The Impact of Management Confidence on Capital Structure. Barry R Oliver a* Australian National University, Canberra, 0200, Australia

Transcription

1 The Impac of Managemen Confidence on Capial Srucure Barry R Oliver a* a School of Finance and Applied Saisics, Faculy of Economics and Commerce, Ausralian Naional Universiy, Canberra, 0200, Ausralia Absrac This paper represens one of he firs sudies o documen he empirical relaion beween capial srucure and managemen confidence. In a unique sample of US firms ha have exised for over 25 years we find managemen confidence, as proxied by Universiy of Michigan Consumer Senimen Index, is highly significan in explaining firm financing decisions. When managemen confidence is higher firms have higher levels of deb. This resul is robus o differen definiions of leverage and confidence as well as differen model specificaions. This confirms he heoreical argumens in behavioural finance ha overconfiden managers will end o issue more deb. Marke-o-book is found o be a significan deerminan of capial srucure as documened by previous sudies. I is likely ha his is due o marke iming raher han growh opporuniies as he sample of firms used in his sudy are no expeced o be growh firms. In addiion, his sudy provides addiional evidence ha iming of equiy issues is no persisen as documened by Baker and Wurgler (2002. The lack of significance in persisence of iming of equiy issues in he sample of firms indirecly suppors he resuls of Hovakimian (2005, ha he measure is a proxy for growh opporuniies. JEL classificaion: G32 Keywords: Capial srucure, leverage, marke senimen, confidence * Conac deails: Tel.: ; fax: address: Barry.Oliver@anu.edu.au. (B. Oliver Acknowledgemens: Special hanks o Tom Smih for commens on earlier drafs. 1

2 1. Inroducion Corporae capial srucure remains a conroversial issue in modern corporae finance. Since he seminal work by Modigliani and Miller (1958, a plehora of research has been underaken in aemping o idenify he deerminans of capial srucure. Tradiionally, his has revolved around agency coss, asymmeric informaion and ransacion cos issues. Recenly, behavioral finance has begun o ake a more prominen posiion in aemping o explain aspecs of finance ha radiional research has failed o explain. For example, classical finance heory leaves no role for invesor senimen in sock reurns. A perusal of aricles forhcoming in he Journal of Finance provides ample evidence of he imporan impac ha behavioral finance is having on mainsream research in finance. The purpose of his paper is o add o he growing body of research in behavioral finance and o provide addiional empirical suppor o radiional capial srucure research by considering he impac of managemen confidence on capial srucure decisions. The paper is organized as follows. Secion 2 considers he major radiional deerminans of capial srucure and inroduces he hypohesized impac ha managemen confidence has on capial srucure. Secion 3 describes he mehod and daa and secion 4 documens he resuls. Secion 5 provides conclusions. 2. Capial srucure deerminans and variable definiions Modigliani and Miller (1958 are widely regarded as he pioneers in modeling he relevance of capial srucure o firm value. Since hen he debae has progressed from heoreical modeling o pracical realiy. I is recognized ha by relaxing he assumpions of Modigliani and Miller (1958 capial srucure is relevan o firm value. These assumpions generally relae o firm characerisics such as firm size, growh opporuniies, collaeral value of asses and profiabiliy. The empirical evidence on he impac of hese characerisics is mixed (Hovakimian, Generally, he relaion firm characerisics 2

