EARNINGS SMOOTHING, EXECUTIVE COMPENSATION, AND CORPORATE GOVERNANCE: EVIDENCE FROM THE PROPERTY-LIABILITY INSURANCE INDUSTRY

Transcription

1 EARNINGS SMOOTHING, EXECUTIVE COMPENSATION, AND CORPORATE GOVERNANCE: EVIDENCE FROM THE PROPERTY-LIABILITY INSURANCE INDUSTRY David L. Eckles Marin Halek and Rongrong Zhang July

2 Insurer Reserve Error and Execuive Compensaion David L. Eckles, Marin Halek, and Rongrong Zhang July 2007 JEL Classificaion: C23, G22, J33, M41 Absrac This paper invesigaes incenives ha managers of insurance firms have o manipulae accouning resuls in order o maximize heir oal compensaion. Insurance company execuives are in a relaively unique posiion in ha hey are able o impac heir oal compensaion by exercising discreion in loss reserve pracices. We find evidence suggesing ha managers whose compensaion packages are more incenive-laden are associaed wih larger loss reserve errors. Specifically, we find managers ha receive larger bonuses and sock awards end o have larger loss reserve errors. We specifically show ha managers who hold more sock or exercise more sock opions end o undersae loss reserve. In addiion, we examine he monioring effec of he corporae board srucure in miigaing managers reserve manipulaion pracices. We find ha managers are more likely o manipulae reserves in he presence of a weak board. David L. Eckles Terry College of Business Deparmen of Insurance, Legal Sudies, and Real Esae Universiy of Georgia 206 Brooks Hall Ahens, GA deckles@uga.edu Marin Halek Terry College of Business Deparmen of Insurance, Legal Sudies, and Real Esae Universiy of Georgia 206 Brooks Hall Ahens, GA mhalek@uga.edu Rongrong Zhang College of Business Adminisraion Deparmen of Finance and Quaniaive Analysis Georgia Souhern Universiy PO Box 8151 Saesboro, GA rzhang@georgiasouhern.edu 2

3 1. Inroducion This paper invesigaes he relaionship beween he compensaion of insurance company execuives, he board srucure of insurance firms and he loss reserve pracices of insurance firms. In he wake of public scandals surrounding accouning pracices of major corporaions (including a leas one insurer, AIG), he Unied Saes Congress passed he Sarbanes-Oxley Ac (SOX) in One significan oucome of SOX requires hose signing a firm s financial saemens o be accounable for he informaion included. Previously, execuives, direcors, officers, ec. could claim ignorance of he accouning decisions and oher such pracices made a lower levels of managemen. Perhaps as a resul of his pre-sox veil of proecion, execuives were able o engage in various accouning manipulaions for a variey of reasons. Along wih he criminally prosecued fraud involving accouning manipulaions a Enron, WorldCom, and oher firms, academic researchers have found evidence of more suble manipulaion of accouning resuls. Raionale for such manipulaion includes avoiding regulaory scruiny (for insurers, Grace (1990) and Peroni (1992)), smoohing ax liabiliies (Grace (1990), Peroni (1992), and Peroni and Shackelford (1999)), and even increasing he compensaion of execuives (Healy (1985), Holhausen e al. (1995), and Eckles and Halek (2007)). Addiionally, here exiss a lieraure invesigaing he oversigh capaciy ha corporae governance mechanisms have in he manipulaion of earnings. Some find abnormal accruals are negaively associaed wih boh audi commiee independence and board independence (Klein 2002). While ohers show earnings manipulaion is less likely in firms wih more ouside direcors (Xie e al. (2003) and Peasnell e al. (2005)). However, he execuive compensaion/earnings managemen lieraure does no consider he oversigh of corporae governance mechanisms nor does he earnings managemen/corporae governance lieraure consider incenives of execuives as a resul of compensaion packages. To 3

4 our knowledge, Corne, Marcus, and Tehranian (2007) is he only paper o invesigae earnings manipulaion wih respec o boh corporae governance oversigh and managerial compensaion incenives. Alhough Corne, Marcus, and Tehranian (2007) is close in spiri o our curren sudy, we are able o uilize acual errors made in generaing published accouning resuls. Specifically, sauory reporing for insurers requires insurers o esimae losses, and requires he insurers o ulimaely repor he observed values of he esimaed losses. As such, here is no need o esimae he accouning error, as is done in many sudies regarding accouning manipulaion. We fuse hese hree relaed lieraures o deermine wheher corporae governance sraegies can manage execuives and heir incenives o manipulae earnings. We hypohesize ha he well documened incenives managers have o manipulae earnings is mued (or exacerbaed) when corporae governance is srong (weak). Specifically, we find board monioring and execuive compensaion play imporan roles in curbing managers managemen of insurance companies earnings. The remainder of he paper is organized as follows. Secion 2 discusses relaed lieraure. Secion 3 develops he research hypoheses and he economic models. Secion 4 provides deails on he sample daa and analysis mehodology. Secion 5 presens he empirical resuls, and he final secion concludes. 2. Relaed Lieraure Our paper merges he noions ha one, managers use discreionary accouning componens o affec firm performance (for regulaory avoidance, ax smoohing, or compensaion enhancemen) and wo, ha firms wih sronger boards can exer more conrol over 4

