Delisting Risk Analysis: Empirical Evidence from the Thai Listed Companies

Transcription

1 Advances n Economcs and Busness 4(8): , 6 DOI:.389/aeb Delstng Rsk Analss: Emprcal Evdence from the Tha Lsted Companes Thtvadee Chaawat,, Pornpan Samranruen Chualongkorn Busness School, Chulalongkorn Unverst, Thaland UOB Bank, Thaland Coprght 6 b authors, all rghts reserved. Authors agree that ths artcle remans permanentl open access under the terms of the Creatve Commons Attrbuton Lcense 4. Internatonal Lcense Abstract Delstng has rased an attenton n recent ears. Ths s because delstng ma post a negatvel drect or ndrect mpact on shareholders, managers, emploees, frms, and other stakeholders. Therefore, the needs to understand the smptoms of fnancal dstress n a compan and to be able to predct the frm delstng are crucal. The stud global objectve s to provde a fnancal operaton analss of The Stock Exchange of Thaland (SET) delstng and lstng frms and to uncover the essental varables and rsks whch ma be helpful n montorng ther corporate governance and fnancal strength. Ths stud uses logstc regresson to predct the delstng and lstng status. The data used n ths stud are drawn from the annual reports fled b SET lsted frms n Thaland. The sample conssts of SET-lsted frms n Thaland operatng from 5 to. Scope of data ncluded n ths stud s lsted frms from all ndustr groups and all sectors except fnancals ndustr group classfcaton structured as bankng, fnance and securtes, and nsurance sectors. There are man ratos, representng rsks, statstcall have an effect on delstng. Rsks whch are lkel to affect frm delstng are lqudt rsk, operatng effcenc rsk, proftablt rsk, leverage rsk, credt rsk, and fnancal nsolvenc rsk. The result shows that net workng captal to total asset, debt to equt, and gross proft margn statstcall sgnfcantl have an mpact on a delstng of SET-lsted frms. Important varables as mentoned above show that lqudt rsk, leverage rsk, and proftablt rsk have crucal mpacts on frm delstng. The - Log lkelhood s equal to 7.3, Cox and Snell R s equal to.59, and Nagelkerke R s equal to.896. Logstc regresson model above s used to classf the delstng and lstng durng the ear of 5-. The result shows that logstc regresson model s able to predct 99.% correctl classfng lstng or survvng frms and 9% correctl classfng delstng frms. In addton, earl warnng model of logstc regresson s able to predct % correctl classfng lstng or survvng frms and 83.3% correctl classfng delstng frms durng ther operaton n the ear of. Kewords Rsk, Delstng, Fnancal Ratos. Background/Objectves and Goals Delstng has rased an attenton n recent ears. Ths s because delstng ma post a negatvel drect or ndrect mpact on shareholders, managers, emploees, frms, and other stakeholders. Jarrell (984) found that delsted stocks market value declned 9% on average. Ths delstng of the lsted frms occurred both voluntarl and nvoluntarl. In the context on voluntar delstng s caused b frm experence of fnancal dstress or havng been merged or acqured. Corporate governance has been recognzed as one of ke factors assocated wth fnancal dstress (Johnson, et al., ). Therefore, the needs to understand the sgns of fnancal dstress and to be able to predct potental frm delstng are crucal. The frm would earl understand the rsks of delstng and hence would be able to prevent and/or lessen the ncomng negatve effects on the frm market value. The predcton of frm delstng emplos varet methodologes. A varet of statstcal methods and neural networks approach have been appled for predctng problems. Statstcal methods; e.g., lnear dscrmnant analss (LDA), multvarate dscrmnant analss (MDA) (Altman, 968; Snke, 975; Altman, 977; Lam and Mo, ), factor analss (FA), probt, and logstc regresson (logt) (Zmjewsk,984; Martn, 977; Ohlson, 98; Thomson, 99; Gonzalez-Hermosllo, 999; Kolar, et al., ; Montgomer et al., 5; Canbas, 5; Konstandna, 6; Dogana et al., 6) are wdel used. Ths stud emplos logstc regresson to predct the probablt of delstng of a frm from Securtes and Exchange Commsson (SEC) and attempts to assess the predcton power of logstc regresson. The stud global objectve s to provde a fnancal operaton analss of The Stock Exchange of Thaland (SET) delstng and lstng frms and to uncover the essental