3 have o capial srucure is consisen wih radeoff heories. 1 However, he radeoff heories are no universally acceped. Shyam-Sunder and Myers (1999 repor ha a simple pecking order model ouperforms he arge adjusmen model in explaining he ime-series variaion in capial srucure. Typically, many of he exising empirical sudies rely on firm fundamenal characerisics in explaining capial srucures, largely ignoring he possible role ha individual managers may play in capial srucure choice. Ineresingly, managemen influence on invesmen decisions has been researched quie exensively. The seminal work in his area is Roll (1986. Alhough a considerable amoun of research has been done on managemen influence on invesmen decisions, very lile has been done unil recenly on managemen influence on financing decisions. Roll (1986 uses CEO overconfidence o explain why many mergers are ex-pos value desroying. Roll s hubris hypohesis argues ha managers are oo opimisic abou he performance of heir acquisiions as hey overesimae he benefis of he akeover. This leads o excessively high bids for arges and in urn ex-pos loss making akeovers. Malmendier and Tae (2003 show ha overconfiden CEOs are more likely han raional CEOs o underake value-desroying acquisiions. The issue of overconfidence and he impac i has on invesmen decision making is reviewed in Englmaier (2004. From a financing and capial srucure perspecive, Hackbarh (2004 shows heoreically ha overconfiden managers choose higher deb levels and end o ime capial srucure decisions. Graham and Harvey (2001 surveyed 392 CFOs and found ha recen sock price performance is he hird mos popular reason for equiy issuance decisions. 2 Baker and Wurgler (2002 documen empirically ha managers are more likely o 1 See, for example Timan and Wessels (1988, Rajan and Zingales (1995 and Hovakimian ( Ineresingly, he second mos imporan issue is he amoun by which he sock is undervalued or overvalued by he marke. The mos imporan issue is earnings per share diluion. 3

4 issue equiy when heir sock marke values are higher han book and pas marke values. 3 In oher words managers ime he marke when making capial srucure decisions. However, wheher capial srucure is he cumulaive oucome of pas aemps o ime he equiy marke, and he persisence of marke iming has been debaed by Baker and Wurgler (2002, Hovakimian (2005 and Ali (2006. Irrespecive, boh Baker and Wurgler (2002 and Hovakimian (2005 find a significan negaive relaion beween leverage and marke-obook. This provides suppor for a marke iming effec- ha managers beliefs abou sock prices may influence heir decisions abou issuing deb or equiy. An alernaive view is ha his relaion is driven by growh opporuniies (Hovakimian, In relaion o confidence and sock reurns, considerable research has been conduced. Neal and Whealey (1998 examine he forecas power of hree measures of individual confidence or senimen. They find some evidence of predicabiliy of sock reurns, paricularly for small firms. Brown and Cliff (2005 find srong and consisen suppor for he hypohesis ha asse values are affeced by invesor senimen. Charoenrook (2005 also finds predicabiliy in consumer confidence or senimen and sock reurns. Therefore, if here is a relaion beween equiy values and capial srucure, and marke confidence and equiy values, wha effec does managemen confidence have on capial srucure? To explore his relaion furher we need o consider managemen decision making wih overconfiden managers. 2.1 Overconfidence Psychologiss have deermined ha overconfidence causes people o overesimae heir knowledge, underesimae heir risks and exaggerae heir abiliy o conrol evens (Nofsinger, There are numerous examples of overconfidence. For a good review see Englmaier ( This follows from Myers and Majluf (1984 and asymmeric informaion beween managemen and invesors. 4

5 Psychologiss believe ha emoions and moods ha are no relaed o he decision a hand can exer influence on ha decision. Nofsinger (2003 suggess ha he general level of opimism/pessimism in sociey affecs financial decision-makers mood and can lead o marke wide phenomenon. Opimism (and overconfidence leads corporae managers o make corporae invesmen, use more deb financing, and conduc more acquisiions. Hackbarh (2004 also shows heoreically ha overopimisic and/or overconfiden managers choose higher deb levels and issue new deb more ofen. An opimisic sociey is more willing o ake on addiional deb and increase spending. However, lile empirical research is available o deermine if a relaion exiss beween managemen confidence and capial srucure. 3. Mehod and Daa 3.1 Mehod The ordinary leas squares regression of leverage wih robus -saisics reflecing sandard errors adjused for Heeroscedasiciy is: 4 LEVERAGE α + αconf + α MB + α EFWAMB α SIZE + α TNG + α PRF + ε (1 This follows Baker and Wurgler (2002 and Hovakimian (2005 amongs ohers. Leverage for firm i a ime, is defined as: LEVERAGE Long erm deb + Shor erm deb. Rajan and Zingales (1995 discuss he pros and cons of various leverage measures. We consider he impac of alernaive proxies on he resuls laer in he paper. The independen variables used in equaion (1 are as follows. 4 Differen specificaions and ess of equaion (1 and differen definiions of he dependen and independen variables are considered laer in he paper. 5