5 managers, hereby poenially miigaing some of hese effecs. We provide a brief review of hese separae lieraures. Wihin he insurance lieraure, managers have been shown o use discreionary loss reserve esimaes in an effor o influence earnings reporing (Beaver e al. (2003) and Grace (1990)). Grace (1990) shows ha insurers use reserve esimaion o smooh earnings in an effor o minimize heir ax liabiliy. Beaver e al. (2003) also find evidence of insurers using reserve esimaion o minimize heir ax liabiliy and o manipulae heir earnings repors. Specifically, hey show ha insurers are more likely o manipulae earnings o avoid having o repor small losses. Gaver and Paerson (2004) and Peroni (1992) show insurers manipulaing earnings in an aemp o emporarily avoid regulaory scruiny. Boh Gaver and Paerson (2004) and Peroni (1992) find financially weak insurers under-reporing heir esimaed losses in an effor o avoid riggering regulaory inervenion. There are also several sudies ha consider he effecs of managerial compensaion packages on earnings manipulaion. Healy (1985), Gaver e al. (1995), and Holhausen e al (1995) find evidence of managers manipulaing earnings in an effor o maximize heir performance-based compensaion. Healy (1985) finds evidence consisen wih managers increasing he earnings o increase heir bonus or decreasing he earnings (o have more room in he fuure) when he bonus is eiher ou of he money or capped. 1 Gaver (1995) finds evidence ha managers wih bonus plans make earnings-increasing decisions, even if he firm s performance is below he hreshold ha would resul in a payou from he bonus plan. Holhausen e al. (1995) find evidence of earnings-increasing manipulaion when managers are receiving bonus payous and earnings-decreasing manipulaion when he bonus plans have been 1 Mos bonus schemes are subjec o minimum and maximum performance requiremens, and hence have minimum and maximum payous. Healy shows ha only when managers are obaining more income from he bonus do hey manipulae he resuls o increase earnings. 5

6 capped. Holhausen e al. (1995) find no evidence of manipulaion when he bonus plans have no been riggered. In he insurance lieraure, Eckles and Halek (2007) find evidence of more earnings manipulaions by managers of sock propery/casualy insurers who have more incenive-based compensaion componens. Specifically, insurer accouning errors are shown o be relaed o bonus payous, exercising sock opions, and he amoun of sock held by managers. Browne, Ma, and Wang (2007) also find insurer reserve errors o be relaed o he opions graned o execuives. However, none of hese papers consider he role of corporae governance mechanisms in overseeing managerial behavior or seing he erms of he execuive compensaion packages. Corporae governance lieraure offers several sudies on he relaionship of earnings manipulaion and corporae governance srucure. Klein (2002) finds abnormal accruals are negaively associaed wih boh audi commiee independence and board independence. Xie, e al (2003) sudy a sample of firms from he S & P 500 and find, among oher resuls, ha earnings manipulaion is less likely for hose firms whose boards have more ouside direcors. Using a sample of Unied Kingdom firms beween 1993 and 1996, Peasnell e. al. (2005) documen a negaive relaion beween he likelihood of managers making income-increasing abnormal accruals o avoid reporing losses and earnings reducions, and he proporion of ousiders on he board. However, none of hese papers consider ha he compensaion packages of execuives, which ofen lead o accouning manipulaions, are hemselves a resul of decisions made by boards of direcors. Finally, Corne, Marcus, and Tehranian (2007) do consider he effecs managerial compensaion packages have on accouning manipulaion and he role corporae governance srucures play in seing hese compensaion packages in addiion o providing oversigh wih 6

7 respec o he accouning decisions made. As do mos sudies ha uilize non-insurance accouning daa, Corne, Marcus, and Tehranian (2007) esimae heir measure of earnings manipulaion. Our curren paper exends Corne, Marcus and Tehranian (2007) by uilizing acual accouning errors made by insurers in prior periods. To our knowledge, his is also he firs sudy o consider he combined effecs of boh corporae governance and execuive compensaion on earnings manipulaion in insurance companies. 3. Hypoheses Gaver e al. (1995), Holhausen e al. (1995), and Healy (1985) all indicae some level of manipulaion of repored earnings by managers hrough he use of discreionary accouning pracices. Presumably, his manipulaion is ulimaely done in order o maximize managers oal uiliy. While benefis of ax and regulaory avoidance cerainly indirecly accrue o managers, he incenive-based compensaion componens are equally affeced by earnings manipulaion and accrue direcly o managers. Loss reserve esimaion by insurance execuives is one such discreionary pracice. Therefore, ceeris paribus, we posulae ha insurance execuives who derive a larger porion of heir oal compensaion from incenive-based componens will have larger reserve esimaion errors. Furher, recen corporae governance lieraure documens ha managers of firms wih srong (weak) governance mechanisms are less (more) likely o manipulae earnings. For example, Klein (2006) finds ha earnings managemen is posiively relaed o CEO s siing on he board's compensaion commiee, and earnings managemen is negaively relaed o CEO s equiy holdings. There also appears o be less earnings managemen when a large ouside shareholder sis on he board's audi commiee. She concludes ha hese resuls sugges ha boards srucured o be more independen of he CEO may be more effecive in monioring he corporae financial accouning process. Corne, Marcus, and Tehranian 7

8 (2007) show ha sock opions compensaion encourages earnings managemen, and board independence and insiuional monioring discourage earnings manipulaions. In his secion, we examine how various execuive compensaion componens and several corporae governance characerisics relae o execuives discreionary accouning pracices in insurance firms. The compensaion componens examined include bonuses, resriced sock held, sock opions exercised, and sock opions awarded. Prior research indicaes ha each componen does no necessarily induce he manager o ac in a consisen manner. We herefore disinguish beween hose compensaion elemens ha should induce over-reserving (e.g. sock awards), under-reserving (e.g. exercising opions), and eiher over- or under-reserving (e.g. srucured bonus plans). 3.1 Execuive Compensaion and Reserve Errors Several observable compensaion iems can be characerized as long-erm incenive schemes. Compensaion componens mean o align he long-erm incenives of he execuive and company are he value of opions and resriced sock awarded o he manager during he year. These are no opions (or sock) immediaely convered o cash, bu raher he value of he securiies given o he manager ha heoreically aligns he manager s incenives wih he long run incenives of he firm. The ulimae value of hese securiies is no obained during he year in which hey are awarded, bu raher will be realized in he fuure, coningen on he overall value of he firm. Hence, long erm compensaion variables should provide moivaion for managers o over-reserve for losses in he curren year. Over-reserving creaes he percepion ha he insurer is safe (i.e. has more surplus) and allows fuure resuls, such as lower realized losses han expeced, o have a posiive impac on he firm s value in subsequen years. Over- 8