2 46 Delstng Rsk Analss: Emprcal Evdence from the Tha Lsted Companes varables, representng rsks, whch ma be helpful n montorng ther corporate governance and fnancal strength. Delstng predcton models can assst SEC regulator determne whether a frm s n danger of falng and dstngush good frms from bad frms. In addton, nformaton on fnancal dstress and delstng can also help audtors decde whether a frm s a gong concern. Detecton of frm delstng s of mportance to consumers, regulators, legslators, shareholders, bondholders, audtors, and the general publc. To know n advance about the factors affectng frm delstng SEC regulator can better assess and montor the frm s delstng status and take prompt actons. The second purpose of ths stud s to provde the nformaton about the predctng accurac performance of the delstng model.. Methods The data used n ths stud are drawn from the annual reports fled b SET-lsted frms n Thaland. The sample conssts of SET-lsted frms n Thaland operatng from 5 to. Scope of data ncluded n ths stud s lsted frms from all ndustr groups and all sectors except fnancals ndustr group classfcaton structured as bankng, fnance and securtes, and nsurance sectors. Data are separated nto two groups. Frms n the frst group are subject to possble delstng and delstng frms whch SET had gven an advce to accelerate the fulfllment of The Crtera for Removal from beng Delsted or The Extenson Crtera durng 5-4. The delsted frms n the frst group are not ncluded frms whch are voluntarl delsted, merged, and not compled the SEC regulatons; e. g, submttng fnancal statement beond deadlne, fnancal statement not met audtors aspects. Frms n the second group are companes normall operatng n the same perod and sharng the same value of the assets as frms n group. The amount of group and group data are shown n Table. The dependent varable s not a contnuous varable but a bnar varable that takes a value of one when a frm s subject to possble delstng or becomes delsted and zero for survvng frms. There are man fnancal ratos representng rsks and havng an effect on delstng. Rsks whch are lkel to affect frm delstng are lqudt rsk, operatng effcenc rsk, proftablt rsk, leverage rsk, and other rsks (Jensen (986) and Wer et al. (5)). Ths stud uses fnancal ratos to prox the above rsks. Frm ma encounter lqudt rsk when frm has a chance not havng enough cash or cash equvalent to meet the needs of carrng out da-to-da operatons. Current rato (CURRENT), quck rato (QUICK), and net workng captal to total asset rato (NWCTA) are the prox for lqudt measurement. Table. Average Asset Sze and Amount of Sample n Group and Group Classfed b Industr Group durng 5- Year Industr Group Group Group Sample Sze Average Asset (Baht) Sample Sze Average Asset (Baht) 5 Agro and Food 4,36,97 9,447,585 Technolog,,584,,44 Consumer Products 44, ,79 Propert and Constructon 8,65,84 8,667,44 7 Servce 3,83,7 3,887,975 8 Agro and Food 834, ,47 Resources 8,959,96 3 8,3,73 Consumer Products,86,76 6,96,35 Propert and Constructon 3,33,69 3,4,78 9 Consumer Products,5,45 4,73,48 Servce,98,683 7,76,357 Industrals,88,563 9,845,849 Industrals 3,7, ,3,658 Technolog 3,3,66 3 9,875,645 Resources,685,4 5,89,88 Servce,47,6,48,79 Industrals,4,58 6,33,8 Consumer Products 3,5,96 3,, Total 6 4,73,45 5 3,67,448

