EARNINGS MANAGEMENT IN NON-PROFIT HOSPITALS - EVIDENCE FROM TAIWAN

Transcription

1 Internatonal Journal of Electronc Busness Management, Vol. 9, No. 3, pp (2011) 243 EARNINGS MANAGEMENT IN NON-PROFIT HOSPITALS - EVIDENCE FROM TAIWAN Hu-Fang Tan * Department of Health Care Admnstraton Chang Jung Chrstan Unversty Tanan (71101), Tawan PhD Program n Busness Feng Cha Unversty Tachung (40724), Tawan ABSTRACT Ths paper nvestgates whether earnngs management occurs n non-proft hosptals. Usng manually collected fnancal data of the foundaton hosptals n Tawan over the perod from 2006 to 2008, I fnd non-proft hosptals tend to manage earnngs to a range just above zero by examnng the dstrbuton of reported ncome on a sample of 45 hosptals and 133 hosptal-year observatons. The study further analyzes the characterstcs of non-proft hosptals that have reported earnngs slghtly hgher than zero. It shows that the relgon- and busness-group hosptals and hosptals that receve more donatons ntend to manage earnngs to a range just above zero. In addton, I test the assocaton between dscretonary accruals (estmated n aggregate accruals model as well as specfc accruals model) and pre-managed earnngs, the evdence suggests that managers of non-proft hosptals use ther manageral dscreton over accountng choce to meet earnngs target. These results provde evdence consstent wth non-proft hosptals smlar to for-proft organzatons, conduct earnngs management. Keywords: Non-proft Organzatons, Hosptals, Earnngs Management * 1. INTRODUCTION Non-proft organzatons are generally thought not to take the maxmzaton of ther profts as ther goal for operaton as proft may not be ther crtcal concern. However, beng dfferent from other types of donor-based non-proft organzatons, non-proft hosptals have busness behavors, for mantanng the organzaton s sustanablty whle stll provdng socal welfare functon at the same tme. Therefore, the top managers have to earn a reasonable proft, confronted wth the pressure from the board and the tenure of duty, and then face fnancal performance ncentves lke managers of other proft-orented organzatons [28,12,6,13]. Although managers n non-proft hosptals need to concern themselves wth the organzatonal fnancal performance, unlke the lsted companes n the market, non-proft organzatons do not receve captal nvestment from fnancal markets, and thus do not need to present transparent fnancal nformaton to the publc. That s, the agency problem n non-proft organzatons s worse than that n for-proft organzatons [6]. Under the montorng pressure from nterested partes, the * Correspondng author: jule@mal.cjcu.edu.tw managers wthn non-proft hosptals may undertake earnngs-management behavor to meet the earnngs goal. The lterature has mostly focused on the study of proft organzatons, and less attenton has been pad to non-proft organzatons, especally n Tawan. Thus, the purpose of ths study s to detect the earnngs-management behavor of non-proft hosptals. In Tawan, non-proft hosptals are establshed by foundaton hosptals. It s regulated that non-proft hosptals cannot dstrbute ther earnngs n any type of form. Owng to the non-proft hosptals double denttes n both the socety and the busness world; ther fuzzy characterstc enhances ths study, whch takes non-proft organzatons as ts research objects, so that ths research topc becomes an nterestng ssue worthy of study. Pror studes have showed that for-proft organzatons attempt to manage ther earnngs to acheve or match ther pre-set earnngs objectves or thresholds. Burgstahler and Dchev [7] suggested that as managers have strong ncentves nducng them to meet or beat earnngs benchmarks, the earnngs-management results n terms of earnngs dstrbuton tend to be dscontnuous, whch s named the threshold effect [10]. Burgstahler and Dchev [7]

2 244 Internatonal Journal of Electronc Busness Management, Vol. 9, No. 3 (2011) ndcated that managers of lsted companes n the US are nclned to avod reportng a defct and to avod reportng earnngs declne. Based on the research results of Burgstahler and Dchev [7], many studes have taken a step further n provng the evdence ndcatng the phenomenon that companes do have the nclnaton to avod a declne n earnngs. Other studes extensvely had the noton that managers of lsted companes tend to manpulate earnngs to avod loss [10], prevous earnngs or beat or meet analysts forecast [10,2,23]; ther man purpose s to nfluence the stock prces. Although non-proft organzatons do not need to be too concerned wth stock prces unlke for-proft organzatons, managers of non-proft hosptals receve supervson and montorng from dfferent nterested partes. If non-proft hosptals ncur losses, managers have ncentves to ncrease reportng earnngs n order to reduce the cost of debt [14] and to keep the assessment of ther management ablty to sustan the organzaton as a gong concern. On the other hand, when non-proft hosptals report hghly proftable, t would cause a greater level of scrutny as the news about ther excessve profts are made publc. Managers wll face greater pressure from the supervson of the government, whch may be turned nto forms of ncur rgorous examnaton of ther payments receved from the Natonal Health Insurance and the requrement of tax advantage and lead to the socety s doubtful atttudes towards ts non-proft poston. Thus, they may reduce ther earnngs on purpose. Leone and Van Horn [29] suggest that non-proft hosptals make 0 as ther earnngs benchmark, whch becomes ther earnng reportng strategy. In addton, they ndcated that managers from non-proft and for-proft organzatons smlar to use dscretonary accruals as a tool for earnngs management. Based on the research results of Burgstahler and Dchev [7] and Leone and Van Horn [29] and a relevant lterature revew n the feld of earnngs management, ths study manually collected the pooled cross-sectonal fnancal data of foundaton hosptals n Tawan over the perod to detect ther earnngs management behavor. The research procedures were as follows: frst, accordng to Burgstahler and Dchev [7], the author used the earnngs dstrbuton approach to test the trend of the non-proft hosptals small postve earnngs; second, ths study utlzed the probt regresson model to test whether the hosptals wth earnngs just above zero have a hgher earnngs management level than other hosptals and then to explore these hosptals characterstcs further. Moreover, ths study computed the dscretonary accruals (usng the aggregate accruals model as well as specfc accruals model) as the proxy varable for estmatng the earnngs management. The author further used the attaned proxy varable and earnngs before manpulaton to conduct the multple regresson analyss n order to understand characterstcs of the non-proft hosptals whch have the suspect to manage earnngs. The results n ths paper show that: (1) non-proft hosptal CEOs may manage earnngs just above zero. Moreover, hosptals that tend to attan zero earnngs do have sgnfcantly more dscretonary accruals. (2) Hosptals establshed by relgous