9 reserve leads o negaive reserve errors. Hence, we predic a negaive associaion beween he percen of long-erm compensaion and he reserve error. Formally, our firs esable hypohesis is saed as: H1: Firms whose managers have larger proporions of long-erm compensaion componens relaive o heir oal compensaion in period, are more likely o over-reserve for losses in period in order o shif firm value ino period +n, when he long-erm compensaion is realized. In addiion o long-erm compensaion componens, we observe incenive based compensaion elemens ha are realized or exercised during he curren year, such as bonuses and exercised sock opions. Resuls from Gaver e al. (1995), Holhausen e al. (1995), and Healy (1985) indicae managers may have incenives o make boh earnings-increasing as well as earnings-decreasing decisions as a resul of hese compensaion elemens. Their argumens hinged upon where firm earnings were relaive o hresholds se in he compensaion packages. A manager was shown o have an incenive o implemen earnings-decreasing pracices if he earnings were above an upper bound se by he compensaion packages (Healy (1985), Gaver e al. (1995), and Holhausen e al. (1995)) or if he earnings were well below he hreshold where he manager would receive some performance based compensaion (Healy (1985)). When earnings are ouside of hese hresholds, he manager does no gain a larger bonus by making earnings-increasing decisions. Alernaively, managers were shown o implemen earnings- 9

10 increasing policies if he firm s earnings were in he range where increased earnings would lead o increased compensaion (Healy (1985), Gaver e al. (1995), and Holhausen e al. (1995)). 2 In our sample, execuives are given wo ypes of bonus plans. The mos common bonus plan which we simply refer o as bonus plans are hose described above (i.e. hose plans sudied in Holhausen e. al (1995), Healy (1985)). The execuives in our sample are also given bonuses ermed long-erm incenive plans. These plans are essenially bonuses paid in one year based upon he performance in pas years. Unforunaely, our daa does no provide sufficien deail regarding he srucure of he execuive bonus plans or long-erm incenive plans in order for us o make absolue predicions on he effecs hese plans may have on loss reserving pracices. 3 We furher remove he long-erm incenive plan bonuses from consideraion since prior research has found hem o be insignifican wih respec o reserving pracices (Eckles and Halek (2007)) and since hey make up a very small proporion of he execuive compensaion packages for our execuives (around 4% of oal compensaion). Wih respec o he bonus plans, managers may have incenives o eiher over- or under-reserve, depending on where he firm s performance falls relaive o he srucure of he bonus plan. Managers have an incenive o under-reserve losses (i.e. make earnings-increasing decisions) if operaing wihin he bonus hreshold and upper limi of he bonus. However, managers have an incenive o over-reserve (i.e. make earnings-decreasing decisions) if he upper limi of he bonus has already been 2 The payoff schemes of he bonuses invesigaed here and by Healy (1985), Gaver e al. (1995), and Holhausen e al. (1995) resemble a generic call spread. Tha is, he payoff of he bonus is fla (presumably zero) unil he firs srike price (bonus hreshold) is reached. The payoff hen increases (hough no necessarily in a linear fashion) unil he second srike price (he upper bound of he bonus) is reached, a which poin he bonus payoff again flaens. 3 Alhough we do no have specifics on he hresholds and upper bounds of all of he compensaion packages, we do observe he plan deails for a subse of he firms in he daa. The bonus schemes repored in he 10-k s of hese firms are exacly he same as he schemes sudied in Healy (1985), Gaver e al. (1995), and Holhausen e al. (1995). In hese bonus schemes, here exiss a lower and upper bound of earnings beween which managers are rewarded a bonus, and ouside of which he bonus is eiher no paid or capped. Alhough we are unable o obain he specifics of every insurer s bonus scheme, for every year, we are confiden ha his common bonus scheme is generally employed by our sample of insurers. 10

11 aained or if he bonus hreshold is unaainable (Holhausen e. al (1995), Healy (1985)). We herefore es he wo compeing hypoheses: H2(a): Condiional on receiving income from a bonus plan, managers who have exhaused compensaion from hese plans in period will have an incenive o make earnings-decreasing reserving decisions, and will herefore over-reserve losses incurred in period. H2(b): Condiional on receiving income from a bonus plan, managers who have no reached heir plan limi in period will have an incenive o make earnings-increasing reserving decisions, and will herefore under-reserve losses incurred in period. Analogously, opions exercised are by definiion in he money, and are herefore in he range where Healy (1985), Gaver e al. (1995), and Holhausen e al. (1995) would sugges earnings-increasing discreionary behavior is occurring. Moreover, hese generic opions are presumably sraigh call opions wih no upper bound (unlike bonuses), which provide only incenives for earnings-increasing discreionary behavior. Thus, we expec managers of insurers o make earnings-increasing reserve errors, i.e. under-reserve, when exercising hese call opions. This leads o he firs par of our anicipaion hypohesis and our hird esable hypohesis, H3, formally saed as: H3: Firms whose managers derive a larger porion of oal compensaion from he exercising of opions in period have more incenive o under-reserve losses incurred in period. 11