6 3.1.1Managemen confidence (CONF There are many measures of senimen. For example, Neal and Whealey (1998 consider discouns on closed-end funds, he raio of odd-lo sales o purchases and ne muual fund redempions o reflec individual invesor confidence. Brown and Cliff (2005 use daa from Invesor s Inelligence, which racks he number of marke newsleers ha are bullish, bearish or neural and he bull-bear spread as measures of confidence. Generally, hese measures are no available over a sufficien period for meaningful resuls from equaion (1 o be obained. We proxy managemen confidence (CONF as he average of he pas 12 monhs Consumer Senimen Index (CSI from he Universiy of Michigan from 1978 o This measure of confidence is based on a direc survey of public percepions abou curren and expeced economic condiions, and i has a sufficienly long ime period for meaningful resuls o be obained. The monhly CSI is based on an ongoing naionally represenaive survey based on approximaely 500 elephone inerviews wih adul men and women living in households in he coerminous Unied Saes (48 Saes plus he Disric of Columbia. The sample is designed o maximize he sudy of change by incorporaing a roaing panel sample design in an ongoing monhly survey program. For each monhly sample, an independen cross-secion sample of households is drawn. The respondens chosen in his drawing are hen re-inerviewed six monhs laer. A roaing panel design resuls, and he oal sample for any one survey is normally made up of 60% new respondens, and 40% being inerviewed for he second ime. The CSI is an average of scores from five survey quesions ha ask respondens abou heir curren financial siuaions; he expeced change in heir financial siuaions over he nex year; heir views on expeced business condiions in he nex year and he nex five years; and wheher hey hink his is a good ime or a bad ime o make big-icke household 6

7 purchases. The acual survey quesions and consrucion mehod of he index are presened in Appendix 1. For a half-cenury, he CSI has consisenly ouperformed oher measures in anicipaing changes in he GDP six o nine monhs before hey occur (Weiss, Qiu and Welch (2005 find i he only good measure of UBS/Gallup invesor senimen. 5 Virually all published academic research on consumer senimen focuses on he CSI (Charoenrook, Furhermore, he US federal governmen has included i among is leading economic indicaors since 1989 (Weiss, We consider he CSI o be a suiable proxy for US managemen confidence or senimen Marke-o-book (MB Baker and Wurgler (2002 documen ha marke-o-book affecs leverage. Generally, firms are more likely o issue equiy when heir marke values are high, relaive o book values and repurchase equiy when heir marke values are low. Marke-o-book is defined as: MB i. -book equiy +marke equiy However, marke-o-book is also a common measure for growh opporuniies (Rajan and Zingales, Therefore, marke-o-book can be regarded as a proxy for growh opporuniies and marke iming. However, if firms ha are less likely o have growh opporuniies have a significan coefficien on MB hen i could be concluded ha marke-obook is evidence of marke iming no growh opporuniies. Firms less likely o have growh opporuniies are older more esablished firms. 5 UBS/Gallup invesor senimen index is only available from

8 3.1.2 Exernal Finance Weighed Average Marke-o-Book (EFWAMB Exernal-finance-weighed-average-marke-o-book has been used by Baker and Wurgler (2002 o measure he persisence of long-pas marke o book raios. Baker and Wurgler (2002 conend ha a negaive relaion on his variable occurs because firms do no subsequenly adjus heir leverage owards he arge, he changes in leverage induced by equiy marke iming persis. As a resul long-pas marke o book raios have a negaive effec on curren deb raios. I is defined as: EFWAMB ( e + d 1 s s xmb 1 is, s 1 ( er + dr r 1 where e and d denoe ne equiy and ne deb issued, respecively. Ne equiy issued is defined as: e (Book equiy -book equiy -1-(reained earnings -reained earnings -1 Ne deb issued is defined as: d - (bookequiy -book equiy -1-(reained earnings -reained earnings-1 Hovakimian (2005 argues ha EFWAMB is a proxy for growh opporuniies, as has been suggesed for marke-o-book, raher han a measure of marke iming persisence. If EFWAMB is a proxy for growh opporuniies hen we would expec EFWAMB for older more esablished firms o be no significanly relaed o leverage. If EFWAMB is significanly relaed o leverage for older more esablished firms (ie non growh firms hen i could be concluded ha marke iming is persisen. 8