organzaton or enterprses and hosptals wth hgher donaton revenue have a greater ntenton to make small postve earnngs. (3) Managers of non-proft hosptals use dscretonary accruals (both proxes of aggregate accruals and specfc accruals) as a tool to conduct earnngs management. The results show that non-proft hosptals, lke for-proft enterprses, exstng earnngs management behavor. Because the foundaton hosptals n Tawan have the propertes of non-proft organzatons, the government usually offers many types of tax advantages to them accordng to the tax codes and relevant regulatons. However, managers opportunsm-orented earnngs-management behavor may lead to the msallocaton of the economc resource or even the volaton of the non-proft organzaton s socal objectves. Therefore, ths study expects that through the extenson of the lterature n the feld of earnngs management, the nterest partes wll be able to understand the possble bas n the reportng earnngs of non-proft hosptals, and that the research results can be useful reference data for the government and other relevant nterested partes. The remander of the study s organzed as follows: Secton 2 develops the ntuton underlyng the testable hypotheses; secton 3 descrbes the emprcal desgn; secton 4 presents the source of the data and emprcal results and the fnal secton (secton 5) s the concluson of the study. 2. LITERATURE REVIEW AND HYPOTHESES Earnngs management means that n order to mslead related partes percepton of the operaton performance of the company or of the contract result connected to the content of fnancal reports, managers use ther dscreton over the accountng to affect the content of fnancal earnngs [21]. In the past, a great deal of lterature has provded evdence that the phenomenon of earnngs management does exst n prvate enterprses; however, there have been few studes on non-proft organzatons. Although the prmary operaton goal of non-proft organzatons s not proft maxmzaton, the exstng lterature has ndcated that non-proft hosptals are concerned about ther fnancal performance; some of the studes even provde

3 H. F. Tan: Earnngs Management n Non-Proft Hosptals - Evdence from Tawan 245 evdence for non-proft organzatons earnngs-management behavor. For example, n the developed object functon model of non-proft organzatons, Deneffe and Masson [11] suggested that proft and socal welfare both have postve margnal utlty. In addton, the compensaton of the chef executve offcer (CEO) n non-proft hosptals s postvely correlated wth the fnancal performance [12,28] and both the turnover of CEOs [6,13] and the turnover of boards [13] are related to poor fnancal performance, suggestng that the fnancal performance ncentves n for-proft and non-proft hosptals are smlar and thus managers of non-proft hosptals are lkely to conduct earnngs management out of earnng goal. Moreover, Trussel [38], Krshnan, Yetman and Yetman [27] and Keatng Parsons, and Roberts [24] ever carred out research on Amercan chartable nonproft organzatons (excludng hosptals) and found evdence of accountng manpulaton. Omer and Yetman [34] ndcated that the non-proft hosptals are more lkely to report near zero taxable ncome when the taxable actvty has a tax-exempt counterpart than are chartable nonprofts. Study on UK nonproft hosptals by Ballantne, Forker and Greenwood [1], and study on Amercan nonproft hosptals by Leone and Van Horn [29] both suggested that non-proft hosptals use earnngs management tools smlar to for-proft organzatons. All the lterature revewed above suggests that earnngs management does exst n non-proft hosptals. However, most of prevous studes nether smultaneously took nto account the proxes of earnngs management-aggregate dscreton accruals and specfc dscreton accruals, nor explored the attrbutes of non-proft hosptals whch undertake earnngs management for achevng the earnngs goal. In Tawan, health care ndustry s heavly regulated, so when faced wth the rapdly growng competton and changes n the natonal health nsurance system, as well as other external envronmental changes, the elements of commercalsm wll ncrease, whch lead to consstency between managers of non-proft organzatons and managers of for-proft organzatons n terms of ther operaton behavors [20]. Besdes, the transparency level of hosptals fnancal nformaton s much lower than that of lsted companes, and the accountng prncples for non-proft hosptals are dfferent from those of other ndustres. In other words, managers of non-proft organzatons have a larger space for manpulatng earnngs. Therefore, t s possble that managers of Tawanese non-proft hosptals also have earnngs-management behavors, though there has not been any relevant research so far. Ths paper extends the percepton of pror lterature and goes a step further to explore the phenomenon of earnngs management n Tawanese non-proft hosptals. The research on earnngs management can generally be dvded nto two strands: research on examnng the earnngs-threshold of the frms and research on managers utlzaton of dscretonary accruals. Ths study expects to detect the earnngs-management behavor of non-proft hosptals from these two aspects. 2.1 Earnngs Threshold n Non-proft Hosptals Effect of Earnngs Threshold n Non-proft Hosptals Burgstahler and Dchev [7] argue that managers have strong ncentves to cross over the exstng earnngs goal and proceed wth earnngs management, the manpulated earnngs-management results could lead to the phenomenon of dscontnuous earnngs dstrbuton, n whch the number of observed values n the range lower than the threshold value s fewer than expected and the number of observed values n the range hgher than the threshold value s greater than expected. Ths phenomenon provdes evdence ndcatng that lsted companes tend to avod loss or declnaton of proft. Degeorge, Patel, and Zeckhauser [10] named ths knd of phenomenon the threshold effect, whch has snce been subject to extensve research. Normally, non-proft hosptals do not make the maxmzaton of proft ther operatonal goal; they do not dstrbute earnngs, have no stock-based compensaton and lack ncentves from the captal market. Hoerger [22] suggested that the operaton of non-proft organzatons s only nfluenced by the restrctons to acheve fnancal breakeven, whch only ams to mantan the balance between revenue and expendture. For hosptals n U.S., Leone and Van Horn [29] argue that there are costs to the non-proft hosptals for reportng ether losses or profts. When they have a defct, the managers wll confront the problem of ncreasng the cost of debt, and face wth the pressure that the supervsors of the organzatons may request a change n the managers/managng team. On the other hand, when the proftablty of the organzaton s relatvely hgh, hosptals may have ther name struck off the governmental lst for tax advantages; moreover, the Natonal Health Insurance wll revew non-proft hosptals clams n a more rgorous way and the outsde donatons may decrease. Hence, most non-proft hosptals wll set the break-even (avodance of loss) pont as the earnngs goal for operaton. Compared wth the U.S. non-proft hosptals, the CEOs of foundaton hosptals n Tawan have met nternal and external pressures related to the fnancal reports about the net proft of ther organzaton. Especally when the organzatons have a great amount of ncome from operaton, the socety and mass meda wll pay much attenton to ths fact and may thus generate a negatve mpresson and a doubtful atttude towards the operaton of non-proft