12 Resriced socks represen sock currenly owned by he execuives ha may be sold or held a heir discreion. The incenives of execuives o manipulae earnings will be deermined by heir holding period preferences for hese socks. Execuives who prefer o coninue holding he sock will be more inclined o underake earnings-decreasing decisions in an effor o shif posiive resuls o fuure periods when he sock will be sold. Conversely, execuives who prefer o liquidae in he shor-run are expeced o make earnings-increasing decisions in order o maximize he curren value of he firm, and herefore heir sock. Therefore, our wo compeing hypoheses regarding sock held are formally saed as: H4(a): Managers who hold sock, and anicipae liquidaing heir shares in he long-erm will have an incenive o make earnings-decreasing reserving decisions, and will herefore overreserve losses incurred in period. H4(b): Managers who hold sock, and anicipae liquidaing heir shares in he shor-erm will have an incenive o make earnings-increasing reserving decisions, and will herefore underreserve losses incurred in period. 3.2 Board Governance and Reserve Errors To examine how corporae governance mechanisms affec insurers reserve pracices, we focus on hree aspecs of corporae governance: he size of he board, he relaive number of independen direcors on he board, and wheher he CEO is also he chairman of he board (i.e. CEO dualiy). Lipon and Lorsch (1992) sugges ha boards of eigh or nine members are mos effecive. They argue ha he larger he board, he more difficul for all members o express 12

13 heir views in he limied ime available a he board meeings. Hence, smaller boards are more effecive in monioring. For our sudy, we hypohesize ha smaller boards are beer suied for monioring insurer execuives, because he insurance firms in our sample engage in complicaed underwriing business. Hence, he smaller he board size, he greaer is abiliy o monior, and he smaller he subsequen absolue reserve errors. However, his associaion may be eiher negaive or posiive. Tha is, for execuives who over-(under) reserve o maximize personal benefis, we expec independen direcors will consrain he over-(under) reserving pracice. Mayers, Shivdasan and Smih (1997) find ha muual insurers employ more ouside direcors han sock insurers. Since ownership of muual firms is non-ransferable, he exernal governance sysem, such as he hrea of a ake-over, is relaively ineffecive in miigaing agency problems. The choice of more independen direcors a leas parially combas he inadequacy of discipline from oher governance mechanisms. He and Sommer (2007) documen ha among muual insurance companies, hose wih greaer separaion of ownership and conrol use more ouside direcors. The increased monioring by ouside direcors can somewha offse he enrenchmen effecs resuling from he ownership and conrol separaion. Wih respec o board independence, numerous empirical sudies sugges ha independen boards provide effecive monioring. Hence, we predic a significan relaion beween he relaive number of ousiders on he board and reserve error. Again, he associaion could be eiher negaive or posiive. Exan lieraure suggess ha CEO dualiy resuls in higher agency coss. Hence, we predic ha when CEOs hold dual posiions, hey end o increase reserve errors, however, he direcion of he reserve depends on how hey derive heir compensaion. When over-reserving increases execuive payoffs, all else equal, we expec firms wih CEO dualiy will reserve even 13

14 more. On he oher hand, when under-reserving provides larger execuive benefis, all else equal, we expec firms wih CEO dualiy will furher under reserve. Our board governance hypohesis is saed as: H5: Firms wih srong boards (smaller boards, more ousiders on board, CEOs who do no hold he posiion as he chairman of he board) will have smaller reserve errors in absolue erms. To es H1, H2, H3, H4, and H5, we employ he following model: RE = β + β LNASSETS + β BONUS 4 + β BSIZE β BOUT + β RSTKA β NETINCOME + β RSTKH + β DUAL i + β STOPA 7 + ν + ε + β LONGTAIL 3 + β STOPX 8 (1) where i represens he individual insurers; represens he year (1992 o 2000); RE represens he reserve esimaion error associaed wih reserves repored for firm i in year ; LNASSETS is he naural log of firm asses; NETINCOME is he ne income of he firm; LONGTAIL is he proporion of business wrien in long-ailed lines; BONUS is he percen of oal execuive compensaion received in he form of a bonus; RSTKA is he percen of oal execuive compensaion received in resriced sock awarded; RSTKH is he raio of he value of resriced sock held o oal execuive compensaion; STOPA is he percen of oal execuive compensaion received in opions awarded; STOPX is he percen of oal execuive compensaion received in opions exercised; BSIZE is equal o he naural log of he number of direcors on he board; BOUT is he proporion of he board ha is made up of independen (nonexecuive) direcors; DUAL is a dummy variable ha is se o one if he CEO also holds he posiion as he chairman of he board and zero oherwise; ν i is he company specific error and ε is he random error. 4 4 The loss reserve esimaion error in year is he developed loss reserve in year n + for losses incurred in year and earlier less he original loss reserve esimae in year for losses incurred in year and earlier. In order o 14

15 A negaive (posiive) coefficien of an independen variable in (1) indicaes ha an increase in ha variable is associaed wih over-reserving (under-reserving). For example, if a long-erm incenive compensaion variable is found o have a negaive coefficien, i implies ha he larger he proporional compensaion in his long-erm componen, he smaller he difference beween developed loss reserves and iniial loss reserves. In fac, he more negaive his difference becomes, he larger he amoun of over-reserving. H1 predics he coefficiens of RSTKA and STOPA o be negaive. Finding any significance on he coefficiens of BONUS is consisen wih managers impacing reserving pracices in response o hese compensaion iems. More specifically, however, finding a negaive coefficien on BONUS is evidence in suppor of H2(a). Conversely, a posiive coefficien suppors he accepance of H2(b). H3 predics he coefficien of STOPX o be posiive. Tha is, he larger he value of opions exercised by managers relaive o heir oal compensaion, he larger he difference beween developed loss reserves and iniial loss reserves (i.e. incurred losses are iniially under-reserved). A negaive coefficien on RSTKH is evidence in suppor of H4(a) while a posiive coefficien on he coefficien of his variable suppors he accepance of H4(b). Again, any significance is indicaive of a relaionship beween he reserve errors and he sock held. In esing he corporae governance hypohesis, we expec he coefficiens on BSIZE, BOUT, DUAL o be significanly differen from zero, bu H5 does no predic a posiive or negaive sign for any of hese coefficiens. If he governance srucure exiss o miigae over-reserving (under-reserving) we would expec o find a posiive (negaive) coefficien. conrol for variaion in insurer size and o reflec he loss reserve esimaion error as a percenage, his difference is scaled by hree facors: 1) oal admied asses, 2) he five-year developed reserve, 3) 1.6 imes (he greaer of oal ne income and oal admied asses) 2/3. We uilize a five-year developmen period ( n = 5 ). This follows he loss reserve esimaion error mehodology uilized by Peroni (1992) and Gaver and Paerson (2004). 15

