0 FUNDAÇÃO GETULIO VARGAS ESCOLA DE ADMINISTRAÇÃO DE EMPRESAS DE SÃO PAULO MÁRCIA LIMA BANDEIRA EMPIRICAL EVIDENCE OF TRADE CREDIT USES OF BRAZILIAN PUBLICLY- LISTED COMPANIES SÃO PAULO 2008
1 MÁRCIA LIMA BANDEIRA EMPIRICAL EVIDENCE OF TRADE CREDIT USES OF BRAZILIAN PUBLICLY- LISTED COMPANIES Dssertação apresentada à Escola de Admnstração de Empresas de São Paulo da Fundação Getúlo Vargas, como requsto para a obtenção de título de Mestre em Admnstração de Empresas Campo de Conhecmento: Mercados Fnanceros e Fnanças Corporatvas Orentador: Prof. Dr. Rchard Sato São Paulo 2008
2 MÁRCIA LIMA BANDEIRA EMPIRICAL EVIDENCE OF TRADE CREDIT USES OF BRAZILIAN PUBLICLY- LISTED COMPANIES Dssertação apresentada à Escola de Admnstração de Empresas de São Paulo da Fundação Getúlo Vargas, como requsto para obtenção de título de Mestre em Admnstração de Empresas Campo de Conhecmento: Mercados Fnanceros e Fnanças Corporatvas Data da Aprovação / / Banca Examnadora: Prof. Dr. Rchard Sato (Orentador) FGV EAESP Prof. Dr. Rafael Felpe Schozer FGV EAESP Prof. Dr. Paulo Renato Soares Terra UFRGS
3 Bandera, Márca Lma. Emprcal Evdence of Trade Credt Uses of Brazlan Publcly-Lsted Companes / Márca Lma Bandera. - 2008. 49 f. Orentador: Rchard Sato. Dssertação (mestrado) - Escola de Admnstração de Empresas de São Paulo. 1. Socedades por ações - Fnancamento. 2. Empresas - Fnancamento. I. Sato, Rchard. II. Dssertação (mestrado) - Escola de Admnstração de Empresas de São Paulo. III. Título. CDU 347.725
4 Aos meus pas, pelo amor e dedcação.
5 AGRADECIMENTOS Agradeço ao Professor Doutor Rchard Sato pela orentação e ensnamentos transmtdos. Agradeço aos professores Joaqum Rubens Fontes Flho, Edmlson Alves e Hsa Hua Sheng pelos ensnamentos. Agradeço aos professores Rafael Schozer e Paulo Renato Terra por suas contrbuções. Agradeço à mnha famíla pelo amor, apoo e compreensão de sempre. Agradeço ao Wllam pelo afeto, apoo e companhersmo. Agradeço à mnha rmã Marana e ao professor Rogéro Sobrera pelo ncentvo ao ngresso no mestrado. Agradeço aos meus amgos que compartlharam desse período de aprendzado, e em especal à Adrana e à Sabrna que colaboraram na a realzação deste trabalho. Agradeço aos funconáros da FGV-SP. Agradeço ao CNPq pela ajuda fnancera concedda durante o curso.