4 246 Internatonal Journal of Electronc Busness Management, Vol. 9, No. 3 (2011) objectves. Thus, foundaton hosptals wll am to manpulate ther fnancal earnngs n the form of small amounts of proft, n whch the cross-sectonal earnngs are mostly dstrbuted between the ranges a lttle hgher than zero. Accordng to the dscusson and lterature revew above, ths study tests the hypotheses as follows: H1: Non-proft hosptals tend to make ther reported earnngs slghtly hgher than zero; therefore, the earnngs dstrbuton of these non-proft organzatons presents as dscontnuous near the zero ranges. H2: Non-proft hosptals that have reported earnngs slghtly hgher than zero more lkely to manage ther earnngs Relatonshps of Non-proft Hosptals and Earnngs Threshold Furthermore, ths paper scrutnzes the characterstcs of non-proft hosptals that have reported earnngs slghtly hgher than zero. As far as the donators characterstcs are concerned, n accordance wth Tawan Medcal Law, medcal foundaton refers to ncorporate foundatons that are lcensed to be establshed for medcal purpose wth certan amount of wealth donated by donors. Accordng to dfferent propertes of orgnal donators, relevant lterature are lkely to dentfy non-proft hosptals n Tawan as enterprse-based, relgon-based and general types of medcal foundatons (e.g. Wang, Chou, and Tsu [39]; L, Ln and Tan [30]). An enterprse-based medcal foundaton s funded by a for-proft busness group. Past studes verfed that transactons of related partes generally exst among subordnate enterprse wthn the group [25,19,32], and the enterprse group tends to present hgh earnngs management[26],whle Wang, Chou and Tsu [39] ndcated that managers of enterprse-based foundaton hosptals tend to operate them n the form of proft-orented busness organzatons, thus the hosptals are lkely to conduct earnngs management. Besdes, Tawanese meda and local supervson groups mght consder these non-proft hosptals as beng managed out of proft as ther earnngs present to be hgh. Therefore, managers wll endeavor to make ther earnngs low. That s one reason why these non-proft hosptals have made obvous efforts to make ther proftablty touch on the goal of zero earnngs, hence study propose the hypothess of H3a. On the other hand, a relgon-based medcal foundaton refers to a medcal foundaton founded by a relgous organzaton, whch always dedcate to mportant relgous belefs due to relgous organzatons characterstcs of relgous chartes [35]. Socety usually has a hgher level of expectaton and standards n judgng the non-proft hosptals governed by relgous organzatons, thus these non-proft hosptals are faced wth more rgorous supervson from socety. Therefore, ths study suggests that relgon-based medcal foundatons have stronger ncentves to acheve zero earnngs, hence the study propose the hypothess of H3b. H3a and H3b are as follows. H3a: Enterprse-based foundaton hosptals have stronger ntentons to acheve a net proft level that s slghtly hgher than zero. H3b: It s more apparent that relgon-based medcal foundatons tend to match the threshold value of zero earnngs. In addton to the factor of donors, those non-proft hosptals that receve more donatons are ntended to acheve earnngs mnmzaton. Trussel [38], Krshnan et al. [27] and Keatng et al. [24] all noted that the general publc s nclned to donate resources to the non-proft organzatons that perform better n terms of provdng socal welfare servces. Frank, Salkever and Mtchell [16] also found the evdence that the ncome from donatons s negatvely related to the fnancal performance of non-proft hosptals. That means as donators make decsons on the object of donatons, proftablty s a sgnfcant factor that nfluences ther decson-makng behavor; f non-proft hosptals have a hgh proftablty, they would be consdered not to need donatons from the outsde world or not to have approprately reached ther target of undertakng publc beneft actvtes. Based on the descrptons and lterature revewed above, ths study proposed that managers of non-proft organzatons attempt to make ther reported earnngs not too hgh so as to mantan or ncrease donatons, hence the hypothess of H4 s proposed as follows. H4: Non-proft hosptals that receve a hgher level of donatons have stronger ntentons to report a net proft level that s slghtly hgher than zero. 2.2 Managers Utlzaton of Dscretonary Accruals The generally accepted accountng prncples (GAAP) normally utlze an accrual bass and entrust managers wth certan flexblty n accountng polcy, whch gves them the space to manpulate and manage earnngs, and they are able to use dscretonary accruals of accountng or select the accountng method as a tools for earnngs management. The manpulaton of earnngs management through dscretonary accruals of accountng s not easly detected by the relevant and nterested partes. Much of the lterature has used dscretonary accruals as the proxy varable to detect the earnngs-management behavor of managers, and the aggregate dscretonary accruals model (ADAM) [8] to measure

5 H. F. Tan: Earnngs Management n Non-Proft Hosptals - Evdence from Tawan 247 earnngs-management behavor. The research results of ADAM have ndcated that the motves of managers workng n prvate enterprses nclude nfluencng the market prce of stocks [15,36,37] and guaranteeng the salary level and the managers postons and loan contracts; these types are contract-based ncentves. Besdes, there are also some emprcal studes regardng the earnngs-management behavor n specal ndustres n the aspects of dscretonary accrual under a specfc accountng ttle. For nstance, the accountng ttle provson for loan loss serves as a measure of management s manpulaton earnngs n banks [3] and the account Reserve for Polcy Clam Losses s usually used to measure the dscretonary accruals n nsurance ndustry [4,5,17,18,33]. In recent years, the dscretonary accrual model has been appled to non-proft hosptals. Ballantne, Forker and Greenwood [1] employed dstnct ways to measure the dscretonary accrual and found that the CEOs of NHS hosptal trusts n England tend to carry out earnngs management flexbly through accountng manpulaton to match the break-even pont between the demands of laws and the pursuts of proft. At the same tme, Leone and Van Horn [29] used both aggregate accruals and specfc accruals to detect Amercan non-proft hosptals earnngs-management behavors and dscovered that when facng ncentves sourcng from fnancal performance, the managers of these hosptals use dscretonary accruals as the tool to manpulate ther organzatonal earnngs. Research n the past has ndcated that when organzatons have a lower level of earnngs before management, to ncrease ther earnngs, the managers wll ncreasngly use the tool of dscretonary accruals to manpulate the earnngs, and vce versa. That s to say, both for-proft and non-proft organzatons use dscretonary accruals as a tool to manpulate ther earnngs. Accordngly, ths study proposes hypotheses as follows. H5a: The earnngs before management of non-proft hosptals and accrued net proft are negatvely correlated. H5b: The earnngs before management of non-proft hosptals and accrued expenses are postvely correlated. 