-
European Business ReviewDeterminants of corporate debt maturity in Latin AmericaPaulo Renato Soares Terra
Article information:To cite this document:Paulo Renato Soares Terra, (2011),"Determinants of corporate debt maturity in Latin America", EuropeanBusiness Review, Vol. 23 Iss 1 pp. 45 - 70Permanent link to this document:http://dx.doi.org/10.1108/09555341111097982
Downloaded on: 24 February 2015, At: 08:01 (PT)References: this document contains references to 53 other documents.To copy this document: [email protected] fulltext of this document has been downloaded 1238 times since 2011*
Users who downloaded this article also downloaded:Wenjuan Ruan, Grant Cullen, Shiguang Ma, Erwei Xiang, (2014),"Ownership control and debt maturitystructure: evidence from China", International Journal of Managerial Finance, Vol. 10 Iss 3 pp. 385-403http://dx.doi.org/10.1108/IJMF-06-2013-0064Michael R. Powers, Joshua Abor, (2007),"Debt policy and performance of SMEs: Evidence fromGhanaian and South African firms", The Journal of Risk Finance, Vol. 8 Iss 4 pp. 364-379 http://dx.doi.org/10.1108/15265940710777315Jos Paulo Esperana, Ana Paula Matias Gama, Mohamed Azzim Gulamhussen, (2003),"Corporate debtpolicy of small firms: an empirical (re)examination", Journal of Small Business and Enterprise Development,Vol. 10 Iss 1 pp. 62-80 http://dx.doi.org/10.1108/14626000310461213
Access to this document was granted through an Emerald subscription provided by 505203 []
For AuthorsIf you would like to write for this, or any other Emerald publication, then please use our Emerald forAuthors service information about how to choose which publication to write for and submission guidelinesare available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.comEmerald is a global publisher linking research and practice to the benefit of society. The companymanages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well asproviding an extensive range of online products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committeeon Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archivepreservation.
*Related content and download information correct at time of download.
Dow
nloa
ded
by E
scue
la d
e A
dmin
istra
cion
de
Neg
ocio
s par
a G
radu
ados
ESA
N A
t 08:
01 2
4 Fe
brua
ry 2
015
(PT)
-
Determinants of corporatedebt maturity in Latin America
Paulo Renato Soares TerraSchool of Management, Universidade Federal do Rio Grande do Sul,
Porto Alegre, Brazil
Abstract
Purpose The purpose of this paper is to test the main theories of corporate debt maturity in amulti-country framework, in an attempt to understand country-specific constraints.
Design/methodology/approach Dynamic panel data analysis estimated by the generalizedmethod of moments, techniques that account properly for cross-section and time series variationallowing for dynamic effects.
Findings There is a substantial dynamic component in the determination of a firms maturitystructure; firms face moderate adjustment costs towards its optimal maturity, and the determinants ofmaturity structure and their effects are similar between Latin American countries and the USA;and there is a partial empirical support for each of the theoretical hypotheses tested.
Research limitations/implications Firm ownership, accounting standards, financial marketdepth, and the degree of supervision on financial reporting may vary across countries, which may affectthe quality and consistency of some variables.
Practical implications Firms face costs in adjusting the maturity of their debt, which gives suchdecision a long-term character, and the determinants of debt maturity do not seem very sensitive to acountrys business and financial environment.
Originality/value The paper focuses on a sample of developing countries that have so far beenignored in empirical studies, employs empirical techniques that account properly for cross-section andtime series variation, and the model allows for dynamic effects that have seldom been considered inprevious research.
Keywords Debts, Capital structure, Data analysis, South America
Paper type Research paper
1. IntroductionThe breakthrough work of Modigliani and Miller (1958, henceforth MM) laid the basisfor what is conventionally regarded as the modern corporate finance. In their influentialpaper and the ones that followed (Miller and Modigliani, 1961; Modigliani and Miller,1963; Miller, 1977), these authors laid down the conditions under which the firm wouldbe largely indifferent to the sources of its financing. In the past 50 years, several papershave explored both theoretically and empirically the implications of their famousPropositions I, II and III. Capital structure and dividend policy are perhaps the most
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/0955-534X.htm
The author is thankful to Melcia Ferri and Debora Novak de Souza and Cristiano Kessler Wagnerand Lus Felipe Siebel for general research assistance for this paper. Comments on an earlierversion of this paper from Jack Glen, Jan J. Jorgensen, Cesario Mateus, Andrea M.A.F. Minardi,Antonio Z. Sanvicente, and Joao Zani, as well as from referees and participants in the 2004Meetings of the Brazilian Finance Society and the Brazilian Academy of Management, as well asthe 2005 Meetings of the American Accounting Association and Financial ManagementAssociation are much appreciated. Any remaining errors are the authors responsibility.
Corporate debtmaturity
45
Received February 2009Revised May 2009
Accepted December 2009
European Business ReviewVol. 23 No. 1, 2011
pp. 45-70q Emerald Group Publishing Limited
0955-534XDOI 10.1108/09555341111097982
Dow
nloa
ded
by E
scue
la d
e A
dmin
istra
cion
de
Neg
ocio
s par
a G
radu
ados
ESA
N A
t 08:
01 2
4 Fe
brua
ry 2
015
(PT)
-
studied issues in corporate finance. Much less attention however has been devoted to thematurity structure of the firms financing.
The financial turmoil that began in mid-2007 and scaled up in late 2008 has spreadworldwide. Its consequences over the credit and liquidity of firms are being felt indeveloped and emerging countries alike. Indeed, Campello et al. (2009) document that theimmediate effects of the financial crises has made financially constrained firms from theUSA, Europe, and Japan to burn through cash reserves, to run on their bank credit lines,cut back on capital investment, employment, research and development spending,marketing expenditures, and dividends, and to sell assets to obtain cash. Although nosuch survey has been published so far focusing firms in emerging markets, it is wellknown that they face generally harsher financial constrains than similar firms indeveloped markets. So, it is fair to conjecture that the dire effects of this crisis may be evenmore pronounced in these countries. In such context, understanding how firms managetheir debt becomes thus more than an academic question to become a real-world problemfor practicing managers.
This paper contributes to the existing body of knowledge in several ways. Here, I testa few theories of debt-maturity structure in a multi-country framework, in an attempt tounderstand country-specific differences. I focus on a sample of developing countries thathave so far been ignored in empirical studies. Moreover, I do so by employing empiricaltechniques that account properly for cross-section and time series variation. Also,the model allows for dynamic effects that have seldom been considered in previousresearch. Finally, I compare my results for Latin American countries to a sample of firmsfrom the USA.
My main findings are that there is a substantial dynamic component in thedetermination of a firms maturity structure, firms face moderate adjustment coststowards its optimal maturity, and the determinants of maturity structure and theireffects are similar between Latin American countries and the USA, despite obviousdifferences in the financial and business environments of these countries. The study alsofinds some empirical evidence for each theoretical hypothesis tested, although notheoretical proposition alone is able to explain the maturity decision.
The remaining of the paper is structured as follows: the next section presents thetheoretical framework, while Section 3 details the methodology, presents the datasources, and describes the variables used in the empirical model. Section 4 reports andcomments the estimation results. Section 5 concludes the paper.
2. Theoretical frameworkA number of explanations for the maturity structure of corporate debt have been putforward. The main criticism that can be made of this body of literature is that it does notemerge from a general equilibrium theory, but as a set of partial explanations that havenot been unified into one single theory. I do not intend to tackle such an ambitious task inthis paper. However, in order to concisely understand the many and disperse theoreticalcontributions to the question of an optimal maturity, I classify the literature into fourmajor groups: the tradeoff hypothesis, the agency hypothesis, the signaling hypothesis,and the maturity-matching hypothesis. Of course, such simplification is open tocriticism, but my classification is ample enough to encompass most theoretical workdone so far, yet discriminating enough to point out the fundamental differences betweeneach group of hypotheses.
EBR23,1
46
Dow
nloa
ded
by E
scue
la d
e A
dmin
istra
cion
de
Neg
ocio
s par
a G
radu
ados
ESA
N A
t 08:
01 2
4 Fe
brua
ry 2
015
(PT)
-
Theoretical explanations for the choice of corporate debt maturity are already impliedin MMs original paper, but are eventually formalized by Stiglitz (1974). MMs paper doesnot consider a multi-period setting, and Stiglitz (1974) provides a rigorous analysis of theMM model in such circumstances. His conclusions are that, under a fairly general set ofconditions (absence of taxation, transaction costs, bankruptcy costs, and other frictions),the maturity choice of the firm is irrelevant, just as MMs findings regarding the firmsleverage ratio under the same conditions. Of course, once one departs from the idealworld of the financial economists[1], such frictions matter, and therefore the maturitydecision would influence the firms valuation just as would the set of other financialpolicies. A large family of hypotheses explores the tax-based, bankruptcy costs andtransaction costs approaches in order to offer an explanation for the maturity choice.