6 RESUMO EXPANDIDO Trade credt é o prazo conceddo a um comprador pelo fornecedor, durante a realzação de uma venda. Portanto, trata-se de um crédto conceddo por nsttuções não-fnanceras. Esta dssertação testa as hpóteses de substtução, complementaredade e reputação relaconadas ao uso de trade credt por empresas brasleras de captal aberto, e, adconalmente, analsa a relação entre o endvdamento e o fornecmento de trade credt por estas empresas. Sob uma vsão mas tradconal, o Trade credt é tdo como uma fonte alternatva de crédto em relação ao fnancamento bancáro. Estudos mas recentes ressaltam sua mportânca em snalzar às nsttuções fnanceras a boa qualdade da empresa, uma vez que este carregara nformações relevantes possuídas pelos fornecedores sobre seus clentes. Neste sentdo, o trade credt tera também a função de facltar o acesso ao crédto. Alphonse, Ducret and Sévern (2006) sugerem que, num prmero momento, a empresa com restrção ao crédto usara o trade credt como complemento do crédto bancáro (hpótese de complementaredade). Consderando que este sera percebdo pelos bancos como um snal de boa qualdade da empresa (hpótese de reputação), enquanto num segundo momento, os bancos se tornaram dspostos a dar crédto para essa empresa, e o trade credt passara a ser uma fonte substtutva de fnancamento (hpótese de substtução). Os resultados confrmam as hpóteses de substtução, complementardade, e reputação. Além de ser uma mportante fonte de fnancamento, o uso do trade credt parece de fato snalzar a boa qualdade da empresa, mesmo para empresas de captal aberto, facltando assm seu acesso ao fnancamento bancáro. Com relação à concessão de trade credt, encontramos que o acesso ao crédto está negatvamente relaconado ao fornecmento de trade credt. Este resultado é contráro ao encontrado por pesqusas que utlzaram amostra de empresas pequenas e médas, e se justfca pela menor dependênca das empresas de captal aberto em relação a seus clentes, condção que as leva a fazer menos concessões
7 aos mesmos. Com relação às empresas com maor acesso ao mercado de captas (por exemplo, empresas que emtem debêntures), essa relação negatva entre o acesso ao crédto e fornecmento de trade credt fo anda mas sgnfcatva. Este resultado evdenca que quanto maor a reputação menor o fornecmento de trade credt por estas empresas. Palavras-chave: trade credt, empresas de captal aberto, fontes de fnancamento, endvdamento.
8 ABSTRACT Ths research provdes emprcal evdence on the use of trade credt as ether a substtuton or a complement to bank debt for lsted companes n Brazl, controllng for the frms reputaton, as stated by Alphonse, Ducret and Sévern (2006). The sample conssts of 263 publcly-lsted companes for 2006. Our fndngs support all three hypotheses. We provde evdence that trade credt may be used as a sgnal for the frm s qualty. Keywords: trade credt, publc-lsted companes, sources of fnancng, debt.
9 INDEX OF TABLES Table 1 - Testng Hypotheses and Related Varables..... 19 Table 2 - Control Varables and Related Hypotheses of Trade Credt Use.... 23 Table 3 - Control Varables and Related Hypotheses of Trade Credt Supply... 24 Table 4 - Dstrbuton of Sample by Industry........ 26 Table 5 - Descrptve Statstcs...... 27 Table 6 - Accounts Payable and Bank Debt - Unvarate Analyss...... 30 Table 7 - Accounts Recevable and Bank Debt Unvarate Analyss........ 31 Table 8 Correlaton Matrx of Independent Varables - Trade Credt Use....32 Table 9 Correlaton Matrx of Independent Varables - Trade Credt Offer...32 Table 10 - The Effect of Accounts Payable and Bank Debt........ 35 Table 11 - The Effect of Accounts Payable and Bank Debt Reputaton Hypothess. 37 Table 12 - The Effect of Bank Debt and Corporate Bonds on Accounts Payable... 39
10 INDEX 1. INTRODUCTION... 11 2. TRADE CREDIT USE AND SUPPLY... 13 3. LITERATURE REVIEW AND HYPOTHESES... 15 3.1 Trade Credt Theores...15 3.2 Hypotheses...17 3.3 Varables Descrpton...21 3.3.1. Analyss of trade credt use...21 3.3.2. Analyss of trade credt supply...24 4. DATA AND SUMMARY STATISTICS... 26 5. EMPIRICAL ANALYSIS... 29 5.1 Unvarate Analyss...29 5.2 Multvarate Analyss...34 6. CONCLUDING REMARKS... 41 REFERENCES... 42 Appendx A Sample... 45