9 3.1.3 Firm Size (SIZE Firm size has been found o be a posiive deerminan of capial srucure (Agrawal and Nagarajan, 1990; Baker and Wurgler, 2002; Hovakimian, Firm size may be a proxy for numerous variables such as lower ransacion coss in issuing deb, greaer access o deb markes and lower informaion asymmeries. The naural logarihm of oal sales is commonly considered a proxy for he size of each firm: SIZEi, Ln( Toal sales Tangibiliy of Asses (TNG The collaeral value of asses, or angibiliy of asses, held by a firm has found o be a deerminan of leverage (Rajan and Zingales, Firms wih higher angible asses are expeced o have higher leverage. Tangible asses are likely o have an impac on he borrowing decisions of a firm because hey are less subjec o informaional asymmeries and usually hey have a greaer value han inangible asses in case of bankrupcy. Addiionally, he moral hazard risks are reduced when he firm offers angible asses as collaeral, because his consiues a posiive signal o he deb holders. Following Hovakimian (2005 we define he collaeral value of asses for each firm as: TNG Propery plan and equipmen Firm Profiabiliy (PRF Myers (1984 pecking order heory of capial srucure shows ha if a firm is profiable hen i is more likely ha financing would be from inernal sources raher han exernal sources. More profiable firms are expeced o hold less deb, since i is easier and more cos effecive o finance inernally. 9

10 The variable chosen o measure profiabiliy for each firm follows Baker and Wurgler (2002: PRF Earnings before ineres, ax and depreciaion 3.2 Daa The Osiris daabase was searched for he larges 500 US indusrial companies on oal asses as of Since he Michigan Consumer Senimen Index was readily available from 1978 his began he saring year. The Compusa files were searched for he op 500 companies ha exised in 1978 and coninued o exis unil These companies made up he final common sample of 229 companies and 5,508 firm years. 6 These companies represened he larges and longes surviving firms in he US over he pas 27 years. They represen relaively lower growh firms given heir long hisory. Summary saisics for hese firms are provided in Table 1. [TABLE 1 HERE] These saisics are comparable o hose repored in Hovakimian (2005. However, he firms in our sample are hose ha have exised across he whole sample period. They are survivors, a leas from 1978 hrough The average value of managemen confidence over he sample period is approximaely 90. The maximum of occurred in he year 2000 and he minimum of 64.4 occurred in A graph of CONF and he average annual value of LEVERAGE from 1978 o 2004 is shown in Figure 1. 7 [FIGURE 1 HERE] firms for 27 years is 6,183. There were 666 observaions where daa for a specific variable in equaion (1 was no available on Compusa. Nine observaions were deleed where LEVERAGE was in excess of 100%. 7 There are 229 firms in he sample, he CONF variable is se o be he same for each firm in any given year. 10

11 Comparing Table 1 o Hovakimian (2005 firms in our sample have higher leverage, are larger in size, have higher angible asses and are more profiable han he sample used by Hovakimian (2005. As we expec for our sample of well esablished, relaively lower growh firms firms, hey have a lower average marke-o-book (1.35 v 1.54 and a lower exernal finance weighed average marke o book (1.28 v 1.63 han he firms in Hovakimian (2005. On average he sample of our firms have approximaely 30% leverage. The maximum value of leverage is 100%. Tangible asses averaged abou 50% of oal asses, and oal asses provided an average of 14% per annum reurn. A Jacques-Bera es for normaliy rejeced he null of normaliy in he disribuion for each variable, excep SIZE. This is expeced given he consrucion of his variable. 8 Table 2 repors resuls of correlaion beween variables. As shown in Table 2 mulicollineariy is unlikely o be a major concern. The variable MB has a relaively high posiive correlaion wih PRF (0.524 and a relaively high negaive correlaion wih TNG ( The regression equaion (1 was conduced excluding MB and he significance of he oher coefficiens remained similar o hose repored. 9 [TABLE 2 HERE] 8 A es for normaliy on he 27 values of CONF for each year failed o rejec he null of normaliy. 9 A redundan variables es was conduced on MB. The log-likelihood raio is 68.4 indicaing MB is highly significan. 11