Arguments for the tradeoff hypothesis are based on the proposition that the optimalmaturity of debt is determined by the tradeoff between the costs to rollover short-termdebt vis-a`-vis the usually higher interest rate bore by long-term debt. In many senses, thearguments rely on explicit transaction costs of different kinds of debt such as flotationand rollover costs as well as tax-shield benefits and implicit bankruptcy costs. Thetax-based explanation suggested by Brick and Abraham Ravid (1985) and Brick andAbraham Ravid (1991) are perhaps the best known examples.
Another whole family of hypotheses derives from the asymmetric information problemformalized by Jensen and Meckling (1976) and extended by Myers (1977). In this case, thematurity structure is yet another instrument that firms can use in order to solve the agencyproblems faced by the various stakeholders of the firm. The agency hypothesis suggeststhat firms choose the optimal debt maturity in order to solve the information asymmetrythat gives rise to the underinvestment (Myers, 1977; Myers and Majluf, 1984) and/oroverinvestment (Jensen and Meckling, 1976; Jensen, 1986) problems. Barnea et al. (1980)offer an explanation for the debt-maturity choice as well as for complex financialcontracting based on market failure in resolving agency problems costlessly.
Also within the asymmetric information mindset, the maturity structure can alsobe regarded as a means of overcoming the adverse selection problem (Akerlof, 1970) interms of providing a credible signal to the market, alongside the general lines suggestedby Ross (1977). The signaling hypothesis is therefore also rooted on informationasymmetry arguments, but suggests that the maturity choice as for a number of otherpublicly known corporate decisions is used by managers as a way to conveyinformation to the market thus reducing the firms cost of capital. Within this group issituated Flannerys (1986) proposition that risky debt maturity is a valid signal iftransaction costs are positive, because high-quality firms can signal their true quality.
Finally, there is the textbook rule-of-thumb that firms should match the maturity oftheir liabilities to the maturity of their assets. This intuitive recommendation seems tohave emerged from the practitioners experience before it had been rationalized intotheory, relying mainly on liquidity risks, inefficient liquidation, and goods market cyclearguments. An argument combining the agency and signaling hypotheses can be used inorder to explain why the maturity of liabilities should match the maturity of assets. Indeed,several finance textbooks allude to this rule when discussing the investment and financingdecisions (Ross et al., 2002; Brealey and Myers, 2003). Hart and Moore (1994) propose oneexplanation based on the asymmetry of information regarding the entrepreneurs trueintentions. Maturity matching in this case would signal the entrepreneurs commitment toabide by his intentions. Alternatively, firms would match the maturity of their liabilities
Corporate debtmaturity
47
Dow
nloa
ded
by E
scue
la d
e A
dmin
istra
cion
de
Neg
ocio
s par
a G
radu
ados
ESA
N A
t 08:
01 2
4 Fe
brua
ry 2
015
(PT)
-
to that of their claims in order to avoid a liquidity problem that would trigger inefficientliquidation of the firm. Excess liquidity, on the other hand, has a high opportunity cost andis also inefficient in the firms perspective. Diamond (1991) proposes a model of maturitychoice in which highly and poorly rated firms choose short-term debt while middle-ratedfirms choose to match the maturity of their debt to the timing of their operating cash flows.In Diamonds (1991) model, poorly rated firms have no choice other than exposethemselves to premature liquidation because of moral hazard concerns from the lenders,while highly rated firms choose to borrow short-term because they expect good news toarrive and therefore obtain better long-term financing deals. The maturity-matchinghypothesis is also supported by Emerys (2001) insightful paper, which argues that firmsmatch the maturity of their assets and liabilities as a means to avoid the term premium ininterest rates. His arguments are based on the demand cycle in the goods market and thereduction in the firms long-run marginal costs achieved by optimal use of short-term debt.
The predictions of these various theories regarding the effects of each determinant ofmaturity structure are summarized in Table I. It is clear that discriminating amongst thehypotheses is difficult, because in many aspects they lead to the same prediction, and inmany cases the hypotheses are silent about the effect of a particular variable.
Determinantfactors
Theoreticalhypothesis
Predicted effect ondebt maturity Empirical proxy Formula
Leverage Agency/matching/tradeoff
Negative/negative/positive
Debt-equity ratio Long-term debt/book equity
Assetmaturity
Matching Positive Asset maturityratio
(Current assets/cost of goodssold) (net fixed assets/depreciation)
Size Agency/signaling
Positive Log of sales Ln(sales)
Growthopportunities
Agency/matching/tradeoff
Negative/positive/positive
Market-to-bookratio
(Book liabilities marketequity)/(total book assets)
Profitability Agency/signaling
Positive/negative Return on Assets Operating income/total bookassets
Business risk Agency/tradeoff
Positive/negative Degree ofoperationalleverage
Sales/operating income
Dividendpolicy
Agency/tradeoff
Negative/positive Dividend yield Dividend per share/shareprice
Liquidity Signaling Positive Current liquidityratio
Current assets/currentliabilities
Tangibility Tradeoff Positive Degree of assetimmobilization
Net fixed assets/total bookassets
Tax effects Agency/tradeoff
Positive/negative Average effectivetax rate
Taxes/taxable earnings
Industry Controlvariable
Undetermined Dummyvariables
0 or 1
Country Controlvariable
Undetermined Dummyvariables
0 or 1
Year Controlvariable
Undetermined Dummyvariables
0 or 1
Table I.Determinants of debtmaturity, theoreticalhypothesis, predictedeffect, and empiricalvariables
EBR23,1
48
Dow
nloa
ded
by E
scue
la d
e A
dmin
istra
cion
de
Neg
ocio
s par
a G
radu
ados
ESA
N A
t 08:
01 2
4 Fe
brua
ry 2
015
(PT)
-
Most empirical studies have concentrated on the USA. Mitchells (1991) and Morris(1992) pioneer studies have taken different empirical approaches to the problem. WhileMorris (1992) investigates the maturity structure of the firms total indebtedness,Mitchell (1991) focuses on the maturity of single-bond issues. These are the two mostcommon empirical approaches in the literature. The first approach is followed byEasterwood and Kadapakkam (1994), Barclay and Smith (1995, 1996), Stohs and Mauer(1996), Johnson (1997), Scherr and Hulburt (2001) and Lyandres and Zhdanov (2007). Thesecond approach is preferred by Mitchell (1993), Guedes and Opler (1996) and Gottesmanand Roberts (2004), the latter investigating the maturity of bank loans. Baker et al. (2003)also investigate bond issues, and in the aggregate, find evidence of market timing ofbond issues.
Table II summarizes the empirical literature and their main findings. As can be seen,a great deal of conflicting evidence has been found on this issue. In general, very littlesupport has been found for the tradeoff hypothesis. There is a considerable amount ofcontroversy regarding the agency costs and signaling hypotheses, while convincingevidence has been found for the maturity-matching hypothesis.
Few studies investigate debt maturity in an international setting. Schiantarelliand Sembenelli (1997) investigate the maturity structure of 604 non-financial firms fromthe UK and 750 non-financial firms from Italy and find support for thematurity-matching hypothesis. Their results are in line with those of Ozkan (2000)who investigates the maturity issue for 429 non-financial British firms in the period1983-1996 and Heyman et al. (2008) who investigate the maturity of 1,132 Belgian smallfirms. Antoniou et al. (2006) study the determinants of debt maturity for a sample of358 French, 582 German, and 2,423 British non-financial firms and find that debtmaturity depends on both firm-specific and country-specific factors, opening thequestion of the degree of influence of each group of factors on the maturity structure.
Larger sets of countries are studied by Demirguc-Kunt and Maksimovic (1999) whoexplored the hypothesis that the financial development of a country determines thematurity of its firms debt. The authors investigate 9,649 non-financial firms from30 countries including developing ones in the period 1980-1991. They find support forthe hypothesis that legal and institutional differences among countries explain a largepart of the leverage and debt-maturity choices of firms. Fan et al. (2008) also study thesubject for 11 industries in 39 countries in the period 1991-2006. Their results largelysupport Demirguc-Kunt and Maksimovics (1999) findings.
To the best of my knowledge, so far, the only study that focused specifically ondeveloping countries is Erol (2004), who analyzed the strategic content of debt maturity in asampleof 15manufacturingsectors from Turkey during theperiod 1990-2000. Theauthorsfindings are that long-term debt is strategic while short-term debt, despite being devoidof strategic content, is associated with financial constraint. In the next section, I describethe methods, variables, and data I employ in order to investigate the debt-maturitystructure of non-financial firms in the seven biggest economies of Latin America.