11 1. INTRODUCTION Trade credt s credt extended by supplers to buyers, partcularly durng sales negotaton. Although ts usage dates back to the Mddle Age (Le Goff, 1957), trade credt has been nvestgated more at depth only snce recent years. Schwartz (1974) reports that trade credt s a complement to bank debt and captal market debt. It dmnshes the effcency of an aggregate monetary control, but on the other hand t mtgates dscrmnatory effects caused by a restrctve monetary polcy, whch tends to affect manly small and medum enterprses (SME). Because of ths credt constrant, trade credt s vewed as a fnancal source of greater mportance n countres where fnancal nsttutons and captal markets are not fully developed. It does not mean, however, that credt extenson (or trade credt) s not relevant n countres wth well-developed fnancal nsttutons such as the US or Germany. Most papers use small frm data to test and analyze trade credt theores. Indeed, trade credt would be a source of fnancng mostly mportant to frms that face more credt access constrant. Beck, et al. (2006) found that older, larger and foregnowned frms have easer access to fnancng, whle trade credt would target small and medum-szed frms. Despte the relevance of trade credt to Brazlan frms, there s no lterature about ths source of fnancng usng Brazlan data. Ths study ams at fllng ths gap. Another nnovaton ntroduced by the present study s the focus on Brazlan-lsted frms. It s reasonable to expect, therefore, that they have at least the same access to bank loans as prvate frms. In addton, publc lsted frms have access to publc debt whle prvate frms do not. Nevertheless, the complex procedures for accessng these other sources of fnancng, the proportonally larger fnancng needs of publc frms and ther need to dversfy ther fnancng suggest that trade credt may be just as (or almost as) useful for publc, large frms as t s for smaller ones. Ths research nvestgates trade credt from the vewpont of both suppler and buyer based on a sample of Brazlan publc frms data. We test the hypothess of Alphonse, Ducret and Sévern (2006) who argue that credt ratoned frms would use trade credt as complementary and alternatve source of fnancng, whle frms not
12 facng credt ratonng would use t only as an alternatve source. The analyss of trade credt supply (the suppler vewpont) focuses on frms access to credt. Our fndngs suggest that, for publc frms, trade credt supply s nversely related to access to loans and other sources of fnancng. The remanng of ths paper s organzed as follows. Secton 2 shows trade credt usage and supply n Brazl. The lterature revew and hypotheses are presented n secton 3, data descrpton n secton 4 and the emprcal results and the dscusson of results n secton 5. Secton 6 concludes.
13 2. TRADE CREDIT USE AND SUPPLY Some authors emphasze that trade credt s more relevant n countres wth less developed fnancal nsttutons. On the other hand, frms need to have access to fnancngs sources, especally bank debt, so that they can grant credt. Demrgüç- Kunt and Maksmovc (2001) found evdence that frms n countres wth large bankng systems borrow more on credt from ther supplers and lend more to ther customers (p. 30). Furthermore, Wlner (2000) states that trade credt s the largest source of short-term fnancng for Amercan corporatons (p. 2). It appears that trade credt has an mportant role as a fnancng source regardless of the development of the fnancal system of the country. Graph 1 shows outstandng trade credt (measured by accounts payable) for publc frms 1 n relaton to ther total assets, comparng aggregate values for Brazl, as well as for the other Latn Amercan countres and the Unted States. In comparson to these regons, trade credt seems to have a steader growth pattern 2, beng nowadays more ntensvely used n Brazl than n these other countres. Latn Amerca (excludng Brazl) presents data relatvely constant, rangng between 9.4% and 10.8%, whereas the use of trade credt n the U.S. presents decreasng trend untl 2001, remanng roughly constant at around 7.9% of total assets snce then 3. 1 The sample was composed by all publcly-lsted companes of these countres avalable n Economátca database. 2 As llustrated by graph 1, data for 2004 and 2005 appear atypcally large for Brazl; such dscrepancy was caused by the nfluence of a very lmted number of frms. However, we have opted to mantan those frms n the sample. 3 Rajan and Zngales (1995) report that trade credt represented 17.8% of total asset of all Amercan frms n 1991. However, t s mportant to note that that our sample conssts only of publc frms.