16 We furher proxy for insurer size (LNASSETS) and business mix (LONGTAIL) as prior research has generally shown hem o affec reserve esimaion error [business mix is significan in Aiuppa and Trieschmann (1987) and Peroni (1992); size is also significan in Aiuppa and Trieschmann, bu no Peroni or Weiss (1985)]. NETINCOME is used o proxy an insurer s ax liabiliy. Relaed lieraure suggess ha an insurer may manipulae loss reserves in an effor o minimize ax liabiliy hence, we would expec o see higher reserve errors for hose firms wih higher levels of ne income. Finally, according o Grace and Levery (2005) i is possible ha hese conrol variables used are no significanly relaed o reserve errors. Our paper can herefore be considered as a furher es of he recen resuls repored by Grace and Levery (2005). 3.3 The Join Effecs of Board Governance and Execuive Compensaion Alhough we expec o find a relaionship beween he reserve errors and he corporae governance variables, i is possible ha he relaionship beween reserve errors, execuive compensaion, and governance srucure is sufficienly complex, ha differing incenives muddle he observed relaionships. The exan lieraure on corporae governance generally finds ha firms wih more effecive board monioring have lower agency coss. Since he board of direcors has direc influence on execuive compensaion mechanisms, we raionalize ha he associaions beween various compensaion componens and insurers reserve errors should be weaker (sronger) for firms wih srong (weak) boards, i.e., srong boards miigae managers incenives o manipulae reserves whereas weak boards exacerbae hese pracices. This leads o our eighh hypohesis which ess how he ineracions beween board governance and execuive compensaion mechanisms joinly affec reserve errors, formally saed as: 16

17 H6: Firms wih effecive board srucures will have weaker associaions beween heir compensaion componens and reserve errors, whereas firms wih ineffecive board srucures will have sronger associaions beween heir compensaion componens and reserve errors. To es H6, we firs use he following model: RE = β + β LNASSETS + β BONUS + r BONUS 1 + r RSTKH 3 + r STOPX β RSTKA 5 + β NETINCOME + β RSTKH * WEAKDUMMY + r WEAKDUMMY + ν + ε * WEAKDUMMY + r RSTKA * WEAKDUMMY + r STOPA β LONGTAIL + β STOPA 7 3 * WEAKDUMMY * WEAKDUMMY i + β STOPX 8 (2) where variables from model (2) remain he same. The dummy variable WEAKDUMMY measures he srengh of he board and is ineraced wih he compensaion variables. Firms wih WEAKDUMMY equal o 1 are considered o have a weak board srucure. In consrucing WEAKDUMMY, we analyze he relaive size and independence of he insurer s board. We define a firm o have a board ha is oo large o mainain effecive oversigh if he board is larger han he median board size of our sample. Furher, we define a firm s board o be no independen if i has fewer han he median percen of ouside board members. Finally, we define a board o be no separaed if he CEO also holds he posiion of Chairman of he Board (i.e. if DUAL equals 1). If he firm mees wo or more of hese hree crieria ( oo large, no independen, or no separaed ) he firm is considered o have a weak board srucure, and WEAKDUMMY is se equal o one. We hypohesize ha he sum of ( β + r ) i will be differen from β for a leas one of he i. However, he sign on r is undeermined. If execuives over- 17

18 reserve, hen r will be negaive for firms wih weak boards. If execuives under-reserve, hen r will be posiive for firms wih weak boards. 4. Daa and Descripive Saisics Since Ke, Peron and Safieddine (1999) show ha more widely held insurance firms are more likely o base heir CEO compensaion packages on objecive accouning measures, we limi our sudy o publicly held insurance companies. Daa for our research come from he Form 10-K ha publicly raded companies file wih he U.S. Securiies and Exchange Commission, annual sauory saemens filed wih he Naional Associaion of Insurance Commissioners (NAIC), COMPUSTAT s Execuive Compensaion (ExecuComp) daa, and he Invesor Responsibiliy Research Cener, Inc. s (IRRC) daabase. The Form 10-K and ExecuComp provide execuive compensaion on insurers, he IRRC daabase provides our corporae governance variables, while oher NAIC financials provide necessary insurer accouning informaion such as reserves, premiums, admied asses and incurred losses. We analyze sock propery-liabiliy insurers beween 1992 and As such, our analysis can be considered a pre-es of sors for he effecs of SOX on insurers. Our iniial sample of 1,750 execuive-year compensaion observaions reduces o 218 insurer-year observaions for five-year loss reserve error esimaions for he following reasons. 6 Firs, ExecuComp repors individual annual compensaion figures for execuive officers while he NAIC repors firm specific as well as consolidaed informaion for insurers who are 5 We are limied o using daa years 2000 and earlier because of he mehodology employed in calculaing he reserve errors. Life insurers are excluded because hey do no repor he reserve deail required o calculae reserve errors. 6 Compensaion for one year for one execuive is defined o be one execuive-year compensaion observaion. Accouning informaion for one year for one insurer is defined o be one insurer-year observaion. 18