3. RESEARCH METHOD From two aspects, ths study explores the earnngs-management behavors of non-proft hosptals; the two aspects are frst the examnaton of whether non-proft hosptals have the ntenton to manage earnngs towards a small amount of postve earnngs, and second the examnaton of whether the managers of non-proft hosptals tend to manpulate earnngs through dscretonary accruals n order to nfluence the reported earnngs. Ths study goes a step further n explorng the organzatonal characterstcs of non-proft hosptals that have earnngs-threshold psychology. The followng s a detaled descrpton of the research method adopted by ths study. 3.1 The Detecton of Earnngs Threshold In the past, the research regardng earnngs-threshold has usually employed the method of dstrbuton of earnngs frequences. Based on the research method proposed by Burgstahler and Dchev [7], ths study employs earnngs (deflated by total asset of year t-1) after management as the analytcal bass to group all sample hosptal-years nto 2 percent-wde proftablty ntervals and plots the frequency that a nonproft reports ncome n these ntervals n order to understand the frequency dstrbuton stuaton from a cross-sectonal perspectve. Burgstahler and Dchev [7] presumed that the frequency dstrbuton dagram for managers (before earnngs management) should be flat. However, f the company carres out earnngs management frequently to reach ts proftablty goal, the number of observatons that are a lttle lower than the threshold values seems to be lower more of the observatons are n the range a lttle hgher than the threshold value. These two blended phenomena result n a dscontnuty of dstrbutons near the threshold values. H1 ams to test the hypothess assumng that the earnngs dstrbuton near the zero range s contnuous. Furthermore, ths study tests the presumpton and calculates ts Z-statstc, whch estmates the dfference between each group s true and expected frequences, and then dvdes the dfferences by the standard devaton. The detaled computaton follows. Z AQ EQ (1) SD Defntons of the varables: AQ : The real number of hosptals n group ; EQ : The expected number of hosptals n group, whch s the average value of the former and latter groups. Below s the computaton formula of EQ. ( AQ 1 AQ EQ 1) (1)-1 2 SD : the estmaton of the standard devaton of the dfferences between the real frequences and the expected frequences n each group. Below s the computaton formula of SD. 1 2 [ NP (1 P ) (1/ 4) N ( P 1 P 1)(1 P 1 P 1)] SD (1)-2

6 248 Internatonal Journal of Electronc Busness Management, Vol. 9, No. 3 (2011) Where N s the number of total observatons; P s the probablty of one observaton beng n earnngs nterval ; P -1 s the probablty of one observaton beng n the former nterval; P +1 s the probablty of one observaton beng n the latter nterval. Ths study manly concerns the statstcal testng of the small negatve earnngs and small postve earnngs. Nevertheless, one of the other possble explanatons for the dscontnuous dstrbuton can be sample bas. Therefore, ths study analyzes the earnngs-management behavor of hosptals that are located wthn the range near the threshold value. At the same tme, the study also analyzes the characterstcs of hosptals located wthn the range near the threshold value. Employng probt regresson, the analyss proceeds as follows. EM = DA AMJ RELIGION GROUP RECEIPT LEV SIZE (2) 1 _ Defntons of the Varables: EM: a dummy varable whch equals 1 f the deflated earnng s postoned n the nterval of (0.0, 0.2), otherwse t equals zero. DA_AMJ: the absolute value of dscretonary accruals and the proxy varable representng the degree of earnngs management of managers. In ths study, the dscretonary accruals are estmated through the modfed Jones model [9], whch processes the regresson analyss of the total accruals on cash sales and gross fxed assets, n whch the resduals are assgned as the estmates of dscretonary accruals; the followng s the formula for estmaton. Tacc 1 Sales AR ) 2 ( PPE (2)-1 Where Tacc s total accruals, Tacc = (contnung operatng net proft cash flow from operatons); Sales s the changes n the medcal ncome; AR s the changes n the account recevables and PPE s the gross fxed assets. The varables wthn the regresson (2)-1 are all deflated by the total assets of year t-1. The absolute value of the resduals (ε) from ths model s the DA_AMJ. RELIGION: a dummy varable, representng relgon-based foundaton hosptals; the expected drecton of the sgn s postve. GROUP: a dummy varable, representng enterprse-based foundaton hosptals; the expected drecton of the sgn s postve. RECEIPT: the ncome from donatons scaled by year t-1, the expected drecton of the sgn s postve. Control Varables: LEV: the long-term labltes scaled by total assets of year (t-1), representng the degree of accommodaton, representng the fnancng dstress of the organzaton. SIZE: log of total assets, representng the hosptal sze. 3.2 Measures of Dscretonary Accruals n Detectng Earnngs Management Ths paper uses ordnary least squares (OLS) to analyze the correlaton between dfferent dscretonary accruals proxes and the correspondng earnngs before manpulatng. Both aggregate accrual model and specfc accrual model are employed to measure dscretonary accruals for conductng the analyss Aggregate Accruals Model The aggregate accruals model s based on the modfed Jones model [9]. The orgnal number (not absolute value) of the unexpected accountng accruals (.e. the estmated dfference between actual level of total accruals and the expected accruals estmated by modfed Jones model), denoted by DA_MJ, s used to measure the dscretonary accruals. The OLS equaton s demonstrated as equaton (3). DA_ MJ EBDA_ MJ LDA MJ RELIGION GROUP RECEIPT LEV SIZE YEAR (3) 1 2 _ Defnton of the Varables: DA_MJ: the orgnal number of ntegrated dscretonary accruals wthn Modfed Jones Model [9], estmated by the same procedure as equaton (2)-1. EBDAMJ: earnngs before dscretonary accruals, whch s the amount of current contnung operatng ncome deducted by DA_MJ and scaled by total assets of pror perod. The expected sgn of ß 1 s negatve. Control Varables LDA_MJ: lagged dscretonary accruals, whch s used to control for lkely autocorrelaton n dscretonary accruals. RELIGION: a dummy varable whch equals 1 representng relgon-based non-proft hosptals, and 0 otherwse. GROUP: a dummy varable whch equals 1 representng enterprse-based non-proft hosptals, and 0 otherwse.