14 Graph 1 - Trade Credt Use 1997-2006 Elaborated by the authors Graph 2 shows the amount of trade credt granted by publc frms n relaton to ther total assets. Smlar to trade credt usage, trade credt supply seems to be on an ncreasng trend n Brazl. In Latn Amerca (excludng Brazl), trade credt supply seems to be roughly constant n ths ten-year perod, whereas n the US trade credt declned through 2001 and remaned relatvely constant n the subsequent years. Graph 2 - Trade Credt Offer 1997-2006 Elaborated by the authors
15 3. LITERATURE REVIEW AND HYPOTHESES 3.1 Trade Credt Theores Frms justfy supplyng trade credt on the bass that, n the absence of such fnancng, clents who do not have access to credt from fnancal nsttutons wll not buy ther products (NADIRI, 1969). Emery (1984) argues that trade credt provson boosts sales, representng a strategy to reduce frms nventores wthout prce cuts. In ths sense, such strategy would avod costs assocated wth prce reducton and preserve the frms dscretonary prcng ablty, as ponted out by Petersen and Rajan (1997). An nstrument that facltates ths practce s the trade bll, whch can be dscounted n advanced. Another relevant aspect of trade credt s that t plays a role as a product guarantee when suppler reputaton s not well known, snce clents can observe the qualty of products before effectvely payng for them (SMITH, 1987; LONG, MALITZ and RAVID, 1993). Wlner (2000) addtonally argues that, for a suppler that reles on a specfc customer, t may be n the suppler s long term nterests to grant trade credt for ths customer even f the suppler s under a temporary fnancal dstress. In ths sense, the author verfed that large buyers faced wth credt restrants may use ther poston (as a large costumer) to obtan trade credt from ther supplers. Although these reasons justfy the nterest of non-fnancal frms n extendng trade credt, t s stll nterestng to ask why they do so n the presence of fnancal frms specalzed n sellng credt. Accordng to Petersen and Rajan (1997), the explanaton s mostly related to nformatonal advantages that supplers experence, n comparson to banks, due to ther busness relatonshp wth ther clents (The supplers vsts to ther clents may provde useful hnts about nventores, employee redundancy, clents complants and other problems that the buyer may be eventually facng). Frst, supplers may obtan nformaton about ther customers n the natural course of busness, hence n a more tmely and less costly way than banks. As Cook (1997)
16 ponts out, ths can be explaned by the fact that supplers vst ther customers on a regular bass: n such vsts, supplers representatves (salespeople, commercal managers) get n touch wth buyers employees and representatves of other frms, rasng nformaton about ther relatonshp wth the buyer. Another facltatng factor s that supplers and buyers frequently belong to related ndustres, whch makes t easy for the supplers to have a good assessment of ther customers market condtons. Also, Petersen and Rajan (1997) pont out that the frequency and volume of purchases provde a relevant sgnal about the customer s stuaton. The busness relatonshp also provdes the supplers wth a montorng advantage. Man and Smth (1992) state that, faced wth a lkely default, the suppler can cut out futures sales. Ths atttude can be harmful and even trgger the frm shutdown, especally f there are few alternatve supplers of that nput. On the other hand, fnancal nsttutons retalaton does not have an effect as mmedate as the supplers actons. Knowng that supplers retalaton can lead to ther shutdown, buyers have an extra ncentve to avod defaultng on trade credt. Furthermore, Buckart and Ellngsen (2005) argue that goods are more dffcult to dvert than cash, that s, supplers can overcome moral hazard easer than banks. The last advantage mentoned by Petersen and Rajan (1997) relates to the nature of collateral assets. In case of default, lenders can seze the buyers nventory of goods: however, gven the nature of ther actvtes, such assets are less valuable to banks than to supplers partcularly when they are durable and/or have undergone lttle transformaton from the supplers nputs (see also Frank and Maksmovc, 2004 and Longhofer and Santos, 2003). Fnally, t s mportant to explan why frms use trade credt even f t s consdered more costly than other forms of credt. There are two man reasons for the nterest n ts use: (1) trade credt s an alternatve source of fnancng, especally when frms are credt constraned and (2) t s vewed as a sgnal of good qualty of frms, makng t easer for them to access bank loans. Bukart and Ellngsen s (2004) model predcts that, n the case of wealthy frms, the use of trade credt has a negatve relatonshp to wealth. Moreover, accordng to these authors, the least wealthy frms use bank loans and trade credt up to ther