3. Data, variables, and research methods3.1 Data and variablesAccounting and stock market firm-level data are taken from the Economatica Prow
database (Economatica, 2003). Observations are yearly during the period 1987-2002(subject to availability) and the unit of analysis is each firm. Countries that are the object
Corporate debtmaturity
49
Dow
nloa
ded
by E
scue
la d
e A
dmin
istra
cion
de
Neg
ocio
s par
a G
radu
ados
ESA
N A
t 08:
01 2
4 Fe
brua
ry 2
015
(PT)
-
Th
eore
tica
lfr
amew
ork
Em
pir
ical
stu
die
sH
yp
oth
eses
Mai
nau
thor
sS
tron
gly
sup
por
tsM
oder
atel
ysu
pp
orts
Can
not
sup
por
t
Statictradeoff
Ban
kru
ptc
yco
sts
Flo
tati
onco
sts
Rol
lov
erco
sts
Tax
effe
cts
Tra
nsa
ctio
nco
sts
Mod
igli
ani
and
Mil
ler
(195
8)B
rick
and
Ab
rah
amR
avid
(198
5)
An
ton
iouetal.
(200
6)B
arcl
ayan
dS
mit
h(1
995)
Gu
edes
and
Op
ler
(199
6)S
cher
ran
dH
ulb
urt
(200
1)S
toh
san
dM
auer
(199
6)
Bar
clay
and
Sm
ith
(199
6)G
ued
esan
dO
ple
r(1
996)
Mit
chel
l(1
993)
Ozk
an(2
000)
Agency
costs
Ass
etsu
bst
itu
tion
Jen
sen
and
Mec
kli
ng
(197
6)B
arcl
ayan
dS
mit
h(1
995)
Gu
edes
and
Op
ler
(199
6)H
eym
anetal.
(200
8)C
ontr
acti
ng
cost
sM
onit
orin
gco
sts
Ov
erin
ves
tmen
tU
nd
erin
ves
tmen
t
My
ers
(197
7)B
arn
eaetal.
(198
0)B
arcl
ayan
dS
mit
h(1
996)
Eas
terw
ood
and
Kad
apak
kam
(199
4)G
ued
esan
dO
ple
r(1
996)
Ly
and
res
and
Zh
dan
ov(2
007)
a
Mit
chel
l(1
991)
Mit
chel
l(1
993)
Ozk
an(2
000)
Sch
err
and
Hu
lbu
rt(2
001)
Sto
hs
and
Mau
er(1
996)
Ly
and
res
and
Zh
dan
ov(2
007)
b
Signaling
Asy
mm
etri
cin
form
atio
nC
red
itq
ual
ity
Liq
uid
ity
sig
nal
ing
Ak
erlo
f(1
970)
Dia
mon
d(1
991)
Fla
nn
ery
(198
6)
Got
tesm
anan
dR
ober
ts(2
004)
Hey
man
etal.
(200
8)M
itch
ell
(199
1)
Bar
clay
and
Sm
ith
(199
5)B
arcl
ayan
dS
mit
h(1
996)
Gu
edes
and
Op
ler
(199
6)M
itch
ell
(199
3)O
zkan
(200
0)S
toh
san
dM
auer
(199
6)
An
ton
iouetal.
(200
6)
Maturity
matching
Dem
and
cycl
eD
iam
ond
(199
1)G
ued
esan
dO
ple
r(1
996)
Dem
irg
uc-
Ku
nt
and
Mak
sim
ovic
(199
9)L
iqu
idit
yri
skE
mer
y(2
001)
Hey
man
etal.
(200
8)M
orri
s(1
992)
Ozk
an(2
000)
Sch
err
and
Hu
lbu
rt(2
001)
Sch
ian
tare
lli
and
Sem
ben
elli
(199
7)S
toh
san
dM
auer
(199
6)
Notes:
aU
nd
erin
ves
tmen
th
yp
oth
esis
;bov
erin
ves
tmen
th
yp
oth
esis
Table II.Theoretical frameworkand previous findings inthe empirical literature
EBR23,1
50
Dow
nloa
ded
by E
scue
la d
e A
dmin
istra
cion
de
Neg
ocio
s par
a G
radu
ados
ESA
N A
t 08:
01 2
4 Fe
brua
ry 2
015
(PT)
-
of this study are Argentina, Brazil, Chile, Colombia, Mexico, Peru, and Venezuela(henceforth Latin American 7 or simply LA-7). I also collect data of firms from the USA(henceforth simply US), as a benchmark. The database contains 2,486 firms in total forthese eight countries (1,242 firms from Latin America and 1,244 from the US) over theperiod covered. In this study, I exclude all firms pertaining to the financial industry, suchas financial services and insurance (427 firms), holding and asset managementcompanies (41 firms), and real estate (44 firms), as well as others (seven firms) andnon-classified establishments (four firms). The final sample contains 1,963 firms in total(986 firms from Latin America and 977 from the US). Industry sectors are classified basedon the database documentation (22 industries) and on the North American IndustryClassification System Level 1 Code (20 industries). Additional descriptive data on thefirms activities are available in the database and are used in order to re-classify the firmsin the final 19 industries employed in this research whenever necessary. An overview ofthe number of firms available in the database by country and industry sector is shown inTable III.
From Table III it can be seen that Brazil heavily influences the sample: it has the mostfirms included amongst the Latin American countries and for the longest time period,responding for more than 40 percent of the Latin American sample composition.Venezuela, on the other hand, has little influence on the sample with less than 3 percentof the Latin American firms. Table III also shows that Food and Beverages is thepredominant activity of the Latin American firms with a participation of about 11 percent.Software lies at the other end of the spectrum, with only two firms included. For the US,the predominant sector of activity is the Electronic industry, while Agriculture isrepresented by a single firm.
In this paper, I employ balance sheet data for individual firms with annual periodicity,since balance sheet information for yearly statements are usually more reliable[2].Also, considering the long-term implications of the maturity structure choice, higherfrequency data should not add much to the findings but it might be noisier.
Accounting information in the database is available in local currency (real ornominal) and in US dollars. Since this is a cross-country study, I use figures denominatedin US dollars in order to ease comparisons. In fact, such scaling is irrelevant since mostvariables in this study are ratios. However, a nominal variable such as firm size would begreatly misleading for comparison purposes if stated in local currency.
The dependent variable is a proxy of the maturity of debt carried by each firm,measured in two ways: long-term financial debt over short-term loans plus long-termfinancial debt (Maturity Ratio 1, henceforth simply MR1), and long-term book liabilitiesover total book liabilities, i.e. long-term book liabilities plus current liabilities (MR2).
Strictly speaking, debt-maturity analyses should concentrate on bank loans, bondsand other sources of financial debt. However, trade financing and other short-termoperational liabilities are important sources of funds in many emerging markets, whichcould distort the results if the analyses are limited to strictly defined debt. Hence, theuse of a proxy defined in terms of total liabilities such as MR2 is appropriate.
The dilemma of employing book values versus markets values when studying debtcaters for a lively discussion of its own. On one hand, book values are subject to creativeaccounting and discretionary criteria defined by regulatory authorities. On the otherhand, market values are subject to distortions induced by low liquidity and concentratedtrading in few participants. In this study, I choose book values instead of market values
Corporate debtmaturity
51
Dow
nloa
ded
by E
scue
la d
e A
dmin
istra
cion
de
Neg
ocio
s par
a G
radu
ados
ESA
N A
t 08:
01 2
4 Fe
brua
ry 2
015
(PT)
-
Cou
ntr
y
Arg
enti
na
Bra
zil
Ch
ile
Col
omb
iaM
exic
oP
eru
Ven
ezu
ela
Lat
inA
mer
ica
Lat
inA
mer
ican
firm
s(%
)U
SA
All
firm
s
All
firm
s(%
)P
erio
d
Ind
ust
ry19
90-2
002
1987
-200
219
90-2
002
1992
-200
219
88-2
002
1992
-200
219
92-2
002
1987
-200
219
87-2
002
1994
-200
219
87-2
002
1987
-20
02
Ag
ricu
ltu
re6
123
34
101
484.
91
492.
5C
hem
ical
635
93
79
372
7.3
8916
18.
2C
onst
ruct
ion
314
20
103
032
3.2
1749
2.5
Ele
ctri
city
734
231
010
277
7.8
3711
45.
8E
lect
ron
ic2
201
02
40
292.
914
317
28.
8F
ood
and
bev
erag
es6
3617
423
242
112
11.4
3214
47.
3G
asan
doi
l12
92
20
11
272.
762
894.
5M
ach
iner
y1
150
03
50
242.
433
572.
9M
anu
fact
uri
ng
518
85
28
147
4.8
5097
4.9
Min
ing
513
1410
1126
483
8.4
2610
95.
6P
ulp
and
pap
er4
102
23
02
232.
317
402.
0R
etai
lin
gan
dw
hol
esal
ing
216
183
292
070
7.1
132
202
10.3
Ser
vic
es0
419
316
31
464.
712
817
48.