19 ofen comprised of muliple insurance companies. 7 Since execuive officers represen an enire insurance group (i.e. he publicly raded eniy) and he corporae governance variables are repored by he IRRC a he group level, we uilize consolidaed daa for each insurance group based on he aggregaion of individual insurers wihin each group. 8 Furher, we raionally assume he execuives of each insurance group as a whole were influenial in he reserving pracices of heir respecive group as hey ulimaely approve he repors a he group level. Therefore, we aggregae and average compensaion daa for he execuives of each insurance group whose compensaion daa is available. Second, our calculaion of five-year loss reserve esimaion errors for insurers does no allow us o use daa beyond he year As menioned, we follow mehodology consisen wih Peroni (1992) and Gaver and Paerson (2004). A five-year loss reserve esimaion error is he five-year developed reserve less he original reserve which was repored five years prior. We are able o deermine five-year loss reserve esimaion errors for he years 1992 hrough 2000 only. For example, he 1997 developed loss reserve daa are needed o calculae 1992 loss reserve errors and he 2005 developed loss reserve daa are needed o calculae 2000 loss reserve errors. Finally, we limi our iniial sample o publicly held propery-casualy insurance companies ha are domiciled in he Unied Saes. Life and healh insurers are excluded as managers of hese insurers have much less discreion in esablishing reserves because a larger porion of heir reserves are based upon well-esablished and well publicized acuarial ables (see 7 For example, in 1992 he Allsae Insurance Group consised of eleven individual insurance companies. The NAIC repors financial informaion for each of hese eleven companies as well as consolidaed informaion for he enire group. 8 Some insurer groups do no provide consolidaed daa for heir group o he NAIC and in some cases insurer groups repor muliple consolidaions for he same group. Hence, we creae consolidaed daa for each insurer group. Our resuling aggregaed insurer group daa is consisen wih hose insurers who do repor consolidaed financial informaion o he NAIC. 19

20 Peroni (1992)). Also, he reserve developmen for life insurers is no repored o he NAIC, making he observaion of loss reserve errors more difficul. ExecuComp provides daa on publicly held U.S. and Canadian firms; however Canadian insurers are excluded from our sample since foreign regulaion may effec heir behavior. Muual insurers are excluded from our sample, which also conribues o he relaively low number of insurer groups, for wo reasons. Muual insurers do no have sock or sock opions o offer heir execuives. Furher, incenive compensaion daa from muual insurers (and privaely held sock insurers) is self-repored o he NAIC, and herefore incomplee. The exclusion of muual insurers should no be of significan concern since heir execuive compensaion packages are no compleely comparable wih hose of sock insurers. 9 Moreover, Ke, Peron and Safieddine (1999) show ha more widely held insurance firms are more likely o base heir execuive compensaion packages on objecive accouning measures. The aggregaion of execuives compensaion daa, he uilizaion of consolidaed publicly held sock insurer daa and he daa limiaions creaed by our five-year loss reserve error calculaion mehodology combine o yield a sample of 218 insurer-year observaions based on 55 publicly raded propery-casualy insurers. 10 [Inser Table 1] 5. Hypoheses Tess and Resuls 9 Analysis of loss reserve esimaion error and execuive compensaion srucure may be done on boh muual and sock firms, bu in separae models, wih separae hypoheses. Here, we aemp o address half of his analysis (he sock half) and leave he analysis of muual insurers o fuure sudies ha conain more complee execuive compensaion daa for muual insurers as well as privaely held sock insurers. 10 The resuling sample reflecs a relaively large porion of he U.S. propery-casualy indusry, and reflecs an even larger relaive porion of propery-casualy sock insurers which is he only organizaional form of ineres o us. For example, in 1999 sock insurers accouned for abou 67% of he enire propery-casualy indusry s ne premiums wrien. Ne premiums wrien by our sample of insurers represen approximaely 59% of all sock insurers, or 39% of he enire indusry s ne premiums wrien in These percenages are consisen hroughou he years of our sample daa. 20

21 We use fixed effec models o accoun for unobservable cross-secional differences ha affec reserve error. Table 2 shows he resuls for equaion (1). The models were esimaed using he hree measures of reserve error discussed above. These resuls suppor he findings in Eckles and Halek (2007). As prediced, he coefficien on RSTKA is significan and negaive in all hree specificaions, supporing H1. We furher find suppor for H2(a) wih he negaive and significan coefficien on BONUS. This resul is consisen for wo specificaions of he reserve error. STOPX is posiive and significan in wo of he specificaions, supporing H3. We also find some suppor for H4(b), as he coefficien on RSTKH is posiive and significan in one of he specificaions. However, he coefficiens of all hree measures of board srucure, BSIZE, BOUT, and DUAL, are no significan, herefore, we do no have direc evidence supporing H5. This could be a resul of differing effecs of board srucure on reserve errors, as well as he ineracion of board srucure on execuive compensaion and reserve errors. We conduc empirical ess for hese predicions described in equaion (2) nex. [Inser Table 2] To es he ineracions beween he board srucure and compensaion variables, we again used a fixed effecs model o esimae equaion (2) specificaions. Table 3 shows he resuls of he fixed effecs specificaion and Table 4 repors he ess of he significance of he join execuive compensaion coefficiens. Alhough we could no separae incenives/relaionships beween corporae governance variables and reserve errors in our firs model above, here we find evidence of an ineracion componen beween reserve errors, execuive compensaion and corporae governance. In all hree of he reserve error specificaions, he execuive compensaion 21

22 componens found o be significan in model (1) (BONUS and RSTKA) remain significan. Table 5 suggess ha he use of BONUS schemes is no exacerbaed by he presence of a weak board srucure. However, he manipulaion of reserves because of resriced sock owned is exacerbaed by he presence of a weak board srucure. The coefficien on he ineracion erm beween STOPA and WEAKDUMMY and he coefficien of WEAKDUMMY are significan in several specificaions. Taken ogeher, hese resuls sugges ha a weaker board creaes more opporuniy for execuives o manipulae reserves in an effor o maximize heir compensaion. Finally, we find weak evidence ha suggess he effec of RSTKH and STOPX on reserve error depends on firms exising board srucure. [Inser Table 3] [Inser Table 4] 6. Conclusion This paper examines he relaionships beween he compensaion of insurance company execuives, he board srucure of insurance firms and he reserving pracices of insurance firms. The execuives incenives o manipulae accouning resuls in order o maximize heir personal benefis are well-documened in he earnings managemen lieraure. Our paper adds o his srand of lieraure in wo ways. Firs, mos of he earnings managemen sudies mus esimae he componen of he managed earnings since i is generally unobservable in accouning saemens. We use daa for insurance companies ha repor he developed reserve and acual reserve, he difference represening he acual accouning discreion exercised by insurers managers. Second, our sudy fuses he hree relaed lieraures, i.e., earnings managemen, 22