17 lmt. Cunnnghan (2004) verfy emprcally of Bukart and Ellngsen s model (2004). By defnng wealth as net ncome before tax, she found that, for medum-wealth frms, as the lmt of bank credt ncreases, trade credt s less used; and for low-wealth frms, they are drectly proportonal, whch suggests that ths type of frm exhausts both lmts of bank credt and trade credt. Petersen and Rajan (1997) found that credt constraned frms use more trade credt, whle frms wth more access to credt from fnancal nsttutons extend more credt to ther customers; these results were further corroborated by Bas and Goller (1997), who also suggest that trade credt can mtgate nformaton asymmetry problems. As explaned n the prevous secton, supplers have advantage n obtanng nformaton about ther buyers n comparson to banks, and hence access to trade credt would work as a sgnal of the frm s good qualty, facltatng ther access to bank loans. The results of Cook (1997) support ths ratonale for a country wth less developed fnancal nsttutons (Russa). Alphonse, Ducret and Sévern (2006) verfy ths sgnalng role of trade credt, usng data for the US and smultaneous equatons (due to the endogenety problem between bank debt and trade credt). They argue that frms wth good relatonshp wth banks would use trade credt only as a substtute source of fnancng (substtuton hypothess), whle credt ratoned frms would use t as both, substtute and complementary source of fnancng, but n dfferent moments. These authors suggest that, n a frst moment, credt constraned frms would use trade credt as complementary source of fnancng (complementary hypothess). Banks would consder ths nformaton and become more wllng to loan to these frms (reputaton hypothess). Hence, n a second moment, these frms would use trade credt and bank loans as substtutes sources (substtuton hypothess). 3.2 Hypotheses In order to nvestgate trade credt from the pont of vew of both supplers and customers, ths research s structured nto two parts. The frst ams at testng hypotheses 1, 2 and 3 (descrbed below), whch are related to the trade credt use
18 (vewpont of the customer) and ts relaton to bank debt. The second part s composed by the hypothess 4, whch are related to trade credt offer (vewpont of the suppler). Hypothess 1: Substtuton Hypothess. Trade credt s an optonal source of fnancng to bank loan. In order to test trade credt usage, we use the model of Alphonse, Ducret and Sévern (2006). In ths test, the use of trade credt by a frm was measured by ts accounts payable to total assets rato. Bas and Goller (1997) state that, once trade credt carry nformaton owned by the suppler about ts customer, t should mtgate the nformatonal asymmetry between frms and banks. Followng ths ratonale, trade credt would play a sgnalng role, facltatng the frm s access to bank credt for credt ratoned frms. In ths sense, trade credt would be an optonal source of fnancng for both classes of frms (credt ratoned and non-ratoned). In spte of the fact that publc frms have ther nformaton dsclosed (balance sheets, fnancal statements and other type of nformaton), supplers have access to superor nformaton, even before they are publshed. Therefore, we expect that the use of trade credt can favor access to bank loans also for publc frms and accounts payable/total assets ndcator s negatvely related to total debt. Hypothess 2: Complementary Hypothess. Trade credt s a complementary source of fnancng to bank debt for credt ratoned frms. Accordng to the nformatonal and strategc advantages mentoned before, supplers could be nterested n grantng credt even for credt ratoned frms. In such case, the frm s accounts payable s expected to have a postve relatonshp wth bank debt. Hypothess 3: Reputaton Hypothess. Credt ratoned frms that use more trade credt should have easer access to bank debt than those that use less trade credt. Alphonse, Ducret And Sévern (2006) argue that the two hypotheses above are not mutually exclusve. They argue that they occur to dfferent classes of frms and n dfferent moments. Frms wth bad relatonshp wth banks would use trade credt
