The effect of pension accounting on corporate pension asset allocation. Citation Review Of Accounting Studies, 2010, v. 15 n. 2, p.

Transcription

1 Ttle The effect of penson accountng on corporate penson asset allocaton Author(s) Amr, E; Guan, Y; Oswald, D Ctaton Revew Of Accountng Studes, 2010, v. 15 n. 2, p Issued Date 2010 URL Rghts The orgnal publcaton s avalable at Ths work s lcensed under a Creatve Commons Attrbuton- NonCommercal-NoDervatves 4.0 Internatonal Lcense.; The orgnal publcaton s avalable at

2 The Effect of Penson Accountng on Corporate Penson Asset Allocaton By El Amr* London Busness School Regent s Park London NW1 4SA eamr@london.edu Yanlng Guan School of Busness The Unversty of Hong Kong ylguan@busness.hku.hk Denns Oswald Ross School of Busness Unversty of Mchgan Ann Arbor, MI dennso@umch.edu August 2008 *Correspondng author. We thank Katherne Schpper (Edtor), Anonymous Revewer, Davd Blake, Elroy Dmson, John Pckles, Laurens Swnkels, Rchard Taffler and semnar partcpants at the 2008 CAR&E conference, INSEAD, Unversty of North Carolna, Tel Avv Unversty and Unversty of Warwck for many useful comments. We also thank London Busness School's research fundng for ths project. Address correspondence to El Amr, London Busness School, eamr@london.edu

3 The Effect of Penson Accountng on Corporate Penson Asset Allocaton Abstract We examne the mpact of new penson dsclosures and subsequent full penson recognton under FRS 17 and IAS 19 n the UK and SFAS 158 n the US on penson asset allocaton. These standards requre recognton of the total penson surplus/defct on the balance sheet and perodcal actuaral gans/losses n shareholders' equty. Therefore, these standards ntroduce a large element of volatlty nto company balance sheets and comprehensve ncome. We dentfy a Dsclosure perod durng whch UK companes had to dsclose all the requred data under FRS 17 n the notes wthout formal recognton. We also dentfy a Full Recognton perod startng one year pror to untl one year subsequent to the adopton of FRS 17/IAS 19 (UK) and SFAS 158 (US). We predct a shft of penson assets from equty to debt securtes by UK companes durng the Dsclosure perod due to the hgher vsblty of pensons n the UK and the antcpaton of full recognton. We also predct a declne n penson funds allocated to equty securtes durng the Full Recognton perod, around the adopton of FRS 17/IAS 19 and SFAS 158. We fnd that UK companes, on average, shfted penson assets from equty to debt securtes durng both the Dsclosure and the Full Recognton perods. We also fnd that whle pror to the adopton of SFAS 158 US companes mantaned a stable allocaton to equtes and bonds, these companes, on average, shfted funds from equtes to bonds around the adopton of SFAS 158. Cross-sectonal analyss shows that the shft away from equtes s related to changes n fundng levels, shorter nvestment horzons, ncreased fnancal leverage and the expected mpact of the new standards on shareholders' equty. Keywords: Defned Beneft Plans, Penson Asset Allocaton, Penson Surplus/ Defct, FRS 17, IAS 19, SFAS 158.

4 1. Introducton We nvestgate the effect of new penson dsclosures and subsequent recognton requrements of penson surplus/defct on the allocaton of penson funds to equty and debt securtes. In November 2000, the Accountng Standards Board (ASB) n the UK ssued Fnancal Reportng Standard (FRS) No. 17, Retrement Benefts (ASB, 2000), whch ntally had to be adopted n June In 2002, ASB extended the transtonal perod of FRS 17 to fscal years startng on or after January 2005, whch concded wth the adopton of Internatonal Fnancal Reportng Standards (IFRS) n Europe. Effectvely, UK companes had to adopt the revsed Internatonal Accountng Standard (IAS) No. 19 (IASB, 2004), whch was vrtually dentcal to FRS 17. In September 2006, the Fnancal Accountng Standard Board (FASB) ssued SFAS 158 (FASB, 2006), whch replaced SFAS 87 (FASB, 1985) and became effectve for all US companes n December Between fscal 2000 and the adopton of FRS 17/IAS 19, UK companes had to provde new and detaled dsclosure on the status of ther penson plans. Adopton of FRS 17/IAS 19 mples that UK companes recognze the total penson surplus/defct on the balance sheet (net of deferred tax). Any resultng actuaral gans or losses arsng durng the year are recognzed mmedately n shareholders equty. Smlarly, SFAS 158 requres full recognton of the penson surplus/defct on the balance sheet and mmedate recognton of all actuaral gans/losses and pror servce cost through other comprehensve ncome. The recognton of net penson surplus/defct as an asset/lablty on the balance sheet and actuaral gans/losses n shareholders equty under FRS 17/IAS 19 and SFAS 158 ntroduce materal volatlty to balance sheets of UK and US companes, especally f penson assets are mostly nvested n equty securtes. In partcular, reportng actual, rather than smoothed, penson returns njects volatlty nto shareholders equty whle the recognzed net penson asset/lablty could be a sgnfcant porton of a company s book 1