9S
oftw
are
10
01
00
02
0.2
7779
4.0
Ste
el6
437
210
65
798.
019
985.
0T
elec
omm
un
icat
ion
s3
5910
212
31
909.
149
139
7.1
Tex
tile
335
65
610
469
7.0
1180
4.1
Tra
nsp
ort
and
log
isti
cs1
108
15
01
262.
628
542.
8V
ehic
les
and
par
ts3
230
02
20
303.
026
562.
9A
llfi
rms
7639
516
947
145
126
2898
610
0.0
977
1,96
310
0.0
All
firm
s(%
)3.
920
.18.
62.
47.
46.
41.
450
.249
.810
0.0
LA
-7fi
rms
(%)
7.7
40.1
17.1
4.8
14.7
12.8
2.8
100.
0
Table III.Compositionof the sample
EBR23,1
52
Dow
nloa
ded
by E
scue
la d
e A
dmin
istra
cion
de
Neg
ocio
s par
a G
radu
ados
ESA
N A
t 08:
01 2
4 Fe
brua
ry 2
015
(PT)
-
because the reliability of market-based figures for Latin American firms, especially withrespect to debt valuation, is questionable. Secondary markets in the region are thin, tradeis often infrequent, and data availability is difficult. Given these shortcomings, I findbook values more adequate to the purposes of this research.
Descriptive statistics for MR1 and MR2 are presented in Table IV. It is clear thatmaturity ratios for the US are substantially bigger than for the average Latin Americanfirm. In turn, maturity ratios of LA-7 firms are more volatile (less so for MR2). Mexicanfirms present the larger maturity ratios amongst all Latin American firms. As expected,when trade finance is included in the definition of maturity, the ratios of both samplesdiminish, indicating a bigger dependence of short-term financing. Moreover, for MR2ratios of LA-7 firms become closer to those of their North American peers (althoughstill smaller).
One important aspect to be considered when investigating the debt-maturity choiceof the firm is that it is usually a related decision with the capital structure (amount of debtvis-a`-vis equity) decision. Many empirical studies overlook such aspects; thus, theirresults might be biased.
In order to treat this effect properly, I choose a two-stage strategy to obtain a proxy forcapital structure in which in the first stage the leverage proxy is regressed against the(other) independent variables determining maturity[3] and then, in the second stage, theresiduals of the first stage are introduced as regressors in the maturity equation[4].This way, the leverage effect is taken into account in the maturity equation while notcontaminating it by the capital structure decision since the leverage residuals are byconstruction orthogonal to the remaining independent variables[5].
Countries Obs. Mean Median SD
MR1Argentina 545 0.4181 0.4523 0.3288Brazil 3,598 0.4467 0.4686 0.3100Chile 1,536 0.5014 0.5615 0.3539Colombia 244 0.4651 0.5067 0.3421Mexico 1,236 0.5354 0.6159 0.3280Peru 149 0.3967 0.4063 0.3399Venezuela 146 0.4292 0.4717 0.3112LA-7 7,454 0.4699 0.5059 0.3277USA 4,599 0.7856 0.8844 0.2624MR2Argentina 621 0.3374 0.3076 0.2590Brazil 4,100 0.3786 0.3532 0.2626Chile 1,770 0.4081 0.3910 0.2867Colombia 283 0.3745 0.3776 0.2548Mexico 1,411 0.4326 0.4613 0.2648Peru 1,032 0.2789 0.2452 0.3712Venezuela 175 0.3955 0.4040 0.2217LA-7 9,392 0.3788 0.3609 0.2835USA 5,028 0.5247 0.5769 0.2673
Notes: The table presents the descriptive statistics for each dependent variable in the period1987-2002; LA-7 refers to the pooling together of all firm-level data for Argentina, Brazil, Chile,Colombia, Mexico, Peru, and Venezuela
Table IV.Descriptive statistics
Corporate debtmaturity
53
Dow
nloa
ded
by E
scue
la d
e A
dmin
istra
cion
de
Neg
ocio
s par
a G
radu
ados
ESA
N A
t 08:
01 2
4 Fe
brua
ry 2
015
(PT)
-
Firm-specific determinant factors for the debt-maturity structure are chosen from thoseoften suggested in the extant literature. The set of firm-specific explanatory variablesconsists of the following: leverage, asset maturity, size, growth opportunities,profitability, dividend policy, liquidity, tangibility, tax effects, and business risk. Theseempirical proxies are also presented in Table I, alongside their theoretical predictions.
Firms with negative book equity are excluded from the sample, resulting in a totalof 452 observations (323 observations from the LA-7 and 129 from the US). Descriptivestatistics for exogenous variables are presented in two forms: in Table V variables aregrouped by country while in Table VI the data are presented by variable to easecomparison[6]. Again, figures for US companies are usually bigger than for the typicalLatin American firm. North American firms are more leveraged, riskier, bigger, moreprofitable, and they have shorter-lived assets, more growth opportunities, and paymore taxes. Latin American firms have more tangible assets and pay relatively moredividends. Firms are roughly comparable in terms of asset maturity and liquidity.Volatility of some variables is very high, especially for business risk, asset maturity,liquidity, and, for some countries, tax rate. The fact that Mexican firms present asubstantially negative mean effective tax rate (2408 percent in comparison to the medianof only 24 percent) suggest the presence of large outliers that may inflate the standarddeviation for this variable. In order to account for such cases, in this variable and others,in the data analyses that follows I take appropriate remedial measures.
Table VII presents the correlation matrix of independent variables. Correlations aregenerally low, ranging from 20.1822 (liquidity versus tangibility) to 0.2741 (growthopportunities versus profitability) for the LA-7, and from20.3882 (size versus liquidity)to 0.333 (size versus dividend yield) for the US.
The quality of measurement of these variables, to what extent the data reported isaccurate, is certainly an issue. Annual accounting reports are usually subject to independentauditing and, since all firms present in the sample are public, accounting reports are subjectto supervision of each countrys securities commission. The degree of compliance maynevertheless differ from one country to another depending on how stringent are eachcommissions standards and how much resolve and enforcement power the commissionhas. Similarly, stock market data are also dependent on each markets depth. Anotherpossible source of measurement imprecision is the set of accounting standards adoptedin each country. These issues shall be taken into account when analyzing the results.
Besides the above variables, I employ a set of dummy variables as instruments. First,the sector of activity of each firm is included, given the possible systematic effects thatthe nature of the firms activities may have over its leverage, in particular the totalleverage measures. The sector of activity is represented by a set of dummy variablesbased on the classification informed in the database. Food and Beverages is chosen asthe base-case so that the instrument set may include an intercept. Likewise, countrydummies are used to account for any country-specific variation such as the institutionalframework, business environment, and macroeconomic conditions. Brazil is thenchosen as the base-case. Finally, year dummies are employed in order to account forcommon time shocks to all firms. The year 2002 is chosen as the base-case.
One final remark is that, in determining debt-maturity structure, the nature of theownership of the firm may induce systematic effects. State-owned firms, for instance, mayhave a lower bankruptcy probability due to implicit government guarantees a factorthat according to theory is decisive for the optimal maturity. Similarly, firms that belong
EBR23,1
54
Dow
nloa
ded
by E
scue
la d
e A
dmin
istra
cion
de
Neg
ocio
s par
a G
radu
ados
ESA
N A
t 08:
01 2
4 Fe
brua
ry 2
015
(PT)
-
Ob
s.M
ean
Med
ian
SD
Ob
s.M
ean
Med
ian
SD
Ob
s.M
ean
Med
ian
SD
Var
iab
les
Argentina
Brazil
Chile
Lev
erag
e62
10.
7965
0.19
907.
1732
4,10
41.
2211
0.16
2013
.681
91,
773
0.28
320.
1538
0.98
45A
sset
mat
uri
ty57
824
.164
617
.665
522
.086
73,
059
42.6
419
10.3
246
1,43
0.89
5175
633
.075
315
.742
282
.660
3S
ize
588
11.3
737
11.5
623
1.83
393,
635
11.4
402
11.5
013
1.84
261,
599
10.2
118
10.3
693
1.92
52G
row
thop
por
tun
itie
s50
00.
9920
0.91
770.
4378
3,45
10.
9489
0.73
931.
8685
1,33
02.
2458
1.29
058.
7085
Pro
fita
bil
ity
621
0.03
510.
0268
0.07
154,
093
0.03
160.
0210
0.12
011,
779
0.05
390.
0473
0.11
54B
usi
nes
sri
sk60
10.
7490
5.15
5015
4.79
784,
082
0.78
784.
3584
156.
6401
1,65
412
.624
85.
2892
124.
6428
Div
iden
dy
ield
597
0.02
720.
0000
0.06
233,
515
0.05
510.
0175
0.11
491,
361
0.04
780.
0314
0.10
43L
iqu
idit
y62
11.
6889
1.14
832.
7276
4,09
32.