19 (complementary hypothess) n a frst moment. As banks notce the supplers wllngness to grant credt to the frm, they would also be more nclned to lend money to ths frm. The authors also clam that frms that beneft from long term bankng relatonshp have no ncentves to use trade credt: for these frms only the substtuton hypothess should be relevant (p.9). Accordng to the reputaton hypothess, t s expected that the coeffcent of total debt s negatve and sgnfcant for both classes of frms (frst equaton) and accounts payable s postve and sgnfcant only for frms wth worse fnancal health (second equaton). The varable used as proxy for access to credt by Alphonse, Ducret And Sévern (2006) was the duraton of the relatonshp between frms and banks. However, t s not avalable n the databases used n ths research. The varable used as proxy for credt constrant by frms was the bankruptcy predcton score adjusted by Sanvcente and Mnard (1998) for Brazlan frms. The goal here s not to dentfy whether frms are under fnancal dstress, but smply to seggregate frms between those wth better and those wth worse fnancal health. Table 1summarzes the hypotheses 1 to 3. Hypothess 4: the larger the access to fnancal sources, the larger the trade credt granted, snce the frm s not an ndspensable suppler. Petersen and Rajan (1997), based on a sample of small frms data, found that the easer ther access to credt, the more wllng they were to supply trade credt. Publc frms have access to both bank loans and publc debt. Moreover, snce they have to dsclose ther fnancal statements, they may be consdered more transparent than prvate frms, whch facltates ther access to bank loans. Hence, one can expect that publc frms have more access to fnancng sources than small and medum enterprses. We therefore expect that aggregate ndebtedness would not be as relevant to publc frms ncreasng ablty to grant credt than t s to prvate frms. Furthermore, larger frms tend to be less dependent from ther supplers than SME and for ths reason they grant less concessons to ther buyers (Wlner, 2000). Based on ths ratonale, we expect the level of total ndebtedness not to be relevant, or yet, to have a negatve relatonshp wth trade credt supply n case of publc frms.
20 Table 1- Testng Hypotheses and Related Varables Hypothess Frm Classfcaton Equaton Varable Expected Sgnal Authors 1 - Substtuton Hypothess 2 - Complementary Hypothess 3 - Reputaton Hypothess * NS = Not sgnfcant Better and worse fnancal heath Better and worse fnancal heath Better and worse fnancal heath 1 Total Debt/ Total Assets - 2 Accounts Payable/ Total Assets + 1 Total Debt/ Total Assets - Worse fnancal heath 2 Accounts Payable/ Total Assets + Better fnancal heath 2 Accounts Payable/ Total Assets NS* Nelsen (1994) and Petersen and Rajan (1995) Bas and Goller (1997) and Buckart and Ellngsen (2004) Alphonse, Ducret and Sévern (2006)
21 Damond (1991) states that frms that ssue corporate bonds have better reputaton than the others. Therefore, we expect that such reputaton amplfes the effect of non-dependence, hence bonds to total assets rato s expected to be more sgnfcant than both other loans to total assets and total debt to total assets ratos. 3.3 Varables Descrpton 3.3.1. Analyss of trade credt use In order to test hypotheses 1 to 4, we use a number of varables, the choce of whch fnds support n the lterature on trade credt. Table 2 and Table 3 summarze control varables and ther related hypotheses for trade credt use and supply respectvely. The varables descrbed below appear n the two followng equatons: (1) ACPAY + β GROWTH 6 = β + β TDEBT 1 + β PROFIT 7 + β CURR 2 + β OWNER 8 + β INCOME 3 + β SIZE 9 + β AGE 4 + β IND 10 + β SQRAGE 5 _ 1 + β11 IND _ 2 + used for testng hypotheses 1 and 3. (2) BANK + β DLEASE 6 = β + β ACPAY 0 1 + β DSTOCK 7 + β PROFIT 2 + β IND 8 + β OWNER 3 _ 1 + β 9 IND _ 2 + β SIZE 4 + β DEPR 5 + used for testng hypotheses 2 and 3 (3) ACREC β TDEBT 5 = β 0 + 1SIZE + 2 AGE + 3GROWTH + 4GR _ + β LOANS 6 β β + β BONDS 7 β β PROFIT + used for testng hypothess 4 Based on Alphonse, Ducret and Sévern (2006), we use the rato of accounts payable/total assets (ACPAY) as a proxy for trade credt. Access to other forms of credt are represented by bank debt to total assets (BANK) and total debt to total