12 4. Resuls The resuls of regression equaion (1 are shown in Table 3. The regression of equaion (1 is highly significan wih an adjused R 2 of 25%. This compares wih an R 2 of 17.6% in Hovakimian(2005. [TABLE 3 HERE] The coefficien on he variable CONF is posiive and significan. This indicaes ha when our proxy for managemen confidence is higher, firms have higher leverage. 10 This suppors he noion ha opimism (and overconfidence causes managers o overesimae he probabiliy of success and underesimae he risk of he decision oucomes. This leads corporae managers o use more deb financing. An opimisic sociey is more willing o ake on addiional deb and increase spending. The coefficien on marke-o-book is negaive indicaing ha when marke values are relaively high leverage is low. This suppors he marke iming resuls of Baker and Wurgler (2002 and Hovakimian (2005. Ineresingly, he exernal finance equally weighed marke o book variable (EFWAMB is no significan. Baker and Wurgler (2002 conend ha he significan negaive relaion beween EFWAMB and leverage ha hey find is because firms do no subsequenly adjus heir leverage owards he arge afer marke iming. The changes in leverage induced by equiy marke iming persis. As a resul, long-pas marke-o-book raios have a negaive effec on curren deb raios. The resuls in Table 3 for he coefficien on EFWAMB do no provide evidence of persisence in equiy marke iming. Hovakimian (2005 and Baker and Wurgler (2002 find a significan negaive relaion beween EFWAMB and Leverage. However, Hovakimian (2005 conends ha he negaive relaion is no due o persisence in marke iming bu due o effecs of growh opporuniies. 10 Noe ha he proxy for managemen confidence is lagged one period in equaion (1. 12

13 Hovakimian (2005 and Baker and Wurgler (2002 consider boh non-ipo and IPO firms. A possible reason for he insignificance in our resuls is ha our sample is comprised of all firms ha exised from 1978 o 2004, here are no IPO firms, unless hey lised in 1978, which is unlikely o be a large number. Therefore, he insignifican coefficien on EFWAMB is evidence ha marke iming does no persis and he significan coefficiens documened by Baker and Wurgler (2002 and Hovakimian (2005 are more likely due o growh opporuniies. The hree main conrol variables (size, angibiliy of asses and profiabiliy all show as having highly significan coefficiens. Firms ha are larger are significanly posiively correlaed wih higher leverage. This suppors he hypoheses ha hese firms have lower ransacion coss in issuing deb, greaer access o deb markes and lower informaion asymmeries. Firms wih higher angibiliy of asses have a posiive correlaion wih leverage. This suppors he hypohesis ha hese firms are likely o have relaively lower informaion asymmeries, lower bankrupcy risk and moral hazard risks. More profiable firms are correlaed wih less leverage. This suppors he pecking order heory of capial srucure. 4.1 Robusness ess The firs es of robusness is o consider he impac of differen specificaions of LEVERAGE. The definiion of leverage used in equaion (1 is based on book values. An alernaive specificaion is o base leverage on marke values, paricularly marke value of equiy. Leverage is redefined as: MKT LEVERAGE Long erm debi, + Shor erm debi, ( Long erm deb + Shor erm deb + Marke value of equiy This measure of leverage is applied as he dependen variable in equaion (1. The resuls are consisen wih hose previously repored. 13