3209
1.13
3320
.211
61,
769
5.01
181.
4252
42.8
555
Tan
gib
ilit
y60
40.
4581
0.45
080.
2622
4,09
90.
3537
0.32
850.
2565
1,74
30.
4138
0.40
360.
2883
Tax
rate
351
0.13
700.
0105
1.03
224,
090
0.34
270.
0267
11.1
032
1,50
10.
0291
0.04
780.
8786
Colom
bia
Mexico
Peru
Lev
erag
e28
30.
4484
0.10
851.
7798
1,39
60.
6342
0.34
781.
2782
1,03
20.
9174
0.16
0119
.452
0A
sset
mat
uri
ty14
320
.357
16.
2019
99.2
809
1,28
219
.999
916
.580
017
.287
259
72
0.41
7411
.649
739
8.94
84S
ize
300
11.0
373
11.0
497
1.56
891,
404
12.2
737
12.3
799
1.80
251,
023
10.1
975
10.1
482
1.28
13G
row
thop
por
tun
itie
s20
20.
8311
0.74
940.
4341
879
1.27
871.
0921
0.65
5863
51.
1095
0.89
920.
7238
Pro
fita
bil
ity
290
0.02
070.
0319
0.13
001,
411
0.07
280.
0739
0.08
031,
030
0.05
650.
0514
0.12
08B
usi
nes
sri
sk28
930
.838
97.
0272
721.
9443
1,41
116
.369
47.
8969
191.
8072
1,02
514
.923
76.
8122
264.
8394
Div
iden
dy
ield
211
0.03
490.
0045
0.08
5988
10.
0190
0.00
000.
0839
655
0.02
670.
0000
0.04
79L
iqu
idit
y28
41.
6725
1.32
091.
1693
1,41
25.
0568
1.47
8196
.903
11,
032
1.97
591.
3498
4.19
93T
ang
ibil
ity
277
0.24
780.
2137
0.18
911,
412
0.51
290.
5497
0.26
911,
032
0.47
830.
4731
0.22
25T
axra
te29
00.
1099
0.14
871.
5031
1,41
02
4.02
790.
2361
134.
0558
1,02
90.
3190
0.26
793.
9418
Venezuela
LatinAmerica
USA
Lev
erag
e17
50.
2757
0.15
820.
3367
9,38
40.
8542
0.17
9611
.291
45,
028
1.42
950.
7093
20.0
382
Ass
etm
atu
rity
5620
.032
313
.612
616
.768
66,
471
30.7
274
12.4
473
991.
8092
3,87
230
.503
67.
8454
270.
1606
Siz
e16
610
.919
610
.941
61.
8189
8,71
511
.175
011
.184
51.
9224
5,01
514
.224
814
.315
41.
7925
Gro
wth
opp
ortu
nit
ies
140
0.74
630.
6920
0.35
077,
137
1.24
120.
8770
4.02
214,
879
2.77
001.
8240
3.63
15P
rofi
tab
ilit
y17
50.
0346
0.01
330.
0659
9,39
90.
0447
0.03
830.
1120
5,02
70.
0645
0.07
440.
1813
Bu
sin
ess
risk
171
2.19
024.
6788
51.4
443
9,23
37.
8228
5.61
2721
2.02
665,
027
46.4
644
7.07
042,
837.
0812
Div
iden
dy
ield
207
0.04
530.
0124
0.12
037,
427
0.04
390.
0142
0.10
195,
956
0.01
060.
0000
0.02
08L
iqu
idit
y17
52.
1203
1.46
992.
8290
9,38
63.
1365
1.28
4844
.049
85,
017
2.45
901.
5497
6.68
18T
ang
ibil
ity
175
0.53
550.
5708
0.22
639,
342
0.40
970.
4008
0.26
825,
028
0.31
540.
2495
0.23
03T
axra
te17
40.
0971
0.07
031.
5772
8,84
52
0.43
060.
0811
54.0
800
5,02
70.
3316
0.36
072.
9790
Notes:
Th
eta
ble
pre
sen
tsth
ed
escr
ipti
ve
stat
isti
csfo
rea
chex
pla
nat
ory
var
iab
leb
yco
un
try
;th
ed
ata
cov
erth
ep
erio
d19
87-2
002;
LA
-7
refe
rsto
the
poo
lin
gto
get
her
ofal
lfi
rm-l
evel
dat
afo
rA
rgen
tin
a,B
razi
l,C
hil
e,C
olom
bia
,M
exic
o,P
eru
,an
dV
enez
uel
a
Table V.Descriptive statistics
Corporate debtmaturity
55
Dow
nloa
ded
by E
scue
la d
e A
dmin
istra
cion
de
Neg
ocio
s par
a G
radu
ados
ESA
N A
t 08:
01 2
4 Fe
brua
ry 2
015
(PT)
-
Cou
ntr
ies
Ob
s.M
ean
Med
ian
SD
Ob
s.M
ean
Med
ian
SD
Ob
s.M
ean
Med
ian
SD
Ob
s.M
ean
Med
ian
SD
Leverage
Businessrisk
Assetmaturity
Dividendyield
Arg
enti
na
621
0.79
650.
1990
7.17
3260
10.
7490
5.15
5015
4.79
857
824
.164
617
.665
522
.086
759
70.
0272
0.00
000.
0623
Bra
zil
4,10
41.
2211
0.16
2013
.681
94,
082
0.78
784.
3584
156.
640
3,05
942
.641
910
.324
61,
430.
903,
515
0.05
510.
0175
0.11
49C
hil
e1,
773
0.28
320.
1538
0.98
451,
654
12.6
248
5.28
9212
4.64
375
633
.075
315
.742
282
.660
31,
361
0.04
780.
0314
0.10
43C
olom
bia
283
0.44
840.
1085
1.77
9828
930
.838
97.
0272
721.
944
143
20.3
571
6.20
1999
.280
921
10.
0349
0.00
450.
0859
Mex
ico
1,39
60.
6342
0.34
781.
2782
1,41
116
.369
47.
8969
191.
807
1,28
219
.999
916
.580
017
.287
288
10.
0190
0.00
000.
0839
Per
u1,
032
0.91
740.
1601
19.4
520
1,02
514
.923
76.
8122
264.
839
5972
0.41
7411
.649
739
8.94
865
50.
0267
0.00
000.
0479
Ven
ezu
ela
175
0.27
570.
1582
0.33
6717
12.
1902
4.67
8851
.444
5620
.032
313
.612
616
.768
620
70.
0453
0.01
240.
1203
LA
-79,
384
0.85
420.
1796
11.2
914
9,23
37.
8228
5.61
2721
2.02
76,
471
30.7
274
12.4
473
991.
809
7,42
70.
0439
0.01
420.
1019
US
A5,
028
1.42
950.
7093
20.0
382
5,02
746
.464
47.
0704
2,83
7.08
3,87
230
.503
67.
8454
270.
161
5,95
60.
0106
0.00
000.
0208
Size
Liquidity
Growth
opportunities
Tangibility
Arg
enti
na
588
11.3
737
11.5
623
1.83
3962
11.
6889
1.14
832.
7276
500
0.99
200.
9177
0.43
7860
40.
4581
0.45
080.
2622
Bra
zil
3,63
511
.440
211
.501
31.
8426
4,09
32.
3209
1.13
3320
.211
63,
451
0.94
890.
7393
1.86
854,
099
0.35
370.
3285
0.25
65C
hil
e1,
599
10.2
118
10.3
693
1.92
521,
769
5.01
181.
4252
42.8
555
1,33
02.
2458
1.29
058.
7085
1,74
30.
4138
0.40
360.
2883
Col
omb
ia30
011
.037
311
.049
71.
5689
284
1.67
251.
3209
1.16
9320
20.
8311
0.74
940.
4341
277
0.24
780.
2137
0.18
91M
exic
o1,
404
12.2
737
12.3
799
1.80
251,
412
5.05
681.
4781
96.9
031
879
1.27
871.
0921
0.65
581,
412
0.51
290.
5497
0.26
91P
eru
1,02
310
.197
510
.148
21.
2813
1,03
21.
9759
1.34
984.
1993
635
1.10
950.
8992
0.72
381,
032
0.47
830.
4731
0.22
25V
enez
uel
a16
610
.919
610
.941
61.
8189
175
2.12
031.
4699
2.82
9014
00.
7463
0.69
200.
3507
175
0.53
550.
5708
0.22
63L
A-7
8,71
511
.175
011
.184
51.
9224
9,38
63.
1365
1.28
4844
.049
87,
137
1.24
120.
8770
4.02
219,
342
0.40
970.
4008
0.26
82U
SA
5,01
514
.224
814
.315
41.
7925
5,01
72.
4590
1.54
976.
6818
4,87
92.
7700
1.82
403.
6315
5,02
80.
3154
0.24
950.