5 value and market captalzaton. 1 A cost-effectve way of reducng the volatlty of the penson defct/surplus and the volatlty of comprehensve ncome and shareholders equty s matchng penson assets wth penson labltes. Ths could be acheved by selectng a portfolo of penson assets whose far value s more postvely correlated wth the far value of the penson lablty, namely bonds. 2 Alternatvely, companes could react to the new accountng rules by termnatng penson plans (Klumpes and Whttngton, 2003) or by convertng defned beneft penson plans nto defned contrbuton and cash balance plans (D Souza et al., 2004; Swnkels, 2006). However, these actons may be costly due to tax and labor negotaton costs. We examne the mpact of new penson dsclosures and full penson recognton on asset allocaton of UK and US companes. In partcular, we examne changes n UK companes penson asset allocaton before and after the adopton of FRS 17/IAS 19 ( ). We dentfy two dstnct accountng regmes for UK companes: A Dsclosure perod rangng from fscal 2001 untl one year pror to the adopton of FRS 17/IAS 19, and a Full Recognton regme, startng from one year pror to adopton untl one year after the adopton of FRS 17/IAS 19. We predct that UK companes that sponsor defned beneft penson plans wll shft penson assets from equty to debt securtes durng the Dsclosure perod. Ths s because of the ncreased vsblty of penson plans due to market-based dsclosures and the antcpaton of the effect of full penson recognton on the volatlty of shareholders equty and comprehensve ncome. We also predct a declne n penson assets allocated to equty securtes durng the Full Recognton perod due to the recognton of penson surplus/defct on the balance sheet and the hgher antcpated volatlty of shareholders equty and 1 The volatlty that can be ntroduced nto corporate balance sheets s evdenced by the fact that at the end of 2001 the combned surplus for the FTSE 100 was 5 bllon, but by md July 2002 wth the collapse n world stock markets ths fell to a defct of 25 bllon (Reynolds, 2002). 2 A report by the Commttee on Investment of Employee Beneft Assets has estmated that swtchng to markto-market penson accountng would result n approxmately $290bn n funds beng shfted from equtes to bonds (Brewstern, 2005). 2

6 comprehensve ncome. We also examne whether full recognton penson accountng under SFAS 158 prompted US companes to shft penson assets from equtes to bonds. As US companes had to adopt Full Recognton accountng under SFAS 158 n 2006, we predct a declne n the allocaton to equty securtes by US companes from one year pror to the adopton of SFAS 158 (fscal 2005) untl one year after the adopton of the standard (fscal 2007). The new penson accountng standards n the UK and the US may not be the only reason for the change n the composton of penson assets. Contemporaneous changes n fundng levels due to mposed mnmum fundng requrements, shorter nvestment horzons, changes n plan coverage and ncreased fnancal leverage could trgger smlar asset allocaton changes. Multvarate cross-sectonal analyss reveals that whle other factors have sgnfcant effects on penson asset allocaton, the expected mpact of the new standards s ncrementally nformatve n explanng the shft to debt securtes n UK and US companes; that s, the shft to debt securtes s more sgnfcant n companes wth larger penson schemes relatve to shareholders' equty. To test our hypotheses, we use penson asset allocaton data for large UK and US companes over the perod Usng a sample of 1,509 and 2,128 frm-year observatons for UK and US companes, respectvely, we fnd that durng the FRS 17 Dsclosure perod, UK companes, on average, shfted funds from equty to debt securtes. We also fnd that both UK and US companes shfted penson assets from equtes to bonds durng the Full Recognton perod of FRS 17/IAS 19 and SFAS 158, respectvely. Ths shft reduces the effect of the new standards on the volatlty of shareholders equty and comprehensve ncome. In addton, cross sectonal tests show that UK and US companes for whch the negatve mpact of full recognton was expected to be stronger shfted relatvely more funds from equty to debt securtes. Collectvely, the results support the 3

7 argument that new penson accountng standards n the UK and the US had a sgnfcant mpact on corporate penson asset allocaton, ncrementally to other economc factors. Ths study contrbutes to the lterature n several ways. It examnes the determnants of penson asset allocaton n an nternatonal settng usng both UK and US companes. In addton, t s conducted durng a perod of a penson accountng change n both the UK and the US, whch s a powerful settng for our study. Furthermore, the emprcal evdence n ths study s mportant n understandng the possble effects of full recognton penson accountng on captal markets. The remander of ths study s organzed as follows. In secton 2, we develop the hypotheses to be tested. Secton 3 descrbes the research desgn. Sample selecton and descrptve statstcs are ncluded n Secton 4. Secton 5 provdes the emprcal results whle Secton 6 provdes concludng remarks. 2. Hypotheses Followng the ssuance of FRS 17 n 2000, UK companes were requred to dsclose the present value of the penson oblgaton and the market value of penson plan assets, as well as actuaral assumptons and detals on asset allocaton. These new dsclosures meant greater transparency of the penson scheme as well as ncreasng nvestor scrutny of the mpact of penson schemes on fnancal statements. Whle FRS 17 dsclosures provde comfort to nvestors n over-funded plans, under-funded companes are perceved as rsker because penson defcts are a form of debt. 3 Also, the exstence of Mnmum Fundng Requrements ncreases the lkelhood of addtonal penson contrbutons, especally n under-funded plans. Hgher future penson contrbutons and the uncertanty assocated wth ther tmng and magntude also ncreases the perceved rsk of sponsorng companes. Furthermore, 3 See Harrs et al. (2001) for a dscusson of the effects of FRS 17 on fnancal statements and valuaton. 4

8 under FRS 17, employees have access to more accurate penson dsclosures, whch could trgger pressure to reduce plan rsk, as the employees bear the defct, whle the company enjoys the surplus. Thus, new penson dsclosure requrements are potentally costly and companes would prefer to hedge themselves aganst potental penson defcts by reducng the volatlty of the penson surplus/defct and mprovng the matchng of penson assets and labltes (Blake, 2001). Better matchng s acheved by allocatng more penson assets to debt (.e., bonds and other nterest bearng securtes) nstead of equty securtes. As the market value of bonds s generally less volatle than that of equtes, and snce penson labltes largely depend on prevalng yeld on bonds, such a polcy would result n lower volatlty of the penson defct/surplus. In addton, a shft from equtes to bonds s lkely to be exacerbated by the antcpated adopton of FRS 17 and the recognton of the full penson lablty on subsequent balance sheets. In contrast, shftng penson assets to bonds s lkely to result n hgher future penson contrbutons, as the expected rate of return on bonds s lower than that on equtes. Thus, the shft from equty to debt securtes wll contnue as long as the cost of reportng a penson defct s hgher than the present value of addtonal future contrbutons. Ths leads to our frst hypothess: Hypothess 1: UK companes that sponsor defned beneft penson plans wll shft penson assets from equty to debt securtes durng the FRS 17 Dsclosure perod. Adopton of FRS 17/IAS 19 n the UK and SFAS 158 n the US meant recognton of the entre penson defct/surplus on the balance sheet, recognton of all pror servce costs n net ncome and recognton of all actuaral gans/losses n comprehensve ncome. Therefore, n addton to ncreasng, on average, the amount of debt on the balance sheet, adopton of 5