14 The variable LEVERAGE (and MKT LEVERAGE can only ake on values beween zero and one. A firm wih zero leverage indicaes ha i is all-equiy financed and a firm wih leverage of 1 indicaes i is all-deb financed, echnically bankrup. The regression equaion (1 was re-run using a Censored Normal (TOBIT regression. The resuls using eiher definiion of leverage were consisen as hose previously repored. Also he regression equaion (1 was re-defined as: LEVERAGE α + αconf + α MB + α EFWAMB α SIZE + α TNG + α PRF + α LEVERAGE + ε i. 1 (2 Where ΔLEVERAGE is defined as he change in LEVERAGE from o -1. The oher variables are as previously defined. Baker and Wurgler (2002 argue ha a lagged value of LEVERAGE is required in equaion (2 as leverage is bounded beween zero and one. When leverage is near one of hese boundaries, he change in leverage can only go in one direcion, regardless of he values of he oher variables. No conrolling for lagged leverage may obscure he effecs of he oher variables. Specifically, he purpose of equaion (2 is o idenify he impac of he independen variables on curren financing decisions. The resuls of his regression are shown in Table 4. [TABLE 4 HERE] The variable CONF remains significan as does MB. This suggess ha confidence and marke-o-book raio is significanly relaed o curren financing decisions. This means ha he level of confidence and marke-o-book have a significan effec on incremenal decisions o change he curren capial srucure even afer conrolling for pre-decision leverage and oher relevan facors. The resuls in Table 4 on he conrol variables are quie differen from Hovakimian (2005, Table 7. Firm size is now insignifican as is profiabiliy. EFWAMB also remain insignifican. To ascerain if CONF is driving hese resuls, regression equaion (2 was run 14

15 excluding CONF. The resuls become remarkably similar o Hovakimian (2005, excep for EFWAMB which remains insignifican. These resuls furher suppor he conclusion ha EFWAMB is a proxy for growh and no persisence in marke iming. This also furher suppors he conclusion ha managemen confidence is a significan variable in capial srucure choice. The proxy for managemen confidence was re-defined as he measure of he Consumer Senimen Index from he Universiy of Michigan as a January of each of he years 1978 o 2004 as well as a December of he years 1978 o 2004, raher han he monhly average for he year. In boh cases he coefficien on CONF in equaion (1 remains highly significan. Considering Figure 1, i is apparen ha here is a general upward rend in CSI. LEVERAGE also appears o have a sligh upward rend. Therefore, he posiive significance of CONF may be jus indicaing ha boh leverage and confidence have increased over ime. To assess he impac of ime, a ime variable was included in equaion (1. Alhough he ime variable is significan, he confidence variable remains posiive and highly significan. Marke-o-book was redefined as: MB i. Marke equiy Book equiy This measure of marke-o-book is applied in equaion (1. The resuls are consisen wih hose previously repored. Conclusion This paper represens one of he firs sudies o documen an empirical relaion beween capial srucure and confidence or senimen. In a unique sample of US firms ha have exised for over 25 years we find ha managemen confidence as proxied by he Universiy of Michigan Consumer Senimen Index is pervasive and highly significan in explaining firm financing decisions. When confidence is higher firms have higher levels of deb. This 15

16 resul is robus o differen regression models and differen specificaions, as well as differen definiions of leverage and confidence. This suppors he heoreical argumens in behavioural finance ha overconfiden managers end o issue more deb. Marke-o-book is found o be a significan deerminan of capial srucure as documened by previous sudies. I is likely ha his is due o marke iming raher han growh opporuniies as he sample of firms used in his sudy are no expeced o be growh firms. This sudy also provides addiional evidence ha iming of equiy issues is no persisen as documened by Baker and Wurgler (2002. The lack of significance in persisence of iming of equiy issues in he sample of firms indirecly suppors he resuls of Hovakimian (2005, ha he measure is a proxy for growh opporuniies. 16