7 H. F. Tan: Earnngs Management n Non-Proft Hosptals - Evdence from Tawan 249 RECEIPT: ncome from donaton, whch s deflated by total assets of pror perod. LEV: long-term lablty dvded by the lagged total assets. SIZE: log of total assets, representng the hosptal sze. YEAR: a dummy varable whch controllng for years varaton Specfc Accruals Model Specfc accruals Model s proposed by McNchols [31], who consdered that the hghly negatve correlaton between cash flow from operatons and change n workng captal accruals n the same perod wll lead to the problems n estmatng the aggregate accruals. The analyss of the specfc accruals model s often employed by researchers for studes on partcular ndustres. In Tawan, most of the hosptals have joned n the Natonal Health Insurance program. Therefore, the space for recognton of ncome s qute lmted so that t s more approprate to use the specfc accountng accruals model to detect earnngs management behavor n nonproft hosptals. Ths study examnes the Crtera Governng the Preparaton of Fnancal Reports of foundaton hosptals (referred to as CGPFR n the followng sectons) and selects three accrued accounts, unque for foundaton hosptals and requred judgment n determnng proper balances that have the greatest opportunty for managers to conduct earnngs management, ncludng Natonal Insurance Payment Deducton, Research and Development, Human Resource Development, and Health Educaton Expenses, and Medcal Relef, Communty Medcal Servce, and Other Types of Socal Servce Expenses. A detaled descrpton s as follows. 1. Natonal Insurance Payment Deducton: In 1995, the government mplemented a Natonal Health Insurance (NHI) program n Tawan. Health care provders clam ther medcal servces and hosptals rembursements are settled by the Bureau of Natonal Health Insurance (BNHI) after revewng hosptals clams. It can lead some elmnaton when the clams are mproper. CGPFR requres foundaton hosptals to reserve Natonal Insurance Payment Deducton for any adjustments to payments by BNHI that occurs after the year-end. Accordng to Artcle 20 of CGPFR, the account Natonal Insurance Payment Deducton s a contra revenue account and should be reported as the deducton for Medcal Revenue from the NHI payment. 2. Research and Development, Human Resource Development and Health Educaton Expenses: Accordng to Artcle 46 of the Medcal Law, each foundaton hosptal should reserve more than 10% of annual operatng ncome for support of actvtes related to educaton, tranng and research. 3. Medcal Relef, Communty Medcal Servce and Other Types of Socal Servce Expenses: Accordng to Artcle 46 of the Medcal Law, each foundaton hosptal should accrue at least 10% of annual ncome for support of actvtes related to medcal relef, communty medcal servce and other types of socal servce. Based on the CGPFR concernng the second and the thrd accounts, when hosptals accrue the provsons, the expenses and labltes should be recognzed at the same tme. The amount of these accounts nvolves consderable management dscreton. The study by Leone and Van Horn [29] selected expenses of bad debts as the sample objects for analyss. However, the revenue from the Natonal Health Insurance of hosptals n Tawan has occuped as much as 75.59% of ther total operatng revenue (on average). That s, the man ncomes and accruals of the hosptals are from payments of the Natonal Health Insurance. Before the announcement of the CGPFR, natonal nsurance payment deducton or other bad debts all belonged to expenses of bad debts; after the CGPFR were announced n 2005, the payment deducton from the Bureau of Natonal Health Insurance was put under the accountng ttle of Natonal Health Insurance Payment Deducton, and was thus deducted from the medcal revenue nstead of bad debts. Currently, the accountng ttle of Bad Debt Provson s used less frequently. Durng the sample perod, there were only 14 sample objects that desgnated the small amount of deducton of Natonal Health Insurance payment as bad debts (about 0.3% of hosptal s total revenue); therefore, ths paper does not analyze the accountng ttle Expenses of Bad Debts. Dscretonary accruals of the three suspcous accounts are estmated as the proxy of earnngs management n nonproft hosptals. The amount of these accounts s generally proportonal to current medcal revenue; therefore ths study takes the same steps as Modfed Jones Model [9] to regress the specfc accounts respectvely on current medcal revenue and the measures of dscretonary accruals are the resduals of the regresson model. To explore managers earnngs-management behavor through the specal accountng accruals, ths study process the OLS analyss as equaton (4). DA EBDA LDA RELIGION GROUP RECEIPT LEV SIZE (4)

8 250 Internatonal Journal of Electronc Busness Management, Vol. 9, No. 3 (2011) Defnton of the Varables: DA : The specal dscretonary accruals, ncludng three tems: 1. DA_INSUR: estmated dscretonary accruals of Natonal Insurance Payment Deducton, whch s the resduals of the model processng the regresson of current Natonal Insurance Payment Deducton on current Revenue from Natonal Insurance Payment and pror perod Natonal Insurance Payment Deducton. 2. DA_EDU: The dscretonary accruals of Research and Development, Human Resource Development and Health Educaton Expenses, whch s the resdual of the model when processng the regresson current Research and Development, Human Resource Development and Health Educaton Expenses on current Revenue from Natonal Insurance Payment and pror perod Research and Development, Human Resource Development and Health Educaton Expenses. 