9 the new standards n both countres s expected to ncrease the volatlty of shareholders equty and comprehensve ncome. Adopton of full penson recognton could have sgnfcant contractual mplcatons. Frst, as many contracts are based on verfable balance sheet fgures, hgher recognzed debt on the balance sheet ncreases the lkelhood of volatng exstng debt covenants, hence the cost of debt renegotaton. Second, hgher volatlty of shareholders equty ncreases the probablty of volatng equty-based debt covenant because the balance sheet s now exposed to market volatlty through nterest rate changes and changes n market values of equtes. Thrd, as the penson defct s recognzed aganst the company s dstrbuton reserves, adopton of the new standards may decrease dstrbutable reserves and may have a negatve effect on the company s ablty to pay dvdends or to mantan a stable stream of dvdends. Adopton could also have a negatve effect on stakeholders ablty to evaluate management performance. For example, Return on Equty (ROE), a wdely used management performance measure, may be dstorted as shareholders' equty becomes more volatle. If stakeholders consder comprehensve ncome n the numerator of ROE, then the volatlty of ROE s even hgher. In addton, hgher volatlty of shareholders equty and comprehensve ncome could create a percepton of weaker management control. To mtgate the effect of adopton on exstng contracts and to facltate better performance evaluaton, we expect companes to shft penson assets from equty to debt securtes durng the adopton of full penson recognton. Ths argument leads to our second hypothess: Hypothess 2: UK and US companes that sponsor defned beneft penson plans wll shft penson assets from equty to debt securtes durng the Full Recognton perod (adopton of FRS 17/IAS 19 n the UK and SFAS 158 n the US). 6

10 The mpact of the new penson standards s expected to be more (less) sgnfcant n companes wth relatvely larger (smaller) penson plans. For example, n ts 2003 annual report, Charter Plc, a UK-based engneerng company, reported penson assets wth a market value of mllon, whle ts shareholders equty amounted to 24.9 mllon. Ths means that a declne of 5.4% n the market value of penson assets could elmnate the company's entre shareholders equty restrctng ts dvdend payout ablty. Smlarly, a declne n the yeld on the AA-rated corporate bonds could elmnate shareholders equty, as the penson oblgaton ncreases when the dscount rate declnes. Thus, we expect companes wth larger penson schemes relatve to shareholders equty to shft more assets from equty to debt securtes durng the Dsclosure and Full Recognton perods. 4 Companes wth larger penson schemes also experence larger actuaral gans/losses, especally f a majorty of penson assets are nvested n equty securtes. These companes are more senstve to the volatlty effects of the new standards and would therefore have stronger motvaton to shft penson assets from equty to debt securtes. In contrast, companes wth smaller actuaral gans/losses would have a weaker motvaton to swtch penson funds to debt securtes, snce shftng to less rsky assets also mples a lower expected return on penson assets on the ncome statement (Fernandez, 2002). Consequently, we hypothesze that the shft from equty to debt securtes wll be more sgnfcant for companes wth a larger rato of penson assets/labltes to shareholders equty. We expect ths relaton for UK companes durng both the Dsclosure and Full Recognton perods and for US companes durng the Full Recognton perod. 4 In 2001, Boots plc. lqudated all of ts equty holdngs n ts 2.3 bllon penson fund and moved the proceeds nto long-dated bonds (Ralfe, 2002; McLesh, 2001). Smlarly, General Motors reduced the share of equtes n the penson portfolo from 47% n 12/2005 to 38% n 12/2006 and to 26% n 12/

11 Hypothess 3a: The shft of penson assets from equty to debt securtes by UK companes durng the Dsclosure perod s postvely correlated wth the relatve sze of the penson scheme. Hypothess 3b: The shft of penson assets from equty to debt securtes by UK and US companes durng the Full Recognton perod s postvely correlated wth the relatve sze of the penson scheme. Whle ths study hghlghts the asset allocaton effects of new penson accountng standards, other factors may have materal allocatonal effects durng the Dsclosure perod n the UK and the Full Recognton perod n the UK and the US. Pror lterature dentfes three major factors that affect penson asset allocaton: Fundng levels, nvestment horzon and offsettng frm rsk. 5 Both UK and US companes are subject to Mnmum Fundng Requrements (MFR). For example, n August 2006, the Penson Reform Act was sgned nto law n the Unted States. Ths Act requres defned beneft penson plans to attan full fundng status wthn seven years by contrbutng funds to the plans. These requrements are expected to ncrease fundng levels over tme. Pror lterature s rather mxed as to the effect of fundng levels on asset allocaton. Stll, Harrson and Sharpe (1983) suggest that as penson plans become less underfunded, the allocaton to equtes should declne. Thus, ncreasng fundng levels could trgger a declne n the allocaton to equtes ndependent of new accountng standards. Investment maturty could also have a sgnfcant effect on asset allocaton. As penson plans become more mature, the nvestment horzon s shortened, whch s expected to trgger a shft of penson assets from equtes to bonds. Fnally, pror studes documented a negatve 5 Black (1980), Feldsten and Selgman (1981), Tepper (1981) and Harrson and Sharpe (1983), among others. 8