22 assets (TDEBT). Accordng to the substtuton hypothess, the sgnal of the coeffcent for the total debt to total assets rato s expected to be negatve. The complementary hypothess predcts that the coeffcent for the rato accounts payable/total assets should be postve. Ferrs (1981) argues that uncertanty of delvery tme augment nventory needs of goods and money 4 whch contrbutes to ncrease the varaton n payments and revenues. Ths ratonale s related to transacton theory of trade credt use. The varables related to ths theory are current assets to total assets (CURR) and sales growth between 2005 and 2006 (GROWTH). These varables are drectly related to workng captal, as well as accounts payable; therefore, one can expect a postve relatonshp between them. The other varable used s the log of total ncome (INCOME). A proxy for ndustry (IND_1 and IND_2) was ncluded to control for nature of product effects, as ponted by Gannett, Burkart and Ellngsen (2007). Both varables are bnary; IND_1 represents commodty frms whereas IND_2 represents servce frms. Followng Alphonse, Ducret and Sévern (2006), the rato of earnngs before tax to total assets (PROFIT) was used as proxy for retaned earnngs 5. Accordng to the peckng order theory, the greater the frm s retaned earnngs, the lower ts dependance on external fnance (Myers, 1984). Also based on ther work, we use the frm s age as a representatve of the degree of nformatonal asymmetry and of the frm s oppacty, such as be assocated to amount of retanng earnngs, fnancal resources needs and other ssues. Hence, a predcton about the relatonshp 4 The author s argument for the ncrease n money needs s: because money must be used to complete all trades and s held at the cost of forgone nterest, stochastc money flows mply money nventory holdngs (p. 244). 5 As ponted out by these authors, ths may not be the best measure for retanng earnngs; however, the argue that t works suffcently well as a proxy for ths varable.
23 between age and accounts payable would be problematc. The measure used was the log of frms age (AGE) and the square of the log of frms age (SQRAGE). Lke frms age, frms sze can be proxy for some frms characterstcs, for example: more transparency (BERGER and UDELL, 1998), more credtworthness (PETERSEN and RAJAN, 1997) or less fnancng obstacles (BECK et al, 2006). Therefore, t s dffcult to predct the sgn of log of total assets (SIZE), the varable used as proxy for frm s sze, n both analyses. Deprecable assets (DEPR) plays collateral role. When frms credtworthness can be observed by lenders, low-qualty borrowers have to pledge collateral to obtan loans, whle hgh-qualty borrowers do not (Boot, Thakort and Udell, 1991). Two dummy varables are ncluded as control varables due to the nformatonal asymmetry (Akerloff, 1970) and consequent sgnalng effects (Spence, 1973) regardng frms qualty/credtworthness. DLEASE takes the value of 1 f the frm uses captal lease and zero owtherwse, whereas DSTOCK assumes value of 1 f the frm bennefts from loan from stockholders and zero owtherwse. The varable GROWTH, related to transacton theory, s expected to have the same sgnal that n the frst equaton (postve), as a stronger growth would demand more workng captal, mplyng an ncrease n fundng needs. The ownershpshare of the prncpal owner (OWNER) was ncluded as proxy for dfferences n fnancng patterns (HOLMES and ZIMMER, 1994). In order to nvestgate the reputaton hypothess, the sample was dvded by ther fnancal health, and the two equatons are analysed jontly. The relaton of total debt (TDEBT) s expected to be negatve for both classes of frms, whle the relaton between accounts payable (ACPAY) and bank debt would be postve and sgnfcant for credt constraned frms and non-sgnfcant for frms that present a good fnancal health. Ths corroborates the ratonale that n a frst stage of credt ratonng, frms use trade credt to substtute bank loans; at a later stage, banks would become wllng to grant credt to frms that have had access to trade credt.