3. DA_CHARITY: The dscretonary accruals of Medcal Relef, Communty Medcal Servce and Other Types of Socal Servce Expenses, ths study uses a smlar estmatng procedure to DA_EDU, n whch the estmate of ths varable s from the regresson resduals. EBDA : earnngs before dscretonary accruals, whch equals current contnung operatng ncome plus estmated dscretonary accruals (ncludng DA_INSUR, DA_EDU and DA_CHARITY) and scaled by lagged total assets, represented by EBDAINSUR, EBDAEDU and EBDACHAR. The expected sgn of ß 1 s negatve. Control Varables: LDA : lagged dscretonary accruals, represented by LDA_MJ, LDA_INSUR, LDA_EDU and LDA_CHARITY, whch are used to control for lkely autocorrelaton n dscretonary accruals. RELIGION: a dummy varable, representng the relgon-based foundaton hosptals. GROUP: a dummy varable, representng the enterprse-based foundaton hosptals. RECEIPT: The donaton ncome, deflated by lagged total assets. LEV: The long-term lablty dvded by lagged total assets, representng the fnancng dstress of the organzaton. SIZE: Log of total assets, represents the hosptal s sze. 4. SOURCE OF DATA AND RESEARCH RESULTS In Tawan, the Department of Health (DOH) announced and ntroduced the CGPFR n 2006 to regulate medcal foundatons accountng system. Snce 2006, the accountng system has followed the general regulatng prncples and only medcal foundatons have ssued publcly fnancal statements. Therefore, ths study manually collects the fnancal reports (accredted by CPA) of medcal foundatons n 2006 to Durng the research perod, there were 51 medcal foundatons nvolved, ncludng 150 hosptal-year observatons. Elmnatng the sample objects that are not yet n formal operaton (the medcal revenue remans at the 0 level) and operated only durng the mddle of the accountng perod, stopped operaton and lacked suffcent data, the sample objects are reduced to 45 medcal foundatons and 133 hosptal years. 4.1 Descrptve Statstcs Table 1 shows the descrptve statstcs results of ths study, n whch the number of each varable s deflated by the openng total assets. The analytcal results of the descrptve statstcs ndcate that the deflated net proft tends to be 0 (ntervals between 0.00 and 0.02); 41 sample objects (hosptals) are ncluded, equal to 30.8% of the total sample objects, whch s qute a porton of the sample. DA_AMJ s the absolute value of dscretonary accruals; ts average s Some of the hosptals dd not conduct earnngs management durng the research perod (the lowest value they receved s 0); but some of the hosptals managed ther earnngs as hgh as 23.4% (the maxmum value they receved) of ther total assets. Lookng nto DA_MJ (the orgnal value of DA_AMJ), ths study found that the managed earnngs amount s between -23.4% and 13% of the hosptals total assets, n whch the negatve number represents the declned earnngs. The negatve DA_MJ means managers utlze ther dscreton to decrease pre-managed earnngs whch mples the natonal tax revenue declned. Besdes, the number of dscretonary accruals of specal accounts ncludng DA_INSUR (Natonal Insurance Payment Deducton), DA_EDU (Research and Development, Human Resource Development and Health Educaton Expenses) and DA_CHARITY (Medcal Relef, Communty Medcal Servce and Other Types of Socal Servce Expenses) s between and The other varable, RECEIPT (ncome from donaton), occupes of the total assets meanng the donaton ncome of the non-proft hosptal s low. The possble reason s that the hosptals are non-proft organzatons but wth busness operaton behavor, and when they started to run the busness, the donatons they receved from the socety declned. For ths study, there are 60 sample objects of relgon-based non-proft hosptals (RELIGION), whch contan 45 records and make up 45% of the total sample objects; 16% of the total sample objects

9 H. F. Tan: Earnngs Management n Non-Proft Hosptals - Evdence from Tawan 251 are enterprse-based non-proft hosptals (6 hosptals; 21 records; GROUP). The percentage of long-term lablty (LEV) s up to 97.4% of the total assets. There are also some hosptals that do not have any debt (20% on average). The hosptal sze (SIZE) s calculated wth the log of the current total assets, of whch the average s Table 1: Descrptve statstcs Varable N Mnmum Maxmum Mean Std Dev Sum EM DA_AMJ DA_MJ DA_INSUR DA_EDU DA_CHARITY EBDAMJ EBDAINSUR EBDAEDU EBDACHAR LDA_MJ LDA_INSUR LDA_EDU LDA_CHARITY RECEIPT RELIGION GROUP LEV SIZE Varable defntons: EM = 1 f net ncome/ lagged total assets, o otherwse; DA_AMJ = absolute value of resdual of Modfed Jones Model; DA_MJ = resdual of Modfed Jones Model; DA_INSUR = dscretonary accruals of the account Natonal Insurance Payment Deducton; DA_EDU = dscretonary accruals of the account Research and Development, Human Resource Development and Health Educaton Expenses; DA_CHARITY = dscretonary accruals of the account Medcal Relef, Communty Medcal Servce and Other Types of Socal Servce Expense; EBDAMJ = net ncome/lagged total assets less DA_MJ, representng pre-managed earnngs. EBDAINSUR = net ncome/lagged total assets less DA_INSUR; EBDAEDU = net ncome/lagged total assets less DA_EDU; EBDACHAR = net ncome/lagged total assets less DA_CHARITY; LDA_MJ, LDA_INSUR, LDA_EDU and LDA_CHARITY are varous lagged DAs; RECEIPT = ncome from donatons/ lagged total assets; RELIGION = 1 f nonproft hosptal s relgon-based, 0 otherwse; GROUP = 1 f nonproft hosptal s enterprse-based, 0 otherwse; SIZE = log (total assets). 