12 relaton between frm rsk and the allocaton to equtes, as companes offset frm rsk by usng a more conservatve penson asset allocaton polcy. Therefore, a change n penson asset allocaton could be assocated wth changes n frm rsk and not necessarly related to new accountng standards. Although our man emprcal focus s on the mpact of the new accountng standards, we do control for these alternatve economc explanatons n our emprcal tests. 3. Emprcal Desgn Penson assets are classfed n the notes to the fnancal statements nto three man categores: stocks, bonds, and other. Stocks and bonds together account for about 90% of total penson funds n our UK and US samples. Other often ncludes assets such as mortgage-backed securtes, venture captal, prvate placement, propertes, etc. To test hypotheses 1 and 2, we analyze penson asset composton over the perod for a sample of UK and US companes. Specfcally, we test whether UK companes have shfted penson funds from equtes to bonds durng the Dsclosure perod and the Full Recognton perod, and whether US companes shfted funds from equtes to bonds durng the Full Recognton perods. We also examne whether fundng levels, nvestment horzon and frm rsk changed systematcally over the sample perod. To test hypotheses 3a and 3b, we construct a model that explans the cross-sectonal varaton n the percentage of penson funds allocated to equty securtes (requity t ): requity + β DIVP 6 t t = β + β IMPACT 0 + β TAXR 7 1 t t + β SDCF 8 + β FUND t 2 t + β SIZE 9 + β FUND t 3 2 t + β CLOSE 10 + β HOR t 4 + ε t t + β LEV 5 t (1) The dependent varable n Equaton (1), requity t, s the market value of penson assets allocated to equty securtes dvded by the market value of total penson assets for 9

13 frm n year t. The frst ndependent varable n Equaton (1) s IMPACT t, whch measures the potental mpact of the new accountng standards on company n year t. As the mpact relates prmarly to the relatve sze of the penson plan, we use two smple measures of the sze of the penson plan relatve to shareholders' equty: () EXPOS1 t : The far value of penson assets deflated by book value of shareholders equty. Ths varable captures the company s exposure to the volatlty n the market value of penson assets and net penson surplus/defct. () EXPOS2 t : The Projected Beneft Oblgaton (PBO) deflated by book value of shareholders equty. Ths varable captures the company s exposure to the volatlty n dscount rates. To compute EXPOS1 t and EXPOS2 t, we use shareholders equty before the effect of actuaral gans/losses and the recognton of net penson surplus/defct. Tax and regulatory factors play a sgnfcant role n asset allocaton through fundng levels. In general, the tax-deductblty of penson contrbutons should nduce companes to pre-fund ther penson plans; companes that are subject to hgher tax rates should have even greater ncentves to pre-fund ther plans. Black (1980) and Tepper (1981) argue that snce returns on penson assets are not taxed, these assets should be nvested n the most heavly taxed securtes, presumably bonds. Ther argument suggests no assocaton between fundng levels and asset allocaton as all companes nvest n bonds regardless of fundng levels. In contrast, Harrson and Sharpe (1983) argue that the exstence of the Penson Beneft Guaranty Corporaton (PBGC) provdes US companes wth a put opton on extremely under-funded penson oblgaton. Together wth lmted tax deductblty n the case of extremely over-funded plans, they argue that fundng and asset-allocaton decsons are jont and extreme. To maxmze tax benefts on one hand and the value of the PBGC opton on the other hand, companes should ether over-fund the penson plan and allocate all the assets to 10

14 bonds, or under-fund and allocate all the assets to equtes. Although n practce fundng/asset-allocaton decsons are rarely extreme, ths argument supports a negatve relaton between fundng levels and allocatons to equtes. A serous caveat s that the value of the PBGC put opton n the US has declned over tme and n partcular snce the 1986 Tax Reform Act. Also, an nsurance company such as the PBGC dd not exst n the UK untl Therefore, the ncentve to allocate penson assets to equtes n cases of extreme under-fundng may be of a second order nature. Bader (1991) argues that companes strve to mnmze the volatlty of future penson contrbutons. These contrbutons are farly predctable for moderate fundng levels, but less predctable for more extreme levels. To reduce the volatlty of penson contrbutons, he argues that extremely over-funded and under-funded plans should nvest n bonds, whle only moderately funded plans should ncrease allocaton to equtes. Hs argument suggests an nverted U-shape relaton between fundng levels and the allocaton to equtes. Based on these arguments, we nclude both FUND t and FUND 2 t to accommodate the possblty of a non-lnear relaton between fundng levels and allocaton to equtes. 6 The fundng status s measured as the far value of penson assets dvded by the accumulated beneft oblgaton (ABO). 7 Investment horzon plays a sgnfcant role n penson asset allocaton. Whle penson oblgatons to retrees are relatvely short-term and are prmarly affected by nterest rates, oblgatons to actve employees are relatvely long-term and are prmarly affected by salary ncreases. As bonds are more correlated wth nterest rate changes and stocks are more hghly correlated wth salary ncreases, companes wth relatvely young (mature) workforce should nvest more n stocks (bonds). Consequently, we expect a postve correlaton 6 Amr and Benartz (1999) documented an nverted-u relaton between fundng status and the percentage nvested n equtes. 11

15 between nvestment horzon, HOR t, and allocaton to equtes. HOR t, s measured as the natural logarthm of the rato of PBO to current servce cost. Overall, an older (younger) workforce should lead to a smaller (larger) rato of PBO to servce cost, ndcatng a shorter (longer) nvestment horzon. Our horzon varable (HOR t ) may be subject to measurement error. To mtgate ths problem, we nclude another varable that s assocated wth nvestment horzon. Ths varable, CLOSE t, takes nto account recent trends to move away from defned beneft to defned contrbuton penson plans and close the defned beneft plans to new entrants (McSherry, 2006). Thus, we nclude n our model an ndcator varable that s equal to one f the prncpal defned beneft plan s closed to new entrants, and zero otherwse. As the nvestment horzon of closed plans s, on average, shorter, we expect a negatve relaton between ths varable and allocatons to equtes. Our model also ncludes varables that capture the nfluence of debt contracts and dvdend payout polcy. As the penson surplus/defct s recognzed on the balance sheet, the penson asste/lablty and correspondng nvestment portfolo may be affected by certan contractual arrangements. In partcular, companes that are closer to the volaton of debt covenants have stronger motves to mprove asset/lablty matchng n order to reduce recognzed penson defcts. Better asset/lablty matchng would also reduce the volatlty of shareholders' equty and future penson contrbutons, whch n turn would reduce the volatlty of dvdends. We expect companes wth tghter debt covenants and hgher dvdends payout ratos to allocate more to bonds. To capture these effects, we nclude fnancal leverage (LEV t ) and dvdends payout rato (DIVP t ). We expect the coeffcents on these varables to be negatvely assocated wth the amount nvested n equty. Fnancal 7 Snce ABO s not avalable for UK companes, we calculate ABO based on the formula proposed n Amr and Benartz (1999): ABO = PBO / (1+G) N, where G s the assumed projected salary ncrease and N s penson fund s nvestment horzon. 12