2303
Profitability
Taxrate
Arg
enti
na
621
0.03
510.
0268
0.07
1535
10.
1370
0.01
051.
0322
Bra
zil
4,09
30.
0316
0.02
100.
1201
4,09
00.
3427
0.02
6711
.103
2C
hil
e1,
779
0.05
390.
0473
0.11
541,
501
0.02
910.
0478
0.87
86C
olom
bia
290
0.02
070.
0319
0.13
0029
00.
1099
0.14
871.
5031
Mex
ico
1,41
10.
0728
0.07
390.
0803
1,41
02
4.02
790.
2361
134.
056
Per
u1,
030
0.05
650.
0514
0.12
081,
029
0.31
900.
2679
3.94
18V
enez
uel
a17
50.
0346
0.01
330.
0659
174
0.09
710.
0703
1.57
72L
A-7
9,39
90.
0447
0.03
830.
1120
8,84
52
0.43
060.
0811
54.0
800
US
A5,
027
0.06
450.
0744
0.18
135,
027
0.33
160.
3607
2.97
90
Notes:
Th
eta
ble
pre
sen
tsth
ed
escr
ipti
ve
stat
isti
csfo
rea
chco
un
try
by
exp
lan
ator
yv
aria
ble
inth
ep
erio
d19
87-2
002;
LA
-7
refe
rsto
the
poo
lin
gto
get
her
ofal
lfi
rm-l
evel
dat
afo
rA
rgen
tin
a,B
razi
l,C
hil
e,C
olom
bia
,M
exic
o,P
eru
,an
dV
enez
uel
a
Table VI.Descriptive statistics
EBR23,1
56
Dow
nloa
ded
by E
scue
la d
e A
dmin
istra
cion
de
Neg
ocio
s par
a G
radu
ados
ESA
N A
t 08:
01 2
4 Fe
brua
ry 2
015
(PT)
-
Lev
erag
eA
sset
mat
uri
tyS
ize
Gro
wth
opp
ortu
nit
ies
Pro
fita
bil
ity
Bu
sin
ess
risk
Div
iden
dy
ield
Liq
uid
ity
Tan
gib
ilit
y
PanelA
LatinAmerica
LA-7
Ass
etm
atu
rity
20.
0002
1.00
00S
ize
20.
0126
0.01
101.
0000
Gro
wth
opp
ortu
nit
ies
0.03
810.
0548
0.24
511.
0000
Pro
fita
bil
ity
20.
0120
0.00
230.
1881
0.27
411.
0000
Bu
sin
ess
risk
0.00
082
0.00
050.
0109
20.
0068
0.00
381.
0000
Div
iden
dy
ield
20.
0417
20.
0044
20.
0226
20.
1324
0.04
960.
0120
1.00
00L
iqu
idit
y2
0.02
832
0.00
322
0.11
990.
0622
0.04
310.
0010
20.
0019
1.00
00T
ang
ibil
ity
0.00
882
0.01
280.
2410
0.02
220.
0945
0.00
042
0.05
562
0.18
221.
0000
Tax
rate
0.00
210.
0001
20.
0272
20.
0112
0.00
712
0.00
320.
0097
20.
0023
20.
0085
PanelB
USA
USA
Ass
etm
atu
rity
0.00
471.
0000
Siz
e0.
0745
20.
0558
1.00
00G
row
thop
por
tun
itie
s2
0.07
192
0.01
252
0.20
931.
0000
Pro
fita
bil
ity
20.
0180
20.
0317
0.15
480.
1256
1.00
00B
usi
nes
sri
sk2
0.00
372
0.00
070.
0009
20.
0062
20.
0076
1.00
00D
ivid
end
yie
ld0.
0853
20.
0243
0.33
302
0.16
800.
0521
20.
0117
1.00
00L
iqu
idit
y2
0.05
280.
2262
20.
3882
0.17
152
0.08
652
0.01
242
0.15
931.
0000
Tan
gib
ilit
y0.
0867
0.01
840.
1618
20.
1612
0.07
870.
0324
0.17
732
0.23
291.
0000
Tax
rate
20.
0169
0.00
010.
0131
0.00
340.
0169
0.00
262
0.01
852
0.00
412
0.01
27
Notes:
Pan
elA
pre
sen
tsth
eco
rrel
atio
nm
atri
xfo
rfi
rms
inL
atin
Am
eric
aw
hil
eP
anel
Bp
rese
nts
the
corr
elat
ion
mat
rix
for
firm
sin
the
US
A;
LA
-7
refe
rsto
the
poo
lin
gto
get
her
ofal
lfi
rm-l
evel
dat
afo
rA
rgen
tin
a,B
razi
l,C
hil
e,C
olom
bia
,Mex
ico,
Per
u,a
nd
Ven
ezu
ela
wh
ile
US
A
refe
rsto
the
poo
lin
gof
firm
-lev
eld
ata
for
the
US
A;
the
dat
aco
ver
the
per
iod
1987
-200
2
Table VII.Correlation matrices
Corporate debtmaturity
57
Dow
nloa
ded
by E
scue
la d
e A
dmin
istra
cion
de
Neg
ocio
s par
a G
radu
ados
ESA
N A
t 08:
01 2
4 Fe
brua
ry 2
015
(PT)
-
to an industrial conglomerate or that are subsidiaries of powerful multinationalcorporations may face less credit constraints than independent local firms. Also, given thewide privatization process and mergers and acquisitions tide that took place inLatin America in the early 1990s, it would be important to precisely determine when thechange of ownership status occurred for each firm. Despite the relevance of such aspect,the database does not provide reliable detailed information about the ownership of thefirms for most of the countries and periods studied. Therefore, I opt for leaving theownership variable out of the study[7].
3.2 Panel data analysisPanel data analysis presents several advantages for the treatment of economic problemswhere cross-sectional variation and dynamic effects are relevant. Hsiao (1986) raisesthree advantages possessed by panel datasets: since they provide a larger number ofdata points, they allow increase in the degrees of freedom and reduce the collinearityamong explanatory variables; they allow the investigation of problems that cannot besolely addressed by either cross-section or time series datasets; and they provide ameans of reducing the missing variable problem. Baltagi (1995) adds to these the usuallyhigher accuracy of micro-unit data respective to aggregate data and the possibility ofexploring the dynamics of adjustment of a particular phenomenon over time.
Estimation of panel data models can be done by ordinary least squares (OLS) in the caseof simple pooling and fixed-effects formulations and by generalized least squares (GLS) forthe random-effects formulation (Hall and Cummins, 1997). However, in the presence ofdynamic effects (lagged dependent variable amongst explanatory variables) OLSestimators are biased and inconsistent, and the same occurs with the GLS estimator(Baltagi, 1995). In order to overcome such problem, Anderson and Hsiao (1981) suggest afirst difference transformation to the model so that all variables constant through time foreach cross-section unit are wiped out, including the fixed effects intercept. The authorsestimate the transformed model with an instrumental variable approach. Advancing uponsuch approach, Arellano and Bond (1991) suggest a two-step estimation procedure usingGLS in the first step and then obtaining the optimal generalized method of moments(GMM) estimator in the second step (Hansen, 1982). Such estimation is convenient becauseGMM does not require any particular distribution form, solving therefore problems ofheteroskedasticity,normality,simultaneity, andmeasurementerrors (Antoniou etal., 2006).
The main advantage of such method for the investigation of the problem proposed in thispaper is that observations of firms from different countries can be pooled together in orderto increase the degrees of freedom. Pooling together firms, on the other hand, assumes thatparameters (slopes and intercepts) are constant across firms. This is, of course, a very strongassumption and subject to potential biases (Hsiao, 1986). That would be the case if theeffects of a given independent variable are different for different kinds of firms, for instancesmall and large firms. The careful choice of firm-specific variables (such as firm size) helpscontrol for these possible biases. Nevertheless, this remains a limitation of this research.
3.3 Empirical modelThe first step is to define the following general (static) model:
MRit b0i b0t XK
k1b1kY ikt
XL
l1b2lZ ilt y i 1it 1
EBR23,1
58
Dow
nloa
ded
by E
scue
la d
e A
dmin
istra
cion
de
Neg
ocio
s par
a G
radu
ados
ESA
N A
t 08:
01 2
4 Fe
brua
ry 2
015
(PT)
-
WhereMRit is the stacked vector of the dependent variable (the ith-firm maturity ratio onthe tth-period), Yikt is the matrix of K firm-specific independent variables (includingindustry dummies in the simple pooling and random-effects models), Zilt is the matrix ofL country dummies (in the simple pooling and random-effects models for the LA-7),b0i isthe firm-specific intercept in the fixed-effects model, b0t is the period-specific intercept,b1k and b2l are the matrices of coefficients, ni is the firm-specific error term in therandom-effects model, and 1it is a vector of error terms.
