order to ease the access

to financing options based on external equity and venture capital These interventions are likely to be most effec- tive if targeted on highly leveraged firms Summing up, the present analysis contributes to the existing literature

in two ways: first it sheds some light on the complex non- linear financial structure-growth nexus Second, it enters the debate on the interpretation of cash-flow sensitivity by providing evidence that, within this framework, such dif- ferences in sensitivities are indeed meaningful and can be viewed as the result of some firms being constrained both internally and externally There is much scope for further research For example, it would be interesting establish the extent to which this inverted U-shape between growth and leverage is detected in other industries or countries Provided that a tipping point exists, it would be impor- tant to understand what factors account for its empirical value

Acknowledgements

I wish to thank Alessandro Ghillino for his support with the data, and the members of the Research Office

“Pianificazione, Strategie e Studi” at Unicredit Research Centre I am indebted to Frank Windmeijer for invaluable help on econometric issues Thanks go to Giorgio Fagi- olo, Silvia Giannangeli, Graziano Coller, Stefano Schiavo, Alex Coad, Claudia Vittori, Fabio Monteforte and Marcela Umana-Aponte for their comments at various stages of the research process The paper has also greatly benefited from the comments made by three anonymous referees Finally,

Appendix A

See Tables A1-A3

Table Al

Descriptive statistics

EMP_GR

SALES _GR

LEV

SCF

Table A2

Baseline model

See explanatory notes to Table 1 *P<0.1 **P<0.05

7” P<0.01.

Table A3

Additional regressions

with ind dummies

Notes: See also explanatory notes to Table 1 All results in this table are obtained using a system GMM estimator Estimates reported in columns | to III are obtained using the balanced panel Columns IV and V show the results obtained with the unbalanced panel Columns I and II report the estimates obtained when firm’s size was also included as an additional control *P< 0.1

™ P<0.05

™ P<0.01

References

Audretsch, D., Elston, J., 2002 Does firm size matter? Evidence on

the impact of liquidity constraints on firm investment behaviour

in Germany International Journal of Industrial Organization 20,

1-17

Aiavazian, V.A., Ge, Y., Qiu, J., 2005 The impact of leverage on firm

investment: Canadian evidence Journal of Corporate Finance 11,

277-291

Berger, A.N., Udell, G.F., 1998 The economics of small business finance:

the roles of private equity and debt markets in the financial growth

cycle Journal of Banking and Finance 22, 613-673

Bernanke, B.S., Gertler, M., Gilchrist, S., 1996 The financial accelerator

and the flight to quality The Review of Economics and Statistics 78,

1-15

Blundell, R., Bond, S., 1998 Initial conditions and moment restrictions in

dynamic panel data models Journal of Econometrics 87, 115-143

Bond, S., Elston, J., Mairesse, J., Mulkay, B., 2003 Financial factors and the

investment in Belgium, France, Germany, and the United Kingdom: a

comparison using company panel data The Review of Economics and

Statistics 85, 153-165

Bontempi, E., Giannini, S., Golinelli, R., 2004 Corporate taxation and its

reform: the effects on corporate financing decisions in Italy Working

Papers 501, Dipartimento Scienze Economiche, Universita’ di Bologna

Bottazzi, L, Da Rin, M., 2002 Venture capital in Europe and the

financing of innovative companies Economic Policy 17, 229-

270

Bottazzi, G., Secchi, A., Tamagni, F., 2008 Productivity profitability and

financial performance Industrial and Corporate Change, Oxford Uni-

versity Press 17, 711-751

Bottazzi, G., Grazzi, M., Secchi, A Tamagni, F., 2011 Financial and

economic determinants of firm default Journal of Evolutionary Eco-

nomics 21, 373-406

Cabral, L., Mata, J., 2003 On the evolution of the firm size distribution:

facts and theory American Economic Review 93, 1075-1090

Carpenter, R.E., Petersen, B.C., 2002 Is the growth of small firms con-

strained by internal finance? The Review of Economics and Statistics,

MIT Press 84, 298-309

Carpenter, R.E., Guariglia, A., 2008 Cash flow investment and investment

opportunities: new tests using UK panel data Journal of Banking and

Finance 32, 1894-1906

Chirinko, R., Schaller, H., 1995 Why does liquidity matter in investment

equations? Journal of Money Credit and Banking 27, 527-548

Cleary, S., 1999 The relationship between firm investment and financial

status Journal of Finance 54, 673-692

Cleary, S., Povel, P., Raith, M., 2007 The U-shaped investment curve:

theory and evidence Journal of Financial and Quantitative Analysis,

Cambridge University Press 42, 1-39

Coad, A., 2007 Testing the principle of growth of the fitter: the relationship between profits and firm growth Structural Change and Economic Dynamics 18, 370-386

Cooley, T.F., Quadrini, V., 2001 Financial markets and firm dynamics American Economic Review 91, 1286-1310

D’Espallier, B., Guariglia, A., 2009 Does the investment opportunities bias affect the investment-cash flow sensitivities of unlisted SMEs? Open Access publications from Katholieke Universiteit Leuven, Katholieke Universiteit Leuven

Devereux, M., Schiantarelli, F., 1990 Investment, financial factors and cash flow: evidence from U.K panel data In: Hubbard, G (Ed.), Information, Capital Markets and Investment University of Chicago Press

Di Majo, A., Pazienza, M.G., Tirberti, A., 2005 Scelte di Finanziamento delle Imprese Minori Teorie e Analisi del Caso Italiano Working Paper 7, DISEFIN-Dipartimento di Scienze Economiche e Finanziarie, Univer- sity of Genova

Evans, D., 1987 The relationship between firm growth, size and age: estimates for 100 manufacturing industries Journal of Industrial Eco- nomics 35, 567-581

Fagiolo, G., Luzzi, A., 2006 Do liquidity constraints matter in explaining firm size and growth? Some evidence form the Italian manufacturing industry Industrial Corporate Change 15, 1-39

Fazzari, S.M., Hubbard, R.G., Petersen, B.C., 1988 Financial constraints and corporate investment Brooking Papers on Economic Activity 1, 141-206

Fazzari, S.M., Hubbard, R.G., Petersen, B.C., 2000 Financial constraints and corporate investment: response to Kaplan and Zingales NBER Work- ing Papers 5642, National Bureau of Economic Research, Inc Gertler, M., Gilchrist, S., 1994 Monetary policy business cycles and the behavior of small manufacturing firms Quarterly Journal of Eco- nomics 109, 309-340

Gilchrist, S., Himmelberg, C.P., 1995 Evidence on the role of cash flow for investment Journal of Monetary Economics 36, 541-572

Guariglia, A., Liu, X., Song, L., 2008 Internal finance and growth: microe- conometric evidence on Chinese firms IZA Discussion Papers 3808, Institute for the Study of Labor (IZA)

Guiso, L., 2003 Small Business Finance in Italy EIB PAPERS 2, vol.7 Guney, Y., Ozkan, A., Ozkan, N., 2007 International evidence on the non-linear impact of leverage on corporate cash holdings Journal of Multinational Financial Management 17, 45-60

Hernando, I., Martinez-Carrascal, C., 2003 The impact of financial vari- ables on firms real decisions: evidence from Spanish firm-level data Banco de Espana Working Papers 0319

Heshmati, A., 2001 The dynamics of capital structure: evidence from Swedish micro and small firms SSE-EFI Working Paper Series in Eco- nomics and Finance, 440

Honjo, Y., Harada, N., 2006 SME policy financial structure and firm growth: evidence from Japan Small Business Economics 27, 289-300.