16 leverage (LEV t ) s, measured as long term debt dvded by the sum of long term debt and market value of equty for frm n year t. Dvdend payout rato (DIVP t ) s measured as dvdends per share dvded by retaned earnngs per share. If retaned earnngs are negatve, then the varable s measured as the average dvdends per share over the current and past two years dvded by average retaned earnngs per share over the current and past two years. Companes subject to hgher tax rates should have greater ncentves to allocate more penson assets to bonds, as bonds are more heavly taxed. Therefore, we nclude the company's effectve tax rate (TAXR t ), measured as total tax expense dvded by pre-tax ncome n year t. If current pre-tax ncome s negatve, we use the average tax expense over the current and past two years dvded by the average pre-tax ncome over the current and past two years. Fredman (1983) and Bode et al. (1984) fnd that companes offset hgh corporate rsk by nvestng more of the penson assets n bonds. Ths polcy of offsettng rsk through the penson fund may reflect management preference to avod makng contrbutons to the penson fund when operatng cash flows are low. 8 Consstent wth these fndngs, we expect a negatve correlaton between the varablty of operatng cash flows and equty allocaton. In addton, companes wth more dversfed operatons would prefer to assume more rsk n ther penson fund. To the extent that larger frms are more dversfed, we would expect a postve assocaton between frm sze and allocaton to equtes. The volatlty of operatng cash flows (SDCF t ) s measured as the standard devaton of operatng cash flows over the current and past four years, deflated by book value of common equty. Frm sze (SIZE t ) s measured as the natural logarthm of market value of equty. To drectly test our hypotheses, we defne two dependent varables: The frst one 8 Specfcally, f operatng cash flows are volatle and the penson assets are nvested n equtes, the plan s lkely to become under-funded when operatng cash flows are low. As a result, the company would have to make large contrbutons to the penson fund n tmes of low operatng cash flows. 13

17 computes the change n the percentage of assets allocated to equtes durng the Dsclosure perod of UK companes. The Dsclosure perod s defned as the perod between fscal 2001 and the year before the adopton of FRS 17/IAS 19: rdisclose = requity (year 2001) - requity (Pre-Adopton year). 9 The second dependent varable computes the change n percentage of assets allocated to equtes durng the Full Recognton perod. For UK (US) companes, ths perod begns one year pror to adopton of FRS 17/IAS 19 (SFAS 158) and ends one year subsequent to adopton: radopt = requity (Pre-Adopton year) - requity (Post-Adopton year). We estmate Equaton (2) for UK companes only. The ndependent varables are measured as the dfference between the level of the varable at the begnnng of the Dsclosure perod and ts level at the end of the perod. rdisclose + β ΔLEV 5 = β + β ΔIMPACT + β ΔDIVP + β ΔTAXR β ΔFUND 2 + β ΔSDCF 8 + β ΔFUND 3 + β ΔSIZE β ΔHOR + β ΔCLOSE ε t (2) In addton, we estmate Equaton (3) for both UK and US companes around the Full Recognton perod. To control for long-term asset allocaton trends, we added to Equaton (3) another control varable, rdisclose, whch captures asset allocaton changes pror to adopton. Controllng for pror allocatonal effects could allow us to draw more accurate nferences on the ncremental effect of accountng standards on penson asset allocaton. 9 Techncally, the Dsclosure perod begns n year However, we use 2001 as the startng year because many observatons for year 2000 are mssng. 14

18 radopt + δ ΔLEV 5 + δ rdisclose 11 = δ + δ ΔIMPACT 0 + δ ΔDIVP + δ ΔTAXR η 7 + δ ΔFUND 2 + δ ΔFUND 2 + δ ΔSDCF + δ ΔSIZE δ ΔHOR 4 + δ ΔCLOSE 10 (3) The ndependent varables n Equaton (3) are measured as the dfference between the level of the varable before and after adopton. We use ordnary least square (OLS) to estmate the above equatons. Petersen (2008) argues that resduals may be correlated across frms or across tme n panel data and therefore OLS standard errors may be based. In ths study, t s possble that the unspecfed determnants of the dependent varables are correlated both over tme and across frms. For example, f the equty market as a whole s volatle, then all frms may shft away from volatle equtes to less volatle bonds. On the other hand, the strategy of penson asset allocaton could also be drven by some unspecfed frm-specfc factors, whch wll gve rse to a frm effect n the error terms. The presence of postve correlaton among error terms results n underestmated standard errors and thus nflated t-statstcs. To address such econometrc concern, we employ two-dmenson clusterng suggested by Petersen (2008) to accommodate the possblty of both tme effect and frm effect n such panel data. 4. Sample Selecton and Descrptve Statstcs The ntal UK sample contans 250 companes that belong to the FTSE 350 lst and that sponsor defned beneft penson plans durng We deleted from the UK sample seven companes that elected the corrdor method allowed under IAS 19. Informaton on market value of penson assets, actuaral present value of penson labltes, penson actuaral assumptons, actuaral gans/losses and detals of penson asset allocaton 15