17 References Agrawal, A., and Nagarajan, N.J. Corporae capial srucure, agency coss and ownership conrol: The case of all-equiy firms, Journal of Finance 45(4, (1990, Ali, A. How persisen is he impac of marke iming on capial srucure, Journal of Finance, (2006 forhcoming. Baker, M., and Wurgler, J., Marke iming and capial srucure, Journal of Finance 57(1, (2002, Brown, G. and Cliff, M. Invesor senimen and asse valuaion, Journal of Business 78, (2005, Charoenrook, A. Does senimen maer, Working paper , (2005, Universiy of New Orleans. Englmaier, F. A brief survey on overconfidence, Working paper-economics Deparmen, (2004, Universiy of Munich. Graham, J. and Harvey, C. The heory and pracice of corporae finance: evidence from he field, Journal of Financial Economics 60, (2001, Hackbarh, D. Managerial opimism, overconfidence and capial srucure decisions, European Finance Associaion Annual Meeing, (Aug, 2004, Maasich, The Neherlands. Hackbarh, D. Managerial Trais and Capial Srucure Decisions, Social Science Research Nework, (Sepember 15, 2004, hp://ssrn.com/absrac Hovakimian, A., Are observed capial srucures deermined by equiy marke iming, Journal of Financial and Quaniaive Analysis, (2005 forhcoming. Malmendier, U. and Tae, G. CEO Overconfidence and corporae invesmen, Journal of Finance, (2005 forhcoming. Modigliani, F. and Miller, M.H. The cos of capial, corporaion finance and he heory of invesmen, American Economic Review 53, (1958, Myers, S. The capial srucure puzzle, Journal of Finance 39, (1984, Myers, S. and Majluf, N. Corporae financing and invesmen decisions: When firms have informaion ha invesors do no have, Journal of Financial Economics 13, (1984, Neal, R. and Whealey, S. Do measures of invesor senimen predic reurns?, Journal of Financial and Quaniaive Analysis 33, (1998, Nofsinger, J., Social mood and financial economics, working paper-deparmen of Finance, (2003, Washingon Sae Universiy. Qiu, L., and Welch, I., Invesor Senimen Measures, Social Science Research Nework, (June 7, 2005, hp://ssrn.com/absrac Rajan, R.G. and Zingales, L. Wha do we know abou capial srucure? Some evidence from inernaional daa, Journal of Finance 50(5, (1995, Roll, R. The hubris hypohesis of corporae akeovers, Journal of Business 59(2, (1986, Shyam-Sunder, L. and Myers, S. Tesing saic radeoff agains pecking order models of capial srucure, Journal of Financial Economics 51, (1999, Timan, S. and Wessels, R. The deerminans of capial srucure choice, Journal of Finance 43, (1988, Weiss, M. Inside consumer confidence surveys, American Demographics, (Feb,

18 Table 1 Descripive saisics of sample variables This able repors variable descripive saisics of he sample variables for: LEVERAGE Long erm deb + Shor erm deb Toalasses SENT Average of he pas 12 monhs Consumer Senimen Index from he Universiy of Michigan. EFWAMB ( e + d 1 s s i, x MB 1 i, s s 1 ( er + d r r 1 where e and d denoe ne equiy and ne deb issued, respecively. Ne equiy issued is defined as: (Book equiy -book equiy -1 -(reained earnings -reained earnings -1 e i, Ne deb issued is defined as: - (bookequiy -book equiy -(reained earnings -reained earnings di, MB i book equiy+marke equiy SIZE i, Ln ( Toal sales TNG PRF Propery plan and equipmen Earnings before ineres, ax and depreciaion LEVERAGE SENT MB EFWAMB SIZE TNG PRF Mean Median Maximum Minimum Sd. Dev Observaions 5,517 18

19 Figure 1 CSI Senimen and average Leverage (% CSI Senimen/Average leverage (% CSI Leverage Year 19

20 Table 2 Correlaion coefficiens This able repors correlaion coefficiens for he regression variables: LEVERAGE Long erm deb + Shor erm deb Toalasses SENT Average of he pas 12 monhs Consumer Senimen Index from he Universiy of Michigan. -book equiy+marke equiy MBi. 1 EFWAMB ( e + d 1 s s i, x MB 1 i, s s 1 ( er + d r r 1 where e and d denoe ne equiy and ne deb issued, respecively. Ne equiy issued is defined as: (Book equiy -book equiy -1 -(reained earnings -reained earnings -1 e i, Ne deb issued is defined as: - (bookequiy -book equiy -(reained earnings -reained earnings di, -1-1 SIZE i, Ln ( T oal sales TNG PRF i, Propery plan and equipmen Earnings before ineres, ax and depreciaion LEVERAGE SENT MB EFWAMB SIZE TNG LEVERAGE SENT MB EFWAMB SIZE TNG PRF