3 Advances n Economcs and Busness 4(8): , Operatng neffcenc could be the motvaton for delstng. Generall speakng, hgh asset turnover and hgh account recevables turnover mpl that the compan s effcentl deplong asset and generatng revenue and vce versa. Hgh account recevables turnover rate could suggest that compan has a conservatve polc regardng an extenson of credt. It also ndcates that frm operates effcentl and s able to quckl collect ts debt. In addton, hgh return on asset (ROA) reflects management effectveness. On the other hand, low ROA dscredts a clear pcture of corporate fnancal health. Therefore, ths paper uses total asset turnover rato (TATURN), account recevables turnover rato ((ARTURN), and return on asset rato (ROA) as a prox of operatng effcenc rsk and credt rsk. Proft margn s one of the most mportant ndcators to compan operatonal success and fnancal strength. Low proftablt problem could damage the compan prospert. Therefore, ths paper uses gross proft margn rato (GPM), operatng proft margn rato (OPM), net proft margn rato (NPM), earnngs before nterest and tax to total asset rato (EBITTA), and retaned earnngs to total asset rato (RETA) as a prox to proftablt rsk. The use of fnancal leverage can generate both postve and negatve spllovers to the compan s return on equt as a consequence of the ncreased level of rsk. Durng the busness downturn fnancal leverage can generate a serous mpact on compan s solvenc and ma cause a frm delstng. Market captalzaton rato reflects the debt component of a compan s captal structure to support compan s operaton and growth. Market captalzaton s lkel to negatvel correlate to frm delstng. Hence, debt to equt rato (DE), total lablt to total asset rato (TLTA), and market captal to total lablt rato (MKCTL) are used as a prox of leverage rsk. Interest coverage rato (INTC) and earnngs before nterest and tax to total lablt rato (EBITTL) also should take nto account the mpact of fnancal dstress. These two ratos measure the margn of safet n order to survve future fnancal hardshp should t become materalzed. Compan wth hgh debt expense burden faces fnancal nsolvenc rsk. Therefore, the fnancal ratos mentoned above are emploed n ths stud to develop an equaton to measure the mpact on delstng and lstng. That s an equaton as follows: Y α + β x + β x + +β k x k + ε Logstc Regresson Logstc regresson s part of a categor of statstcal models called generalzed lnear model. Logstc regresson (logt) model s a tpe of predctve model that can be used when the target or dependent varable s a dchotomous or categorcal varable wth two categores. The logt model s based on the logt dstrbuton whose cumulatve probablt dstrbuton s gven b the followng expresson: p F(Z ) F(α +βx ) /(+e -Z ) /(+e - (α + βx ) where p s the probablt that SET-lsted frms n Thaland are subject to possble delstng or becomes delsted, X s explanator varables, and Z s a lner combnaton the ndependent varables. We can then estmate the logt probablt model b Z log (p /- p ) α +βx The logstc functon wth z on the horzontal axs and f(z) on the vertcal axs s shown n Fgure below. Source: Wkpeda Fgure. Logstc Functon In ths paper t s used that P takes a value of or when t s a delstng and zero otherwse. Probablt of not delstng or survvng s P and thus. Probablt functon s as follows: f ( ) P ( P ), Ths stud wants to analze the factors whch have an mpact on P whch can separate nto two parts:. Attrbutes of the choce represented b. Attrbutes of the ndvdual represented b w U and U are utlt functon of ndvdual of non-chosen alternatve and chosen alternatve, respectvel. U would represent the average utlt of ndvdual whch can be stated as follows: U U + e z δ + w γ + e U U + e z δ + w γ + e Indvdual would determne for an chosen alternatve when U > U or U U. Then and would be as shown below. f f > z