23 execuive compensaion, and he corporae governance, ogeher by invesigaing he ineracion beween managerial compensaion and board governance and how hey joinly affec mangers pracices in loss reserves. Our resuls sugges ha he presence of a weak board exacerbaes managers earnings managemen pracices. The misalignmen of ineress beween execuives and shareholders has been he focus of corporae governance lieraure. The idea of incenive based compensaion was iniially developed o provide proper incenives o managers o ac in he bes ineres of heir shareholders. However, his incenive based compensaion also encourages managers o use heir discreion over accouning pracices o maximize heir own uiliy. Relaive o indusrial firms, our sample firms operae in a heavily regulaed indusry. However, our resuls sugges ha insurers managers could sill exercise sizable discreion over accouning numbers and heir acions are encouraged by he incenive based compensaion, coupled wih he inadequae monioring from he board. We are in he process of exending his work o deermining he degree o which managers of muual (and non-publicly raded sock) insurers impac earnings o maximize heir compensaion. Analysis of a sample including deailed compensaion daa for boh muual and privaely held sock firms, focusing on bonus plans, would be an ideal exension o his work. Second, we would like o obain more specific informaion regarding he srucure of he bonus plans. This would allow more rigorous esing of our hypoheses regarding he impac of bonuses on he behavior of managers. 23

24 References Aiuppa, Thomas A. and James S. Trieschmann An Empirical Analysis of he Magniude and Accuracy of Incurred-Bu-No-Repored Reserves. Journal of Risk and Insurance, 54 (1), Beaver, William H., Maureen F. McNichols, and Karen K. Nelson Managemen of he Loss Reserve Accrual and he Disribuion of Earnings in he Propery-Casualy Insurance Indusry. Journal of Accouning and Economics, 35, Berger, Allen N., J. David Cummins, Mary Weiss, and Hongmin Zi Conglomeraion versus Sraegic Focus: Evidence from he Insurance Indusry. Journal of Financial Inermediaion, 9, Browne, Mark J., Yu-Luen Ma, and Ping Wang Execuive Sock Opion Compensaion and Reserve Errors in he Propery and Casualy Insurance Indusry. Working paper. Coles, Jeffrey L., Naveen D. Daniel, and Laliha Naveen Does One Size Fi All?, Journal of Financial Economics, forhcoming. Corne, Marcia Milon, Alan Marcus, and Hassan Tehranian Corporae Governance and Pay-for-Performance: The Impac of Earnings Managemen, Journal of Financial Economics, forhcoming. Eckles, David, and Marin Halek, 2007, Insurer Reserve Error and Execuive Compensaion, Working Paper. Gaver, Jennifer J., Kenneh M. Gaver, and Jeffrey R. Ausin Addiional Evidence on Bonus Plans and Income Managemen. Journal of Accouning and Economics, 19,

25 Gaver, Jennifer J. and Jeffrey S. Paerson Do Insurers Manipulae Loss Reserves o Mask Insolvency Problems? Journal of Accouning and Economics, 37, Grace, EV Propery-Liabiliy Insurer Reserve Errors: A Theoreical and Empirical Analysis. Journal of Risk and Insurance, 57, Grace, Marin F. and J. Tyler Levery Propery-Liabiliy Insurer Reserve Error Moive, Manipulaion, or Misake Working Paper. He, Enya and David W. Sommer Separaion of Ownership and Conrol: Implicaions for Board Composiion, Working paper. Healy, Paul M The Effec of Bonus Schemes on Accouning Decisions. Journal of Accouning and Economics, 7, Holhausen, Rober W., David F. Larcker, and Richard G. Sloan Annual Bonus Schemes and he Manipulaion of Earnings. Journal of Accouning and Economics, 19, Hoy, Rober E. and Kahleen A. McCullough Managerial Discreion and he Impac of Risk-Based Capial Requiremens on Propery-Liabiliy Insurer Reserving Pracices. Working paper. Ke, Bin, Kahy R. Peron and Assem Safieddine Ownership Concenraion and Sensiiviy of Execuive Pay o Accouning Performance Measures: Evidence from Publicly and Privaely-held Insurance Companies. Journal of Accouning and Economics, 28, Klein, A Audi commiee, board of direcor characerisics, and earnings managemen, Journal of Accouning Economics, 33,

26 Klein, April, 2006, Audi Commiee, Board of Direcor Characerisics, and Earnings Managemen, NYU, Law and Economics Research Paper No Linck, James S., Jeffry M. Neer, Tina Yang, The Deerminans of Board Srucure, Journal of Financial Economics, forhcoming. Lipon, M. and J. W. Lorsch, 1992, A modes proposal for improved corporae governance, Business Lawyer, 1, Mayers, David, Anil Shivdasan and Clifford W. Smih, Jr., 1997, Board Composiion and Corporae Conrol: Evidence from he Insurance Indusry, The Journal of Business, Vol. 70, No. 1, Peasnell, K.V., P.F. Pope and S.E. Young Board monioring and earnings managemen: do ouside direcors influence abnormal accruals? Journal of Business Finance and Accouning, 32(7-8), Peron Kahy R Opimisic Reporing in he Propery-Casualy Insurance Indusry. Journal of Accouning and Economics, 15, Peron Kahy and Mark Beasley Errors in Accouning Esimaes and Their Relaion o Audi Firm Type. Journal of Accouning Research, 34(1), Peron Kahy R. and Douglas A. Shackelford Managing Annual Accouning Repors o Avoid Sae Taxes: An Analysis of Propery-Casualy Insurers. Accouning Review, 74, 3: Ruque, Mark E. and Michael Ha New York Sues AIG, Greenberg for Improper Accouning Pracices. Naional Underwrier-Propery and Casualy, May 30,