19 are collected from annual fnancal statements. All other fnancal data for UK companes are from Datastream. Data for US companes' penson asset allocaton untl 2004 are collected from Pensons and Investments, a perodcal survey that covers the largest 1,000 penson funds n the US. Of ths 1,000, approxmately 300 penson funds relate to defned beneft plans for publcly traded frms (the remander are sponsored by prvate frms, unons, or government enttes, or are foregn companes lsted n the US). Asset allocaton data for are collected from notes to the annual fnancal statements. Fnancal data for US companes are from Compustat. After removng observatons wth mssng data, the sample conssts of 4,440 frm-year observatons, of whch 1,509 observatons are for UK companes and 2,128 observatons are for US companes. Table 1 provdes descrptve statstcs for the UK and US sub-samples. Over the entre sample perod, both UK and US companes allocate, on average, 62% of ther penson assets to equtes. The sze of the penson plan relatve to shareholders' equty s larger n UK than n US frms, as reflected by hgher means and medans of EXPOS1 and EXPOS2 (statstcal tests not reported). Ths dfference could lead to more sgnfcant allocatonal effects n the UK than n the US. As for other varables, the UK and US sub-samples are smlar to each other n terms of fundng status (FUND), penson horzon (HOR) and effectve tax rates (TAXR). US companes are less rsky as reflected by lower volatlty of cash flows (SDCF) and larger frm sze (SIZE). US companes also have lower dvdend payout ratos, for the medan frm, and are more hghly leveraged than UK companes. 10 (Table 1 about here) Most UK companes adopted IAS 19 n 2005, the mandatory year of adopton. However, 54 UK companes n our sample adopted FRS 17 pror to the mandatory year. 16

20 Pror lterature suggests that the tmng of adopton of new accountng studes s largely affected by the mpact of the new standard on the fnancal statements and by contractng costs (e.g., Amr and Zv, 1997). Usng logstc regressons to dentfy the characterstcs of UK companes that elected early adopton of FRS 17, we fnd (results not tabulated) that early adopters have, on average, smaller (at the 0.05 level) ratos of penson assets to shareholders' equty (EXPOS1). In addton, we fnd that early adopters have longer (at the 0.05 level) nvestment horzons (HOR). All other varables that were examned (requity, FUND, LEV, SIZE, CLOSE) were not sgnfcant (at the 0.10 level) n explanng the adopton decson. These results are consstent wth the clam that UK companes that were less affected by FRS 17 were more lkely to adopt the standard earler. Table 2 provdes data on the composton of penson assets for UK companes (top panel), a sub-sample of 54 UK early adopters of FRS 17 (mddle panel) and US companes (bottom panel) over UK companes gradually decreased ther average allocaton to equtes by 19.8% and ncreased ther allocaton to bonds by 12.7%. UK early adopters of FRS 17 exhbt a smlar pattern untl 2003; however, the declne n the allocaton to equtes s sharper n Durng , US companes had a relatvely stable allocaton to equtes and bonds, however, we observe a declne n the allocaton to equtes n the perod (Table 2 about here) Fgure 1 depcts the mean percentage of penson assets allocated to equty and debt securtes for UK and US companes (Fgure 1a) and for UK early adopters of FRS 17 (Fgure 1b). Fgure 1a shows a gradual decrease (ncrease) n the allocaton to equtes 10 All contnuous explanatory varables are wnsorzed at 1% and 99% to mtgate the effect of extreme observatons, except TAXR t and DIVP t whch are wnsorzed at 5% and 95% to remove negatve values. 11 We also examned changes n asset allocaton usng market values of stocks and bond that exsted n Ths way, we attempt to remove the effects of market value changes on the composton of penson assets and solate the effect of corporate rebalancng. The declne n the allocaton to stocks and the ncrease n the allocaton to bonds are more transparent n both sub-samples. 17

21 (bonds) n UK companes durng the sample perod. The average allocaton n US companes, on the other hand, s not as monotonc. Untl 2003, US companes mantaned a relatvely stable allocaton to equtes. Between 2003 and 2005, average allocaton to equtes ncreased due to hgher stock prces. From 2005 to 2007 we observe a sgnfcant decrease n the allocaton to equty n US companes. Fgure 1b, whch focuses on early adopters of FRS 17, shows a gradual decrease (ncrease) n the allocaton to equtes (bonds). However, we observe a sharp declne n the allocaton to equtes n 2001, the year FRS 17 was released. Overall, early adopters exhbt a more conservatve asset allocaton relatve to other UK companes. (Fgure 1 about here) Fgure 2 depcts the trend of penson asset allocaton around the adopton of FRS 17/ IAS 19 n the UK (Fgure 2a), SFAS 158 n the US (Fgure 2b) and early adopters of FRS 17 (Fgure 2c). For each company, we dentfy the adopton year and denote t as AY(0). Then, we plot mean percentage of penson assets allocated to equtes and bonds around the adopton year. Fgure 2a presents nformaton for the entre UK sample between AY(-3) and AY(+2). The fgure exhbts a declne n penson funds nvested n equtes, and an upward trend n the percentage nvested n bonds. These changes are more transparent n years AY(+1) and AY(+2), consstent wth our argument that the new accountng standard caused UK companes to transfer penson assets from equtes to bonds. Fgure 2b presents nformaton for US companes between AY(-3) and AY(+1). 12 The fgure also exhbts a declne n penson funds nvested n equtes and an upward trend n the percentage nvested n bonds subsequent to adopton of SFAS 158. However, these changes are not as sharp as n the UK. Fgure 2c present movements n asset allocaton for UK early 12 AY(+2) s fscal 2008 and thus unavalable at ths tme. 18