4 464 Delstng Rsk Analss: Emprcal Evdence from the Tha Lsted Companes when γ ( z z ) δ + w ( γ ) + ( e e ) δ ( z z ) δ w + e, γ γ x β + e x s explanator varables vector β s coeffcent vector e s error vector Probablt of would be equal to as follows: [ ] Pr[ > ] Pr[ e > x β ] P Pr Error term would have varet of dstrbutons. If error term has a normal dstrbuton, cumulatve probablt functon (c.d.f.) s as follows: F t / ( t) ( ) exp{ x / }dx p Therefore, t has to be estmated wth Probt functon and the cumulatve probablt functon (c.d.f.) s as follows: F ( t) + exp ( t) It makes F( t) F( t) so that probablt of s as follows: Explanator Varables Pr[ e > xβ ] [ e x β ] F( x β ) P Pr x ( β ) P F We postulate that the possblt of delstng and delstng of SET-lsted frms n Thaland s a functon of several varables. Explanator varables are fnancal ratos whch encompass measures of proftablt, leverage, lqudt, operatng effcenc, fnancal solvenc, and decomposton of assets and labltes used here to examne the delstng of the SET-lsted frms n Thaland. Factors hpotheszed to have an mpact on the SET-lsted frms n Thaland s Delstng SET-lsted frms n Thaland are the rsk-takers. The do concern about the lqudt, proftablt, growth and etc. The followng dscussed varables were prevousl tested and found to have a sgnfcant relatonshp wth frms delstng. Banks are sad to serve as fnancal ntermedares. CURRENT s Current Rato QUICK s Quck Rato NWCTA s Net Workng Captal to Total Asset Rato TATURN s Total Asset Turnover Rato ARTURN s Account Recevables Turnover Rato ROA s Return on Asset Rato GPM s Gross Proft Margn Rato OPM s Operatng Proft Margn Rato NPM s Net Proft Margn Rato EBITTA s Earnngs Before Interest and Tax to Total Asset Rato RETA s Retaned Earnngs to Total Asset Rato DE s Debt to Equt Rato TLTA s Total Lablt to Total Asset Rato INTC s Interest Coverage Rato EBITTL s Earnngs Before Interest and Tax to Total Lablt Rato MKCTL s Market Captal to Total Lablt Rato Dependent Varables The dependent varable s not a contnuous varable but a bnar varable that takes a value of one when a SET-lsted frm n Thaland s subject to possbl delsted or becomes delsted and zero otherwse. That s the dependent varable could be nterpreted as the probablt that frm wll become delstng. 3. Results Hpotheses and Expected Results Total of 6 factors expected to nfluence delstng of SET-lsted frms are to be tested n ths stud. The hpotheses are as follows: H : Current asset to current lablt s negatvel correlated wth SET-lsted frm s delstng. The low current rato reflects lqudt rsk. H : Current asset excludng nventor to current lablt s negatvel correlated wth SET-lsted frm s delstng. The low quck rato reflects lqudt rsk. H 3 : Current asset mnus current lablt to total asset s negatvel correlated wth SET-lsted frm s delstng. The low net workng captal rato reflects lqudt rsk. Lqudt rsk causes frm nablt to pa debt and havng a poor expense performance. Ths ssue ma create frms nsolvenc. H 4 : Sales to total asset s negatvel correlated wth SET-lsted frm s delstng. The hgh asset turnover reflects hgh operatng effcenc. H 5 : Sales credt to account recevables s negatvel correlated wth SET-lsted frm s delstng. The low account recevable turnover reflects hgh credt rsk and low operatng effcenc. Credt rsk can generate bad debt to the frms and fnall ma drve frms n a stuaton of default rsk. Low

5 Advances n Economcs and Busness 4(8): , recevables turnover rate ndcates compan s collecton of account recevables s neffcent. The low proporton of qualt recevables could provoke cash flow llqudt. H 6 : Net proft to asset s negatvel correlated wth SET-lsted frm s delstng. The low return on asset rato (ROA) reflects proftablt rsk. H 7 : Sales mnus cost of goods sold to sales s negatvel correlated wth SET-lsted frm s delstng. The low gross proft margn reflects proftablt rsk. H 8 : Sales mnus cost of goods sold and operatng expense to sales s negatvel correlated wth SET-lsted frm s