The next step is to test the model above for fixed- and random-effects[8]. Once it isestablished that the fixed-effects model provides a good fit for the model, then the laggeddependent variable is added to equation (1), which is then first-differenced yielding thedynamic model below:
DMRit b00iDMRit21 XK
k1b1kDYikt 1it 2
One advantage of this specification is that the rate of adjustment of the firm towards itsoptimal maturity[9] can be estimated as l 12 b00i. If adjustment costs are high, therate of adjustment is expected to be small (l approaching zero), while a very high rate ofadjustment (l approaching one) suggests the presence of negligible adjustment costs.
4. Empirical results4.1 Preliminary specification testsIn order to determine which panel data model (simple pooling, fixed-effects, orrandom-effects) better suits the data, I perform two specification tests: the F-test ofsimple pooling versus fixed-effects model and the Hausman test of random versus fixedeffects. The results are shown in Table VIII.
Panel A: F-test Panel B: Hausman testRegion MR1 MR2 MR1 MR2
LA-7 F(664,3346) F(741,3855) x 2(16) x 2(11)4.7601 * * 7.6756 * * 40.4090 * * 31.6360 * *
p-value 0.000 0.000 0.001 0.001USA F(595,2683) F(624,3024) x 2(10) x 2(10)
6.9207 * * 14.1590 * * 33.3920 * * 31.6810 * *
p-value 0.000 0.000 0.000 0.001
Notes: Significance at: *5 and * *1 percent levels; Panel A presents the F-test of a simple pooled OLSagainst a fixed-effects specification; this test statistic is for testing the null hypothesis that firmsintercepts in the basic fixed-effects panel data model are all equal, against the alternative hypothesisthat each firm has its own (distinct) intercept; the test assumes identical slopes for all independentvariables across all firms, and it is distributed F(df1,df2); Panel B presents the Hausman specificationtest of random-effects against fixed-effects specification; this test statistic is for testing the nullhypothesis of the random-effects specification against the alternative hypothesis of the fixed-effectsspecification in the basic panel data model, and it is distributed x 2(df). LA-7 refers to the poolingtogether of all firm-level data for Argentina, Brazil, Chile, Colombia, Mexico, Peru, and Venezuelawhile USA refers to the pooling of firm-level data for the USA; the data cover the period 1987-2002;dependent variables: MR1 long-term debt/total debt; MR2 long-term book liabilities/total bookliabilities
Table VIII.Specification tests
Corporate debtmaturity
59
Dow
nloa
ded
by E
scue
la d
e A
dmin
istra
cion
de
Neg
ocio
s par
a G
radu
ados
ESA
N A
t 08:
01 2
4 Fe
brua
ry 2
015
(PT)
-
The first step is to determine whether the panel data specification that simply poolstogether all available data for all firms and time periods is adequate to describe the data.As pointed out by Hsiao (1986), simple least squares estimation of pooled cross-sectionand time series data may be seriously biased[10]. The model tested in equation (1)includes firm-specific variables described above, as well as country-specific dummyvariables for the LA-7. The results in Table VIII strongly reject the single intercepthypothesis, both for the LA-7 and for the US.
The next step is to determine which model of variable intercepts across firms betterfits the data. Table VIII also presents the results for a Hausman specification test ofrandom- versus fixed-effects. The test, as suggested by Hsiao (1986, p. 49), is particularlyappropriate in situations whereN (the number of cross-sectional units) is large relative toT (the number of time periods) precisely the case of this study. Again, the model inequation (1) above is employed. The test strongly rejects the random-effectsspecification for both groups of countries.
Given these results, I conclude that the fixed-effects specification is an adequate fitto the data. Therefore, after first differencing equation (1), firm-specific interceptsdisappear and the dynamic model of equation (2) is used in the estimation that follows.
4.2 Dynamic panel data estimation resultsPreliminary runs of the fixed-effects model of equation (1) revealed a substantial presenceof autocorrelation in the residuals. This finding raises the question that the maturitychoice of the firm may be dynamic, i.e. current maturity may depend on past maturity.Antoniou et al. (2006) explicitly model such possibility, and suggest that a dynamic ratherthan static panel data analysis may be more adequate. However, as mentioned above,usual OLS and GLS estimators are biased and inconsistent when the lagged dependentvariable is included in the right-hand side of the panel data model. In order to overcomethis problem, GMM estimation is used instead.
Equation (2) is then estimated by GMM using as instruments first-order laggedvalues of the levels[11] of explanatory variables, sector dummies, country dummies(for Latin America), year dummies, and a constant[12]. In order to control for outliers,I exclude influential observations based on Cooks distance indicator. As a result,26 observations are excluded for MR1 and 28 for MR2 in Latin America (respectively40 and 45 in the US). The number of excluded observations is minimal given the samplessize (less than 2 percent). Nevertheless, I report results with and without outliers inTables IX and X, respectively. Standard errors are heteroskedasticity robust accordingto the method proposed by White (1980)[13] and are also robust to autocorrelation.
One important issue when estimating via GMM is to make sure that the instrumentset is adequate. Tables IX and X report the Sargans test statistic for the null hypothesisthat moment restrictions hold. Results cannot reject the restrictions at usual significancelevels in all cases, with the exception of MR2 for the US when outliers are excluded fromthe sample. Therefore, I conclude that the instrument set is valid in general[14]. Resultsalso indicate that the model provides a reasonable fit for the data. Adjusted R 2 rangeclose to 0.5, being very similar to both samples.
One robust result is that the lagged dependent variable is positive and highlysignificant for both samples and both measures of maturity. The estimated rate ofadjustment to an optimal maturity structure ranges between 0.46 and 0.68, an indicationthat firms in the sample face moderate adjustment costs. Adjustment costs for the measure
EBR23,1
60
Dow
nloa
ded
by E
scue
la d
e A
dmin
istra
cion
de
Neg
ocio
s par
a G
radu
ados
ESA
N A
t 08:
01 2
4 Fe
brua
ry 2
015
(PT)
-
Reg
ion
Lat
inA
mer
ica
US
AD
epen
den
tv
aria
ble
sIn
dep
end
ent
var
iab
les
MR
1t-
stat
isti
csM
R2
t-st
atis
tics
MR
1t-
stat
isti
csM
R2
t-st
atis
tics
Mat
uri
tyt
10.
3134
**
7.06
20.
5056
**
11.5
350.
3897
**
5.09
60.
5412
**
7.46
1R
esid
ual
lev
erag
e t0.
0021
0.48
12
0.00
202
1.28
40.
0033
0.91
90.
0079
1.64
8A
sset
mat
uri
tyt
0.00
01*
*2.
587
20.
0000
20.
704
20.
0000
21.
696
20.
0000
20.
848
Siz
e t0.
0364
1.25
12
0.01
092
0.59
92
0.03
642
1.30
02
0.00
882
0.50
4G
row
thop
por
tun
itie
s t0.
0453
1.79
10.
0016
0.12
50.
0011
0.21
42
0.00
39*
22.
321
Pro
fita
bil
ityt
0.21
911.
934
20.
0579
20.
936
20.
0402
20.
663
20.
1074
21.
559
Bu
sin
ess
risk
t0.
0000
1.63
70.
0000
*2.
559
20.
0000
21.
513
20.
0000
**
27.
037
Div
iden
dy
ield
t2
0.11
262
1.37
22
0.04
582
0.92
20.
9360
0.95
30.
1909
0.28
2L
iqu
idit
yt
0.04
17*
*3.
004
0.01
481.
640
0.00
570.
691
0.01
351.
485
Tan
gib
ilit
yt
0.16
050.
881
20.
0788
20.
757
20.
3243
21.
063
20.
2286
21.
204
Tax
effe
cts t
20.
0001
**
23.
269
20.
0000
**
23.
989
0.00
260.
533
20.
0004
20.
201
Nu
mb
erof
obse
rvat
ion
s2,
814
3,36
02,
136
2,47
9A
dju
sted
R2
0.49
800.
5005
0.49
790.
5016
F-s
tati
stic
246.
2480
**
292.
5154
**
194.
6664
**
224.
2141
**
F(d
f 1;
df 2
)(1
0;2,
803)
(10;
3,34
9)(1
0;2,
125)
(10;
2,46
8)Sargansteststatistic
p-v
alu
e0.
605
0.38
50.
994
0.29
2x
235
.093
539
.924
410
.143
227
.277
5d
f38
3824
24D
urb
in-W
atso
nst
atis
tic
2.29
382.
4133
2.17
572.