22 adopters. As n Fgure 2a, we observe sharp declnes n the allocaton to equtes around the adopton of the new standard. (Fgure 2 about here) 5. Emprcal Results Table 3, Panel A, provdes results for testng whether UK companes changed ther penson asset allocaton durng the Dsclosure perod of FRS 17 (Hypothess 1). The results show that UK companes ncreased (decreased) ther allocaton to bonds (equtes) durng the Dsclosure perod (at the 0.01 level). In partcular, average allocaton to bonds ncreased by 5.4% (sgnfcant at the 0.01 level) whle average allocaton to equtes decreased by 4.3% (sgnfcant at the 0.01 level). A smlar pattern s observed for early adopters of FRS 17. These companes ncreased the allocaton to bonds by 3.4% (sgnfcant at the 0.02 level) and decreased the allocaton to equtes by 3.6% (sgnfcant at the 0.02 level). For comparson, we measured the changes n asset allocaton n US companes pror to the adopton of SFAS 158. We fnd that US companes decreased the allocaton to bonds by 3% and ncreased the allocaton to equtes by 5% (these changes are sgnfcant at the 0.01 level). The results n Table 3, Panel A, support Hypothess 1 UK companes decreased ther exposure to equtes durng the dsclosure perod of FRS 17. Panel B of Table 3 presents results for testng whether UK and US companes changed ther penson asset allocaton durng the adopton of full penson recognton under IAS 19/ FRS 17 n the UK and SFAS 158 n the US (Hypothess 2). The test statstc s constructed as the dfference between %Equty (and %Bond) n the pre-adopton and the post-adopton year (-1, +1). On average, UK companes ncreased ther allocaton to bonds by 3.7% and decreased the allocaton to equtes by 4.6% around the adopton of Full Recognton under FRS 17 / 19

23 IAS 19 (both changes are sgnfcant at the 0.01 level). The change n asset allocaton s more extreme n early adopters of FRS 17. Ths sub-sample of companes ncreased the allocaton to bonds by 4.8% and decreased the allocaton to equtes by 6.6% (both changes are sgnfcant at the 0.01 level). A change n asset allocaton s also observed n US companes around the adopton of SFAS 158. In partcular, US companes ncreased the average allocaton to bonds by 2.5% and decreased the average allocaton to equtes by 3.9% (both changes are sgnfcant at the 0.01 level). The results n Panel B of Table 3 support Hypothess 2. In partcular, the evdence suggests that both UK and US companes modfed ther penson asset allocaton polces by shftng penson assets from equty to debt securtes. In addton, the magntude of ths shft s, on average, smlar n both countres. Ths shft reduces the effects of full recognton accountng on the balance sheet and comprehensve ncome. (Table 3 about here) Whle the results n Table 3 support Hypotheses 1 and 2, there could be alternatve explanatons for the changes n penson asset composton. Pror lterature suggests three alternatve reasons for changes n penson asset composton: () hgher fundng levels, due to recent regulatory changes, caused companes to shft assets from equtes to bonds; () shorter nvestment horzons prompted companes to shft penson assets from equtes to bonds; and () ncrease n overall frm rsk mght have caused companes to shft assets from equtes to bonds. To examne the potental effect of fundng levels, nvestment horzon and frm rsk on asset allocaton, we examne the behavour of these factors startng three years before adopton of full penson recognton untl two year after adopton (one year after adopton for US companes). The results are presented n Table 4. 20

24 As Table 4 shows, average fundng levels (FUND) have ncreased for both UK and US companes around the adopton of the new standards. These ncreases, whch are sgnfcant at the 0.01 level, could explan the shft from equtes to bonds n UK and US companes. Investment horzons (HOR) ncreased around the adopton of the new standards n both countres (sgnfcant at the 0.01 level). However, an ncrease n nvestment horzon s nconsstent wth a swtch from equtes to bonds; thus, changes n nvestment horzons are unlkely to be the cause for the shft n asset composton. We further examned the proporton of companes that closed ther defned beneft plans to new entrants (CLOSE). Closng the fund to new entrants reduces the nvestment horzon over tme and could trgger a shft from equtes to bonds. As Table 4 shows, the proporton of closed plans ncreases n the UK (from 0.51 pror to adopton to 0.60 after adopton) and n the US (from 0.24 pror to adopton to 0.35 after adopton). These ncreases, whch are sgnfcant at the 0.01 level, could explan the movement of penson assets from equtes to bonds. As for frm rsk, we fnd that mean leverage (LEV) decreased over tme and mean frm sze (SIZE) ncreased over tme n both countres. The reducton n frm rsk, as reflected n leverage and sze, s nconsstent wth the swtch of penson assets from equtes to bonds, thus cannot serve as a potental explanaton for the shft of penson assets from equtes to bonds. The mddle panel of Table 4 presents mean varables for UK early adopters. We fnd that fundng levels and frm rsk remaned relatvely stable from three years before adopton untl two years after adopton. Investment horzon ncreased from the year before adopton untl the year after adopton (sgnfcant at the 0.09 level) and the proporton of closed plans ncreased from 0.36 to 0.42 around the adopton of FRS 17 (sgnfcant at the 0.08 level). Thus, the only alternatve explanaton to the swtch of penson assets from equtes to bonds n UK early adopters s the ncrease n the proporton of closed plans. (Table 4 about here) 21

25 So far, the results are consstent wth the argument that the shft n penson assets n UK and US companes s, at least partally, drven by new accountng standards, although and the ncrease n the proporton of closed funds and ncreases n fundng levels could also explan the shft from equtes to bonds. We now turn to cross-sectonal analyss n order to examne these effects n a multvarate settng. Frst, we estmate equaton 2, n whch the dependent varable s the change n the allocaton to equtes durng the Dsclosure perod (rdisclose) pror to the adopton of FRS 17/IAS 19. As Table 5 shows, the coeffcents on the mpact varables (EXPOS1 and EXPOS2) are postve, as expected, and sgnfcant at the 0.05 level. Ths result, whch supports Hypothess 3a, suggests that UK companes wth larger penson schemes relatve to shareholders' equty shft more funds from equtes to bonds. The evdence n Table 5 also hghlghts the role of other factors n explanng changes n asset allocaton durng the Dsclosure perod. Specfcally, UK companes that experenced an ncrease n fundng levels over the Dsclosure perod shfted more assets from equtes to bonds, as reflected by postve (sgnfcant at the 0.10 level) coeffcents on ΔFUND. In addton, companes that experenced an ncrease n the nvestment horzon shfted less assets from equty to debt securtes, as reflected by the negatve coeffcents on ΔHOR (sgnfcant at the 0.05 level). We also fnd that companes that experenced an ncrease n effectve tax rates (ΔTAX) shfted more penson assets to bonds, as expected. Furthermore, companes wth hgher fnancal leverage (ΔLEV) shfted more assets to bonds, probably to mtgate the effect of subsequent recognton on the volatlty of shareholders' equty. Fnally, companes that closed ther penson plans to new entrants durng the Dsclosure perod shfted more assets from equtes to bonds, as reflected by the postve coeffcents on ΔCLOSE (sgnfcant at the 0.01 level). 22