delstng. The low operatng proft margn reflects proftablt rsk. H 9 : Net proft to sales s negatvel correlated wth SET-lsted frm s delstng. The low net proft margn reflects proftablt rsk. H : Earnngs before nterest and tax (EBIT) to total asset s negatvel correlated wth SET-lsted frm s delstng. The low EBIT margn reflects proftablt rsk. H : Retaned earnngs to total asset are negatvel correlated wth SET-lsted frm s delstng. The low net earnngs should reflect proftablt rsk and even ma cause a slow-growng busness due to a lmtaton to retan some earnngs for frm s renvestment. Ths ma mmedatel generate a negatve mpact on share prce and thus frms ma face a plent of hardshp. Proftablt rsk leves frms lmted fnancal capact and busness expanson. Hgh growth rate wth low fnancal capact causes hgh borrowng cost. Therefore, low proftablt ma nduce frm s fnancal dstress and a major challenge of survvng. Consequentl, a busness success ma enormousl be questoned. H : Debt to equt s postvel correlated wth SET-lsted frm s delstng. The hgh DE rato reflects leverage rsk. H 3 : Total lablt to total asset s postvel correlated wth SET-lsted frm s delstng. The hgh lablt rato reflects leverage rsk. H 4 : Market value to total lablt s negatvel correlated wth SET-lsted frm s delstng. The frm s hgh market value reflects lower strategc rsk. Lower proporton of debt could ncrease compan s fnancal flexblt. That s a reducton n leverage rsk and the rsk of nsolvenc. Fnancal leverage rsk can be a double-edged sword of a frm. The materal rsk would arse when compan s ROA does not exceed the nterest pament on loans. Ths would lead a crucal mpact on dmnshng profts. Therefore, confdence n frm s fnancal stablt could be badl shaken. H 5 : Earnngs before nterest and tax to cost of captal s negatvel correlated wth SET-lsted frm s delstng. The hgh nterest coverage capablt reflects nterest pament performance. H 6 : Earnngs before nterest and tax to lablt s negatvel correlated wth SET-lsted frm s delstng. The hgh value of ths rato suggests that compan has strong capablt to pa off ts ncurred debt. Frm wth an outstandng ablt to pa back ts debt shows a good determnant of strong fnancal health and lqudt poston. Therefore, lstng frm wth a sound pament capact would experence low nsolvenc rsk. Table summarzes a result of an earl warnng sstem to predct frm delstng b usng stepwse logstc regresson. Statstcall sgnfcant varable coeffcents, S.E., Wald Test statstcs, P-Value, and Exp (B) coeffcents are presented below. Table. Coeffcent Value, P-Value and the Sgn of Correlaton Coeffcent of Stepwse Logstc Regresson for Predctng Delstng, 5- Explanator Varable Coeffcent Value S.E. Wald P-Value Exp (B) NWCTA DE GPM Constant Note: Sgnfcant at percent level Sgnfcant at 5 percent level Sgnfcant at percent level Net workng captal to total asset rato, debt to equt rato, and gross proft margn rato have statstcall sgnfcantl nfluenced SET-delsted frms. Important varables as mentoned above show that lqudt rsk, leverage rsk, and proftablt rsk have crucal mpacts on frm delstng. The - Log lkelhood s equal to 7.3, Cox and Snell R s equal to.59, and Nagelkerke R s equal to.896. The model of delstng probablt of SET-lsted frms can be stated below. Y Log( OddRato) NWCTA DE.3GPM Value of Y n the equaton of would F( Y) + exp( Y ) lead to the probablt that frms would be delsted. Dfferent crtcal values show dfferent levels of correcton percentage n forecastng SET-lsted frm delstng and survvng. Logstc regresson model above s used to classf the delstng and lstng n the tranng set durng the ear of 5-. The result shows that an earl warnng model of logstc regresson s able to predct 97.8% correctl classfng lstng or survvng and delstng. The predcton of testng set, lstng and delstng frms operatng n the ear of, s 97.7% correctl. The results are summarzed n Table 3.