24 Table 2 - Control Varables and Related Hypotheses of Trade Credt Use Varables Expected Sgnal (Equaton 1) Expected Sgnal (Equaton 2) Reference CURR + Transacton Use (Schwartz, 1974) INCOME + Transacton Use (Schwartz, 1974) AGE + Frm Transparency (Berger and Udell, 1998); Peckng Order (Myers, 1984) GROWTH + + Transacton Use (Schwartz, 1974) PROFIT - Peckng Order (Myers, 1984) OWNER + + Fnancng Patterns (Holmes and Zmmer, 1994) SIZE - + Frm Transparency (Berger and Udell, 1998) DEPR - Collateral Theory (Boot, 1991) DLEASE + DSTOCK + Asymmetrc Informaton (Akerlof, 1970); Sgnalng (Spence, 1973) Asymmetrc Informaton (Akerlof, 1970); Sgnalng (Spence, 1973) 3.3.2. Analyss of trade credt supply The analyss of trade credt supply s based on the model of Petersen and Rajan (1997), and some of the varables were redefned accordngly. We use the rato accounts recevable/total assets (ACREC) as a proxy for trade credt supply. The log of book value of total assets was used to measure frm s sze (SIZE). The frms age (AGE) has the same defnton (log of (1+frm age)). Petersen and Rajan (1997) assert that both sze and age of frms ndcates ther credtworthness, hence they would be postvely related to trade credt supply. The rato of gross proft to sales (GR_PROFIT) ndcates the ncentve that a frm have to sell. Snce trade credt s supposed to ncrease sales, we expect that the greater the gross proft, the greater the ncentve the frm has to nvest n actons that ncrease sales. For ths reason, gross proft should have a postve relatonshp wth
25 accounts recevable. Sales growth (GROWTH) has the same defnton as n the prevous secton and t should have a negatve relatonshp wth accounts recevable, once frms growth may affect workng captal management and therefore cause a declne n frms ablty of supplyng trade credt. TOTAL DEBT s the rato of total debt to total assets. BONDS s the volume of corporate bonds. LOANS s defned as the dfference between total debt rato and bonds rato. As prevously explaned, larger frms are less dependent from ther supplers and thus they would grant less concessons to ther buyers. Therefore, contrary to small enterprses, debt ratos are expected to be negatve, or at least not sgnfcant n case of publc frms. We assume that a better reputaton of frms that ssue publc debt ntensfes ths effect of non-dependence and for ths reason bonds rato coeffcent should be more sgnfcant than the other debt ratos. Table 3 summarzes control varables and ther expected sgns and related hypotheses. Table 3 - Control Varables and Related Hypotheses of Trade Credt Supply Varables Expected Sgnal Reference SIZE + Credt worthness (Petersen and Rajan, 1997) AGE + Credt worthness (Petersen and Rajan, 1997) GROWTH - Prce Dscrmnaton (Schawartz, 1974) GR_PROFIT + Prce Dscrmnaton (Schawartz, 1974)