27 Weiss, Mary A Mulivariae Analysis of Loss Reserving Esimaes in Propery- Liabiliy Insurers. Journal of Risk and Insurance, Xie, B. W. Davidson, and P. DaDal Earnings managemen and corporae governance: he role of he board and he audi commiee, Journal of Corporae Finance, 9,

28 Table 1 Descripive Saisics Mean Median Sd Dev Min Max Insurer Accouning Variables Five-Year Reserve Error Scaled by Admied Asses (000) (RE 1 ) Five-Year Reserve Error Scaled by Developed Reserve (RE 2 ) Five-Year Reserve Error Scaled by [1.6 x (he greaer of admied asses or ne premiums wrien) 2/3 ] (RE 3 ) Admied Asses (000,000,000) Ne Income (000,000) Concenraion in longer-ailed lines of business (LONGTAIL) Compensaion Variables Average Toal Compensaion (ATC) (000) Average bonus awarded as a percen of average oal compensaion (BONUS) Average value of resriced sock awarded as a percen of average oal compensaion (RSTKA) Average value of resriced sock held as a percen of average oal compensaion (RSTKH) Average value of sock opions awarded as a percen of average oal compensaion (STOPA) Average value of sock opions exercised as a percen of average oal compensaion (STOPX) Board Variables Board Size % Ouside Direcors (BOUT) Indicaor for CEO holding posiion of Chairman (DUAL) % of firms wih weak board (WEAKDUMMY) Noes: Obs. =218. The sample consiss of cross-secional daa from 1992 hrough

29 Dependen Variable Table 2 Resuls from Fixed Effecs Esimaion of Equaion (1) Five-Year Reserve Error (RE 1 ) Five-Year Reserve Error (RE 2 ) Five-Year Reserve Error (RE 3 ) Independen Variables Prediced Sign Coefficien Inercep BONUS +/ RSTKA RSTKH +/ STOPA STOPX BSIZE +/ BOUT +/ DUAL +/ LNASSETS +/ LONGTAIL +/ NETINCOME +/- 1.28E-11 ** ** ** * Sandard Error (p-value) Coefficien (0.380) (0.047) (0.012) (0.268) (0.419) (0.014) (0.234) (0.664) (0.730) (0.463) (0.093) E-11 (0.458) 1.26E-11 *** ** ** * ** Sandard Error (p-value) Coefficien (0.147) (0.002) (0.026) (0.016) (0.809) (0.362) (0.685) (0.174) (0.987) (0.074) (0.023) E-11 (0.766) 8.77E-12 * ** *** Sandard Error (p-value) (0.676) (0.227) (0.086) (0.403) (0.823) (0.025) (0.172) (0.561) (0.366) (0.225) (0.008) 3.42E-11 (0.798) F-Saisic 3.91 *** n/a 3.09 *** n/a 3.68 *** n/a R-Squared 18.30% n/a 21.07% n/a 19.54% n/a * Significan a 10% ** Significan a 5% ***Significan a 1% Noes: Obs. = 218 Whie s sandard errors RE 1 = Five-Year Reserve Error Scaled by Admied Asses RE 2 = Five-Year Reserve Error Scaled by Developed Reserve RE 3 = Five-Year Reserve Error Scaled by [1.6 x (he greaer of admied asses or ne premiums wrien) 2/3 ] 29

30 Dependen Variable Table 3 Resuls from Fixed Effecs Esimaion of Equaion (2) Five-Year Reserve Error (RE 1 ) Five-Year Reserve Error (RE 2 ) Five-Year Reserve Error (RE 3 ) Independen Variables Prediced Sign Coefficien Inercep BONUS +/ RSTKA RSTKH +/ STOPA STOPX BONUS*WEAKDUMMY +/ RSTKA*WEAKDUMMY RSTKH*WEAKDUMMY +/ STOPA*WEAKDUMMY STOPX*WEAKDUMMY WEAKDUMMY +/ LNASSETS +/ LONGTAIL +/ NETINCOME +/- 9.62E-12 * ** * * Sandard Error (p-value) Coefficien (0.438) (0.258) (0.627) (0.944) (0.676) (0.814) (0.057) (0.017) (0.221) (0.097) (0.714) (0.071) (0.682) (0.110) E-11 (0.521) 5.91E-12 * * * *** ** * ** Sandard Error (p-value) Coefficien (0.066) (0.089) (0.060) (0.003) (0.273) (0.930) (0.359) (0.441) (0.127) (0.121) (0.954) (0.037) (0.079) (0.015) E-11 (0.902) 9.45E-12 * *** ** Sandard Error (p-value) (0.662) (0.908) (0.486) (0.926) (0.240) (0.667) (0.088) (0.007) (0.318) (0.159) (0.565) (0.108) (0.181) (0.012) 3.42E-11 (0.783) F-Saisic 5.73 *** n/a 2.97 *** n/a 3.16 *** n/a R-Squared 24.48% n/a 23.52% n/a 21.91% n/a * Significan a 10% ** Significan a 5% ***Significan a 1% Noes: Obs. = 218 Whie s sandard errors RE 1 = Five-Year Reserve Error Scaled by Admied Asses RE 2 = Five-Year Reserve Error Scaled by Developed Reserve RE 3 = Five-Year Reserve Error Scaled by [1.6 x (he greaer of admied asses or ne premiums wrien) 2/3 ] 30