6 466 Delstng Rsk Analss: Emprcal Evdence from the Tha Lsted Companes Table 3. Number of Frms and Correcton Percentage of Frm Lstng and Delstng Predcton durng 5- Data Set Tranng Set Testng Set Observed Predcted Correcton Lstng Delstng Lstng Delstng Number of Lstng (Percentage) 4 (99.%) (%) Number of Delstng (Percentage) (.9%) 8 (9%) Percentage 99. Overall Correcton Percentage of Classfcaton (%) (6.7%) (%) 5 (83.3%) Overall Correcton Percentage of Classfcaton 97.7 The result n Table 3 above shows that an earl warnng model of logstc regresson s able to predct 99.% correctl classfng lstng frms and 9% correctl classfng delstng frms n tranng set whch s comprsed of 5 lstng and delstng frms durng the perod of 5-. The percentage correcton s slghtl dfferent n the testng set whch conssts 37 lstng or survvng and 6 delstng frms n the ear of. The model s able to perfectl predct lstng frms whle onl able to predct 83.3% correctl classfng delstng frm n the testng set. The goodness of ft of logstc regresson model s assessed b usng Hosmer and Lemeshow test and the hpotheses are as follows: H : The logstc regresson model fts the data H : The logstc regresson model doesn t ft the data wth a sgnfcance at 5 percentage level The result shows that p-value s greater than.5 whch means that logstc regresson model s approprate and can be used to classf frm lstng and delstng. The result of χ test statstcs s shown n Table 4. Table 4. Hosmer and Lemeshow Testng Goodness of Ft of Logstc Regresson Model b usng χ Statstcs Ch-square ( χ ) Df p-value Ths stud examnes an mpact of rsks on frm delstng. The result shows that lqudt rsk, leverage rsk, and proftablt rsk have crucal effects on delstng. Hgher net workng captal enhances survval probablt of lsted frms. In addton, frms wth hgher gross proft margn have a lower probablt of delstng. On the other hand, lsted frms have a negatve mpact from hgh leverage. Hgh debt to equt wll provoke frms to be delsted. REFERENCES [] Altman, E. I. (968). Fnancal Ratos, Dscrmnant Analss and the Predcton of Corporate Bankruptc, The Journal of Fnance, 3, [] Altman, E. I. (977). Predctng Performance n the Savngs and Loan Assocaton Industr, Journal of Monetar Economcs, 3, [3] Canbas, S., Cabuk, A. and Klc, S. B. (5). Predcton of Commercal Bank Falure va Multvarate Statstcal Analss of Fnancal Structure: The Turksh Case, European Journal of Operatonal Research, 66, [4] Dogana, M. M., Celan, N. B., and Aktas, R. (6). Predctng Fnancal Falure of the Turksh Banks, Annals of Fnancal Economcs,, [5] Gonzalez-Hermosllo, B. (999). Determnants of Ex-Ante Bankng Sstem Dstress: A Macro-Mcro Emprcal Exploraton of Some Recent Epsodes. Internatonal Monetar Fund Workng Paper No. 33. [6] Jarrell, G. A. (984). The Stock Prce Effects of NYSE Delstng for Volatng Corporate Governance Rules. U.S. Securtes and Exchange Commsson Workng Paper. [7] Jensen, M. C. (986). Agenc Cost of Free Cash Flow, Corporate Governance, and Takeovers, Amercan Economc Revew, 76(). [8] Johnson, S., Boone., P. Breach, A. and Fredman, E. (). Corporate Governance n the Asan Fnancal Crss, Journal of Fnancal Economcs, 58, [9] Kolar, J., Glennon, D., Shn, H., and Caputo, M. (). Predctng Large US Commercal Bank Falures, Journal of Economcs and Busness, 54(3), [] Konstandna, N. (6). Probablt of Bank Falure: The Russan Case. Economc Educaton and Research Consortum Workng Paper No.. [] Lam, K. F. and Mo, J. W. (). Combnng Dscrmnant Methods n Solvng Classfcaton Problems n Two-Group Dscrmnant Analss. European Journal of Operatonal Research, 38, [] Martn, D. (977). Earl Warnng of Bank Falure: A Logt regresson Approach, Journal of Bankng and Fnance,,

7 Advances n Economcs and Busness 4(8): , [3] Montgomar, H., Hanh, T. B., Santoso, W., and Besar, D. S. (5). Coordnated Falure? A Cross-Countr Bank Falure Predcton Model. ADB Insttute Dscusson Paper No. 3. [4] Ohlson, J. A. (98). Fnancal Ratos and Probablstc Predcton of Bankruptc, Journal of Accountng Research, 8, 9-3. [5] Snke, J. F. (975). A Multvarate Statstcal Analss of the Characterstcs of Problem Banks, The Journal of Fnance, 3 (), -36. [6] Thomson, J. B. (99). Predctng Bank Falures n the 98s, Economc Revew-Federal Reserve Bank of Cleveland, 7(), 9. [7] Wer, C. M., Lang, D., and Wrght, M. (5). Undervaluaton, Prvate Informaton, Agenc Cost and the Decson to go Prvate, Appled Fnancal Economcs, 5, [8] Zmjewsk, M. E. (984). Methodologcal Issues Related to the Estmaton of Fnancal Dstress Predcton Models, Journal of Accountng Research,, 59-8.