4.2 The Dstrbuton Stuaton of the Earnngs Fgure 1 provdes a plot of the frequency dstrbuton of earnngs, n whch the net ncome are frst deflated by the pror perod total assets and then grouped nto ntervals; the dstance of each nterval s The dagram reveals an apparent spke n the dstrbuton at the group on the rght-hand sde of 0 (the ncome nterval s between 0.00 and 0.02), that means earnngs just above zero, whch has apparently the hghest number of sample objects (41 n total) 30.83% of the total sample. Table 2 demonstrates the frequency dstrbuton and the statstcal test results. Statstcal sgnfcance of the dfference n actual and expected frequency s determned usng a Z-statstc [7]. Hypothess H1 concerns the dscontnuty of earnngs dstrbuton around zero as evdence that managers ntend to manpulate ncome slghtly postve. From the Z-statstcs of table 2, t s apparent that there are two groups devatng from the other smoothly and contnuously dstrbuted sample groups. The frequency of the nterval 0 (deflated net ncome s n the nterval [-0.02, 0)) s sgnfcantly hgher than expected (Z = -3.63) and the frequency of the nterval 0.02 (deflated net ncome s n the nterval (0, 0.02]) s sgnfcantly hgher than expected (Z = 4.17). The analytcal results show that durng the research perod, managers n non-proft hosptals were nclned to make ther earnngs near 0 and to avod loss, consstent wth H1. Notes: Fgure 1 s the frequency dstrbuton dagram made up wth the net ncome deflated by the pror perod total assets; the groupng s based on the 0.02 nterval; and the total range s to Fgure 1: Earnngs frequency dstrbuton dagram

10 252 Internatonal Journal of Electronc Busness Management, Vol. 9, No. 3 (2011) Table 2: Frequency dstrbuton of earnngs Interval AQ (Actual Frequency) % EQ (Expected Frequency) % Z -stat SD (Standard Devaton) % % % % % % % % % % % % % % % % % Total % % Notes: The data (the net ncome, whch s deflated by the lagged total assets) are grouped wth 0.02 ntervals; n table 2, f the nterval between groups s 0, the deflated earnngs are located n the group of to 0. If the nterval between groups s 0.02, that means the deflated earnngs are located n the group of 0 to 0.02, and so on. The computaton formula for the expected frequency EQ, Z value and standard devaton (SD) between the AQ (actual frequency) and EQ are as follows. ( AQ AQ 1) AQ EQ EQ ; Z ; SD [ NP (1 P ) (1/ 4) N( P 2 SD 1 ) 1 / 2 1 P 1)(1 P 1 P The Characterstcs of Hosptals Reportng Earnngs are just above Zero To test further the earnngs-management stuaton of the sample hosptals that manage ther earnngs to be just above zero, as well as the characterstcs of these hosptals, ths study uses bnary logstc regresson to conduct the test and analyss (table 3). The man varables nvolved nclude DA_AMJ (estmated absolute value of the resduals by usng the modfed Jones model), RELIGION, GROUP and RECEIPT, and LEV, SIZE and YEAR (not tabulated) are control varables as ndcated n the prevous sectons. Table 3: The Characterstcs of Hosptals that Have Earnngs Slghtly Larger than Zero EM = 1 DA _ AMJ 2 RELIGION 3GROUP 4 RECEIPT 5 LEV 6 SIZE Dependent Varable: EM Level Count Varable Expected Sgn Estmate Standard Error Ch-Square Pr> Ch-Sq Intercept DA_AMJ ** RELIGION ** GROUP * RECEIPT ** LEV SIZE Number of Samples 133 Log Lkelhood **:p< 5% ; * : p<10% Notes:1.the coeffcent of varable Year are not tabluate. 2. The defntons of the varables refer to equaton (2). YEAR Table 3 reports the analytcal results of H2, H3a, H3b and H4 summarzed as follows. The estmated coeffcent of DA_AMJ s , whch acheves the 5% sgnfcance level (p = 0.029), showng that hosptals that have earnngs near 0 have sgnfcantly hgher degree of earnngs management the research results are as predcted of H2. In addton, regardng the organzatonal characterstcs of non-proft hosptals, the estmated coeffcents of RELIGION and GROUP are

11 H. F. Tan: Earnngs Management n Non-Proft Hosptals - Evdence from Tawan 253 and , respectvely (the sgnfcance levels are 5% and 10%, respectvely, wth p-values of and 0.068); the coeffcent sgn s also as expectaton of H3a and H3b, meanng that relgon- and enterprse-based foundaton hosptals may have a hgher level of concern from socety and thus receve a hgher level of ncentves to pursue the exstng earnngs goal. The estmated coeffcent of RECEIPT s (the sgnfcance level s 5%; p =0.022), suggest that hosptals wth hgher levels of ncome from donatons apparently ntend to make ther reported earnngs around zero to avod donators suspcons about the fact that ther donaton may not be well utlzed and for the purpose of mantanng or ncreasng ther donaton revenue, consstent wth H4. However, the analytcal results descrbed above may not prove well that the hosptals that have the characterstcs ndcated above wll always perform earnngs-management behavors, whch requres further research. 4.4 The Emprcal Results of Earnngs Management through Dscretonary Accruals To test whether managers of nonproft hosptals use ther dscreton n reportng earnngs, ths study at the frst undertakes the procedures of regresson analyss to estmate the dscretonary accruals (DA) of the dependent varables of equaton (3) and equaton (4). The explanatory power of dscretonary accruals estmaton model s overall sgnfcant: DA_MJ (F value = 33.99; adjusted R 2 = 33.32%; p < 0.001); DA_INSUR (F value = 60.12; adjusted R 2 = 57.33%; p < 0.001); DA_EDU (F value = 251.4; adjusted R 2 = 79.14%; p < 0.001); DA_CHARITY (F value = 10.39; adjusted R 2 = 10.78%; p < 0.001). Table 4 demonstrates the OLS results of equaton (3) and equaton (4) usng both aggregate and specfc dscretonary accruals to examne the relatonshp between the measures of managers manpulatng earnngs and earnngs before manpulated. The summarzed results follow. The emprcal test results follow. Table 4: The regresson analyss of the relaton between varous manageral dscreton to carry out earnngs management and net proft before manpulate DA j 0 1 EBDA j 2 LDA j 3 RELIGION 4 GROUP 5 RECEIPT 6 LEV 7 SIZE Dependent Varables DA_MJ Dependent Varable DA_INSUR Dependent Varable DA_EDU Explanatory Parameter Standard Explanatory Parameter Standard Explanatory Parameter Standard t Value Pr > t t Value Pr > t Varables Estmate Error Varables Estmate Error Varables Estmate Error t Value Pr > t Intercept *** Intercept Intercept EBDAMJ <.0001*** EBDAINSUR <.0001*** EBDAEDU LDAMJ ** LDA_INSUR ** LDA_EDU <.0001*** RECEIPT * RECEIPT RECEIPT RELIGION RELIGION RELIGION GROUP GROUP GROUP LEV LEV LEV SIZE *** SIZE SIZE Number of Samples Adj R F value (pvalue) 19.45(< ) Dependent Varable Explanatory Varables 6.26(<0.0001) 15.56(<0.0001) Parameter Estmate DA_CHARITY Standard Error t Value Pr > t Intercept ** EBDACHAR ** LDA_CHARITY <.0001*** RECEIPT RELIGION GROUP LEV SIZE ** Number of Samples 88 Adj R F value (p-value) 5.56(<0.0001) Notes: 1. ***:p<1% ; ** :p<5% ; * : p<10% 2. The defntons of the varables refer to equaton (3) and equaton (4). Table 4 shows the coeffcent on EBDA_MJ s sgnfcantly negatve to DA_MJ at 1%, whch suggests that managers of non-proft hosptals do use dscretonary accruals to conduct earnngs management n order to ncrease the organzatonal net proft. The coeffcents of earnngs before