26 The results n Table 5 suggest that durng the Dsclosure perod, UK companes shfted penson assets from equtes to bonds because of changes n fundng levels, nvestment horzons, effectve tax rates, fnancal leverage and plan scope. Among these varables, the varable that had the most sgnfcant effect on asset allocaton s ΔCLOSE. In partcular, companes that closed ther penson plans to new entrants shfted more funds from equtes to bonds. However, after controllng for these effects, we fnd that UK companes wth larger penson plans relatve to shareholders' equty shfted more penson assets from equty to debt securtes durng the FRS 17 Dsclosure perod. We attrbute these fndngs, whch support Hypothess 3a, to the new dsclosure requrements of FRS 17. In partcular, these new penson dsclosures prompted companes wth larger penson plans relatve to shareholders' equty to reduce the volatlty of more vsble and transparent penson defcts. (Table 5 about here) To test Hypothess 3b, we estmate equatons (3), whch explans the change n the allocaton to equty securtes durng the Full Recognton perod of FRS 17/IAS 19 n the UK and SFAS 158 n the US. Table 6 reports results for the UK sample on the left panel and for the US sample on the rght panel. Startng wth the UK sample, the coeffcents on the man test varables, ΔEXPOS1 and ΔEXPOS2, are postve, as expected, and sgnfcant at the 0.05 level. Ths result, whch supports Hypothess 3b, suggests that UK companes wth larger penson plans relatve to shareholders' equty shfted more assets from equty to debt securtes around the adopton of FRS 17/IAS 19. In addton, the coeffcents on ΔFUND are negatve (sgnfcant at the 0.05 level). Ths result suggests that UK companes that experenced an ncrease n fundng levels shfted less penson assets from equtes to bonds. Also, the coeffcents on ΔTAXR are postve (sgnfcant at the 0.01 level), as expected, suggestng that companes wth hgher effectve 23

27 tax rates shfted more funds to bonds. Furthermore, the coeffcents on ΔLEV are postve (sgnfcant at the 0.05 level), as expected, suggestng that companes wth hgher debt shfted more fund to bonds to reduce the effect of full recognton on the volatlty of shareholders' equty. Fnally, the coeffcents on ΔCLOSE are postve, as expected, and sgnfcant at the 0.05 level, consstent wth the argument that companes that closed ther penson plans to new entrants durng the adopton perod shfted more funds from equtes to bonds. Fnally, the postve coeffcents on rdisclose suggest that companes that shfted more funds from equtes to bonds pror to adopton contnued to shft assets to bonds durng the adopton perod. Overall, these fndngs, whch support Hypothess 3b, are consstent wth the clam that UK companes wth larger penson plans relatve to shareholders' equty shfted more penson assets from equty to debt securtes durng the FRS 17/IAS 19 Full Recognton perod ncrementally to changes n fundng levels, effectve tax rates, fnancal leverage and plan coverage. 13 Table 6 also presents results for estmatng equatons (3) usng US data (rght panel). The coeffcents on the man test varables, EXPOS1 and EXPOS2, are postve and sgnfcant at the 0.10 level or better, whch supports Hypothess 3b. Ths result suggests that, smlar to UK companes, US companes wth larger penson plans relatve to shareholders' equty shfted more assets from equty to debt securtes durng the SFAS 158 Full Recognton perod. The results also hghlght the role of dvdend payout, effectve tax rates and fnancal leverage n asset allocaton of US companes. In partcular, companes that pay more 13 We also estmated equaton (3) for UK early adopters (42 observatons wth complete data). We fnd (results not tabulated) postve coeffcents on ΔEXPOS1 and ΔEXPOS2 (sgnfcant at the 0.10 level), as expected, suggestng that early adopters wth larger penson plans relatve to shareholders' equty shfted more funds from equtes to bonds. We also fnd postve coeffcents on ΔLEV (sgnfcant at the 0.10 level), suggestng that companes wth larger fnancal leverage shfted more funds to bonds possbly to reduce the lkelhood of volatng debt-related covenants. 24

28 dvdends shfted more funds to bonds to reduce the effect of the new standard on the stablty of dvdends. Also, as expected, companes wth hgher effectve tax rates and larger fnancal debt shfted more funds to bonds. In contrast to our results for UK companes, the coeffcents on ΔFUND, ΔCLOSE and rdisclose are not sgnfcant n explanng the shft to bonds. Overall, our fndngs are consstent wth the argument that US companes shfted penson assets from equtes to bonds to reduce the mpact of the new standard on the volatlty of shareholders' equty, whch n turn reduces the volatlty of dvdends and the lkelhood of volatng debt covenants. (Table 6 about here) 6. Conclusons FRS 17 and the revsed IAS 19 radcally changed accountng and reportng of defned beneft plans n the UK by ntally ntroducng new market-based penson dsclosures and subsequently requrng full balance sheet recognton of the penson surplus/defct. In December 2006, the FASB ssued SFAS 158 replacng the partal recognton method of SFAS 87 wth full balance sheet recognton of defned beneft post-retrement plans. To reduce the effect of captal market fluctuatons on reported earnngs, these standards requre that actuaral gans/losses shall be recognzed n comprehensve ncome. We nvestgate whether new market-based penson dsclosures had any mpact on penson asset allocaton of UK companes. We also examne whether asset allocaton of UK and US companes changed around the adopton of the new standards. We dentfy a Dsclosure perod durng whch UK companes had to dsclose all the requred data under FRS 17 n the notes to the fnancal statements wthout formally recognzng the full penson surplus/defct on the balance sheet. In addton, we dentfy a Full Recognton perod around the adopton of ether FRS 17 or IAS 19 n the UK and SFAS 25