2020
Notes:
Sig
nifi
can
ceat
:* 5
and
** 1
per
cen
tle
vel
s;fi
rst-
dif
fere
nce
sm
odel
soth
atid
iosy
ncr
atic
firm
-eff
ects
con
stan
tth
rou
gh
tim
ear
eel
imin
ated
;th
em
odel
ises
tim
ated
by
GM
Mu
sin
gas
inst
rum
ents
firs
t-or
der
lag
ged
val
ues
ofth
ele
vel
sof
exp
lan
ator
yv
aria
ble
s,se
ctor
du
mm
ies,
cou
ntr
yd
um
mie
s(f
orL
atin
Am
eric
a),y
ear
du
mm
ies,
and
aco
nst
ant;
esti
mat
ion
inth
ep
erio
d19
87-2
002;
Lat
inA
mer
ica
refe
rsto
the
poo
lin
gto
get
her
ofal
lfirm
-lev
eld
ata
for
Arg
enti
na,
Bra
zil,
Ch
ile,
Col
omb
ia,
Mex
ico,
Per
u,
and
Ven
ezu
ela;
dep
end
ent
var
iab
les:
MR
1
lon
g-t
erm
deb
t/to
tal
deb
t;M
R2
lon
g-t
erm
boo
kli
abil
itie
s/to
tal
boo
kli
abil
itie
s;re
por
tedt-
stat
isti
csar
eca
lcu
late
du
sin
gh
eter
osk
edas
tici
ty-r
obu
stst
and
ard
erro
rs(W
hit
e)an
dar
eal
soro
bu
stto
auto
corr
elat
ion
(Bar
tlet
tK
ern
el);
Mod
el:DMRitb
00iDMRit2
1P
K k1b
1kDY
ikt1it
Table IX.Panel data analysis of
maturity ratios for LatinAmerica and the USA
Corporate debtmaturity
61
Dow
nloa
ded
by E
scue
la d
e A
dmin
istra
cion
de
Neg
ocio
s par
a G
radu
ados
ESA
N A
t 08:
01 2
4 Fe
brua
ry 2
015
(PT)
-
Reg
ion
Lat
inA
mer
ica
US
AD
epen
den
tv
aria
ble
sIn
dep
end
ent
var
iab
les
MR
1t-
stat
isti
csM
R2
t-st
atis
tics
MR
1t-
stat
isti
csM
R2
t-st
atis
tics
Mat
uri
tyt
10.
3175
**
7.11
50.
4906
**
11.7
970.
3416
**
4.89
70.
3224
**
6.04
2R
esid
ual
lev
erag
e t0.
0069
1.24
60.
0033
0.94
00.
0241
1.73
50.
0361
**
4.45
2A
sset
mat
uri
tyt
0.00
01*
2.55
62
0.00
002
0.90
92
0.00
002
1.47
70.
0000
0.42
5S
ize t
0.03
681.
248
20.
0113
20.
617
20.
0089
20.
262
0.03
081.
761
Gro
wth
opp
ortu
nit
ies t
0.04
431.
754
0.00
160.
127
0.00
450.
977
20.
0030
*2
2.23
6P
rofi
tab
ilit
yt
0.21
581.
879
20.
1040
21.
673
20.
0643
21.
094
20.
1077
21.
710
Bu
sin
ess
risk
t0.
0000
0.99
60.
0000
*2.
575
20.
0000
21.
126
20.
0000
**
25.
056
Div
iden
dy
ield
t2
0.15
742
1.86
32
0.05
352
1.03
41.
1893
1.31
60.
2778
0.40
0L
iqu
idit
yt
0.04
21*
*3.
006
0.01
441.
621
0.05
93*
*2.
839
0.04
90*
*4.
224
Tan
gib
ilit
yt
0.11
980.
661
20.
0989
20.
969
20.
1504
20.
455
20.
0711
20.
424
Tax
effe
cts t
20.
0001
**
23.
238
20.
0000
**
23.
962
0.00
370.
786
20.
0017
21.
053
Nu
mb
erof
obse
rvat
ion
s2,
788
3,33
22,
096
2,43
4A
dju
sted
R2
0.49
820.
5008
0.49
810.
5041
F-s
tati
stic
s27
7.74
67*
*33
5.12
02*
*20
8.94
07*
*24
8.29
50*
*
F(d
f 1;
df 2
)(1
0;2,
777)
(10;
3,32
1)(1
0;2,
085)
(10;
2,42
3)Sargansteststatistic
p-v
alu
e0.
653
0.25
50.
992
0.00
0x
234
.044
443
.304
910
.488
059
.351
6*
*
df
3838
2424
Du
rbin
-Wat
son
Sta
tist
ic2.
2973
2.41
522.
0687
1.98
76
Notes:
Sig
nifi
can
ceat
:* 5
and
** 1
per
cen
tle
vel
s;.fi
rst-
dif
fere
nce
sm
odel
soth
atid
iosy
ncr
atic
firm
-eff
ects
con
stan
tth
rou
gh
tim
ear
eel
imin
ated
;th
em
odel
ises
tim
ated
by
GM
Mu
sin
gas
inst
rum
ents
firs
t-or
der
lag
ged
val
ues
ofth
ele
vel
sof
exp
lan
ator
yv
aria
ble
s,se
ctor
du
mm
ies,
cou
ntr
yd
um
mie
s(f
orL
atin
Am
eric
a),y
ear
du
mm
ies,
and
aco
nst
ant;
infl
uen
tial
obse
rvat
ion
sh
ave
bee
nre
mov
edb
ased
onC
ook
sD
.est
imat
ion
inth
ep
erio
d19
87-2
002;
Lat
inA
mer
ica
refe
rsto
the
poo
lin
gto
get
her
ofal
lfi
rm-l
evel
dat
afo
rA
rgen
tin
a,B
razi
l,C
hil
e,C
olom
bia
,M
exic
o,P
eru
,an
dV
enez
uel
a;d
epen
den
tv
aria
ble
s:M
R1
lon
g-t
erm
deb
t/to
tal
deb
t;M
R2
lon
g-t
erm
boo
kli
abil
itie
s/to
tal
boo
kli
abil
itie
s;re
por
tedt-
stat
isti
csar
eca
lcu
late
du
sin
gh
eter
osk
edas
tici
ty-
rob
ust
stan
dar
der
rors
(Wh
ite)
and
are
also
rob
ust
toau
toco
rrel
atio
n(B
artl
ett
Ker
nel
);M
odel
:DMRitb
00iDMRit2
1P
K k1b
1kDY
ikt1it
Table X.Panel data analysis ofmaturity ratios for LatinAmerica and the USA,excluding outliers
EBR23,1
62
Dow
nloa
ded
by E
scue
la d
e A
dmin
istra
cion
de
Neg
ocio
s par
a G
radu
ados
ESA
N A
t 08:
01 2
4 Fe
brua
ry 2
015
(PT)
-
including trade finance are in general higher than pure financial sources, and this is acommon pattern between the LA-7 and the US samples[15]. This is in line with thereasoning that trade finance maturity depends largely on the business practices of eachsector of activity, which makes it difficult for the firm to change. At the same time,adjustment to the target maturity is by no means costless and instantaneous.
Consistent results are also obtained for liquidity. It appears to have a positive effecton debt maturity, as postulated by theory (signaling hypothesis), once outliers have beenremoved from the sample.
As for the other determinants of debt maturity, no significant effect is detected forsize, profitability, dividend yield, and tangibility. Some weak evidence is found forresidual leverage (positive effect, but only in the US for MR2 without outliers), Assetmaturity (positive effect for MR1 in the LA-7), and growth opportunities (negative,but only in the US for MR2). I conclude from these results that the capital structuredecision does not have a singular effect on the debt-maturity decision, oncesimultaneity is resolved. Also, the significant and positive effect of asset maturitysupports the maturity-matching hypothesis for financial debt in Latin America. This isin line with most of the previous empirical evidence, which finds a positive andsignificant effect of asset maturity. One possibility is that many studies employedmeasures of debt maturity that include trade finance (as MR2), which is more sensitiveto the business practices of each sector, as argued above. The fact that it does not causeany effect in the US sample, nor does it to MR2 in the LA-7, suggests that financial debtin emerging markets is more sensitive to the life of a companys assets than in adeveloped market. A possible interpretation for this result is that debt financing maybe more rationed in emerging markets[16] Finally, the negative effect of growthopportunities on MR2 for the US supports the agency hypothesis. However, the factthat it is significant only for a measure including trade finance in a developed market isdifficult to interpret.
Disparities in signs between the samples are found for business risk. My resultsindicate that riskier firms in Latin America have longer debt maturity (which supportsthe agency hypothesis) while riskier firms in the US have shorter maturity (whichsupports the tradeoff hypothesis). Both results are found only for the measure includingtrade finance. This is the one empirical pattern that clearly differed between emergingand developed markets. At face value, it is awkward that riskier firms in volatile andfinancially constrained markets obtain longer term financing. It suggests that the role oftrade financing in such markets goes beyond the mere provision of funds, but has alsoimplications for the operating risk profile of firms. Nevertheless, a more in-depthinvestigation of this finding is called for.
Finally, tax effects seem to have a consistently significant negative effect over thedebt maturity of Latin American firms, but an insignific