12 254 Internatonal Journal of Electronc Busness Management, Vol. 9, No. 3 (2011) estmatng dscretonary accruals usng two specfc account measures, EBDA_INSUR and EBDA_CHARITY, are postve and acheve the 1% and 5% sgnfcance level respectvely, whch provde the evdence that managers tend to adjust these two correspondng accounts, Natonal Insurance Payment Deducton and Medcal Relef, Communty Medcal Servce and Other Types of Socal Servce Expenses n order to attan the earnngs goal (for example, when the net ncome before dscreton s apparently hgh, managers n nonproft hosptals would ncrease the contra ncome accounts to avod scrutny by related partes). The research results of table 4 are consstent wth the study results of Leone and Van Horn [29], as well as presumptons of H5a and H5b. Although the coeffcents of the varable EBDA_EDU match the expected drecton of sgns (postve), the statstcal analyss result s not sgnfcant n ths part. Regardng the control varables, the current dscretonary accruals level s negatvely correlated wth ts former perod level; that s consstent wth the lterature. Except for the current and last perods of the dscretonary accruals level, the test results of the other control varables are nsgnfcant. In short, table 4 s test results show that as the ncentves for earnngs management are presented to the managers of non-proft organzatons, these managers may be nduced to acheve ther earnngs-management goal through dscretonary accruals. 5. CONCLUSION Managers n non-proft hosptal have ncentves and opportuntes to manpulate reportng earnngs. To sustan as a gong concern, earnngs play an mportant role n non-proft hosptals and varous stakeholders ncludng bank, board of drectors as well as the market for CEOs use fnancal performance to evaluate hosptals performance. Ths ncurs the manageral ncentves smlar to for-proft organzatons to avod loss. Instead, non-proft hosptals n Tawan are subject to government regulators (such as ther tax-exempt status and rembursement) and concerned by other stakeholder. They are much lkely to ntend to avod hgh proftablty for reducng the supervson pressure. There are extensve accountng researches n earnngs management n for-proft settng, and less attenton are been pad to non-proft organzatons, especally n Tawan. Ths paper ams to explore earnngs management n nonproft hosptals. Usng manually collected fnancal nformaton from publc fnancal statements of foundaton hosptals n Tawan for the perod The sample s consstent of 45 hosptals and 133 hosptal years. Based on prevous lterature, ths paper examnes the dscontnuous of ncome dstrbuton to realze earnngs-threshold psychology, uses the probt regresson to analyss the characterstcs of nonproft hosptals and employs OLS regresson model testng the assocaton between dscretonary accruals and pre-managed earnngs n order to nvestgate whether the manageral dscreton from the flexblty of GAAP are utlzed to meet ther earnngs goal n nonproft hosptals. The research results of ths study are summarzed as follows. 1. The earnngs-dstrbuton pattern of the foundaton hosptals n sample perod s consstent wth the earnngs-threshold psychology, whch suggests that managers have ncentves to make ther earnngs n a range of just above zero. 2. Non-proft hosptals reportng near zero proftablty have sgnfcantly hgh level of accountng dscretonary accruals. 3. Relgon- and enterprse-based hosptals and hosptals recevng more ncome from donatons tend to manage ther earnngs to be slghtly larger than zero. 4. Usng dscretonary accruals as the ndcators n assessng the managers earnngs-management behavor, and performng an analyss based on the aggregate accruals model and specal accruals of non-proft hosptals, ths study dscovered that the two specfc varables correspondng Natonal Insurance Payment Deducton and Medcal Relef, Communty Medcal Servce and Other Types of Socal Servce Expenses, whch were the estmated dscretonary accruals and net proft before management of the two accountng accounts, are negatvely correlated wth each other. Besdes, the research results related to the dmensons of estmated dscretonary accruals and net proft before management under the two accountng ttles Natonal Insurance Payment Deducton and Medcal Relef, Communty Medcal Servce and Other Types of Socal Servce Expenses are postvely correlated. In sum, ths study provdes the evdence that non-proft hosptals and for-proft frms are smlar to conduct earnngs management and use the dscreton n fnancal reportng to manage earnngs. Because there are dfferent operatng objectves, governance and fnancal performance between non-proft hosptals and for-proft organzatons, the prncpal dfference n non-proft hosptals fnancal reportng strateges s to manage earnngs to avod loss. REFERENCES 1. Ballantne, J., Forker, J. and Greenwood, M., 2007, Earnngs management n Englsh NHS hosptal trusts, Fnancal Accountablty and Management, Vol. 23, No. 4, pp Bartov, E., Gvoly, D. and Hayn, C., 2002, The rewards to meetng or beatng earnng