21 Table 3 Capial srucure deerminans This able repors he resuls of he following pooled cross-secional ime series regression (Equaion 1 on he full sample of 229 US firms from 1978 unil 2004: LEVERAGE i, α 0 + α 1 SENTi, 1 + α 2 MBi, 1 + α 3 EFWAMB i, 1 (1 + α SIZE + α TNG + α PRF + ε 4 i, 1 5 i, 1 6 i, 1 i, where LEVERAGE Long erm deb + Shor erm deb Toalasses SENT Average of he pas 12 monhs Consumer Senimen Index from he Universiy of Michigan. -book equiy+marke equiy MBi. 1 1 ( e s + d s EFWAMB i, x MB 1 i, s s 1 ( e + d r 1 r r where e and d denoe ne equiy and ne deb issued, respecively. Ne equiy issued is defined as: (Book equiy -book equiy -1 -(reained earnings -reained earnings -1 e i, Ne deb issued is defined as: - (bookequiy -book equiy -(reained earnings -reained earnings di, SIZE i, Ln ( Toal sales TNG PRF i, -1-1 Propery plan and equipmen Earnings before ineres, ax and depreciaion Sample -saisics are Whie (1980 adjused for heeroskedasiciy. Variable Coefficien -Saisic Prob. C SENT MB EFWAMB SIZE TNG PRF Adjused R-squared F-saisic Prob(F-saisic

22 Table 4 This able repors he resuls of he following pooled cross-secional ime series regression (Equaion 2 on he full sample of 229 US firms from 1978 unil 2004: LEVERAGE i, α 0 + α 1 SENTi, 1 + α 2 MB i, 1 + α 3 EFWAMB i, 1 (2 + α SIZE + α TNG + α PRF + α LEVERAGE + ε 4 i, 1 5 i, 1 6 i, 1 7 i, 1 i, where Long erm deb + Shor erm deb Long erm deb 1 + Shor erm deb 1 LEVERAGEi, 1 SENT Average of he pas 12 monhs Consumer Senimen Index from he Universiy of Michigan. -book equiy+marke equiy MBi. 1 1 ( e s + d s EFWAMB i, x MB 1 i, s s 1 ( e + d r 1 r r where e and d denoe ne equiy and ne deb issued, respecively. Ne equiy issued is defined as: (Book equiy -book equiy -1 -(reained earnings -reained earnings -1 e i, Ne deb issued is defined as: - (bookequiy -book equiy -(reained earnings -reained earnings di, SIZE, Ln ( Toal sales -1-1 i Propery plan and equipmen TNG Earnings before ineres, ax and depreciaion PRFi, Sample -saisics are Whie (1980 adjused for heeroskedasiciy. Variable Coefficien -Saisic Prob. C SENT MB EFWAMB SIZE TNG PRF LEVERAGE Adjused R-squared F-saisic Prob(F-saisic

23 Appendix 1 Hisorical daa for he Universiy of Michigan Consumer Senimen Index are available a hp:// and hp:// The procedure used o calculae he Index of Consumer Senimen (ICS as described in Howrey (2001 and Charoenrook (2003. To calculae he Index of Consumer Senimen (ICS, firs compue he relaive scores (he percen giving favorable replies minus he percen giving unfavorable replies, plus 100 for each of he five index quesions. Round each relaive score o he neares whole number. Using he formula shown below, sum he five relaive scores; divide by he 1966 base period oal of ; and add 2.0 (a consan o correc for sample design changes since he 1950s. COM1+ COM2 + COM3+ COM4+ COM 5 ICS COM denoes he componens of he index. COM 1 We are ineresed in how people are geing along financially hese days. Would you say ha you (and your family living here are beer off or worse off financially han you were a year ago? COM 2 Now, looking ahead do you hink ha a year from now you (and your family living here will be beer off financially, or worse off, or jus abou he same as now? COM 3 Now, urning o business condiions in he counry as a whole do you hink ha during he nex welve monhs we ll have good imes financially, or bad imes, or wha? COM 4 Looking ahead, which would you say is more likely ha in he counry as a whole we ll have coninuous good imes during he nex five years or so, or ha we will have periods of widespread unemploymen or depression, or wha? COM 5 Abou he big hings people buy for heir homes such as furniure, a refrigeraor, sove, elevision, and hings like ha. Generally speaking, do you hink now is a good or a bad ime for people o buy major household iems? 23