Working Paper Determinants of occupational pension provision in Germany

Transcription

1 econstor Der Open-Access-Publkatonsserver der ZBW Lebnz-Informatonszentrum Wrtschaft The Open Access Publcaton Server of the ZBW Lebnz Informaton Centre for Economcs Dummann, Kathrn Workng Paper Determnants of occupatonal penson provson n Germany Thünen-seres of appled economc theory, No. 75 Provded n Cooperaton wth: Unversty of Rostock, Insttute of Economcs Suggested Ctaton: Dummann, Kathrn (2007) : Determnants of occupatonal penson provson n Germany, Thünen-seres of appled economc theory, No. 75 Ths Verson s avalable at: Standard-Nutzungsbedngungen: De Dokumente auf EconStor dürfen zu egenen wssenschaftlchen Zwecken und zum Prvatgebrauch gespechert und kopert werden. Se dürfen de Dokumente ncht für öffentlche oder kommerzelle Zwecke vervelfältgen, öffentlch ausstellen, öffentlch zugänglch machen, vertreben oder anderwetg nutzen. Sofern de Verfasser de Dokumente unter Open-Content-Lzenzen (nsbesondere CC-Lzenzen) zur Verfügung gestellt haben sollten, gelten abwechend von desen Nutzungsbedngungen de n der dort genannten Lzenz gewährten Nutzungsrechte. Terms of use: Documents n EconStor may be saved and coped for your personal and scholarly purposes. You are not to copy documents for publc or commercal purposes, to exhbt the documents publcly, to make them publcly avalable on the nternet, or to dstrbute or otherwse use the documents n publc. If the documents have been made avalable under an Open Content Lcence (especally Creatve Commons Lcences), you may exercse further usage rghts as specfed n the ndcated lcence. zbw Lebnz-Informatonszentrum Wrtschaft Lebnz Informaton Centre for Economcs

2 Thünen-Seres of Appled Economc Theory Thünen-Rehe Angewandter Volkswrtschaftstheore Workng Paper No. 75 Determnants of Occupatonal Penson Provson n Germany by Kathrn Dummann Unverstät Rostock Wrtschafts- und Sozalwssenschaftlche Fakultät Insttut für Volkswrtschaftslehre 2007

3 Determnants of Occupatonal Penson Provson n Germany by Kathrn Dummann* Abstract Demographc change causes an undersupply of fnancal old age benefts wthn the statutory pay-as-you-go penson system n Germany. Therefore, the provson of occupatonal as well as prvate pensons has to be enhanced. However, there seems to be an undersupply of occupatonal penson provson partcularly n small and medum szed enterprses (SMEs). Usng survey data of the German Soco-Economc Panel (GSOEP) and the German SAVE survey, the present paper studes econometrcally the determnants of occupatonal penson provson n Germany. It shows that occupatonal pensons depend not only on supply-sde factors such as frm sze and ndustry, but also on demand-sde factors such as ndvdual socodemographc attrbutes and people s savngs motves. JEL-classfcatons: C25, G23, J14 key words: Occupatonal Pensons, Retrement Provson, Demographc Change, SMEs * correspondng author: Unversty of Rostock, Department of Economcs, Ulmenstr. 69, Rostock, Germany, kathrn.dummann@un-rostock.de

4 1 1. Introducton Snce the late 1960s, the German populaton ages whle the rate of reproducton s stablzng at a level of 1.4 chldren per woman. Ths demographc change results n an undersupply of fnancal old age benefts through the statutory pay-as-you-go penson system, f contrbuton rates to ths system and/or retrement age reman constant. Therefore poltcans as well as researchers clam that own responsblty for old age ncome should be enhanced by more use of prvate and occupatonal penson systems. Also people themselves are not satsfed wth ther fnancal stuaton for old age and are wllng to place responsblty for fnancal ncome on to both prvate and statutory organzatons (Wunder and Schwarze 2005, p.25). Incentves are set by the state through nstruments lke tax benefts for both frms and employees, portablty rules or shorter vestng perods. Some researchers recommend a swtch from the pay-as-you-go system to the fundng prncple whch would be useful for each pllar (statutory, occupatonal and prvate). An advantage of ths captal-based system s that t s based on an nterpersonal redstrbuton nstead of an ntertemporal one. However, a postve real nterest rate cannot be guaranteed (Krupp and Weeber 2001, Breyer 2000). Snce January 1 st, 2002, the German occupatonal penson system va Entgeltumwandlung s mandatory for employers f employees ask for t. To save for retrement fve penson plan types (Durchführungswege) are optonal for the employer (BetrAVG 1a I, as amended on (BGBl. I p. 3242)). 1 They allow for three dfferent methods of fnancng (employer-based, employee-based or hybrd fnanced). Up to now, the employer-based system has been domnant n Germany, but because of complexty many employers do not provde any occupatonal penson yet or would lke to swtch to an employee-based system. Ths results n an undersupply of occupatonal old age pensons because employees are not oblged to demand such a system. Ths problem s common also n other countres lke the US, UK and Norway. Especally small and medum-szed enterprses (SMEs) face the problem of undersupply. In descrptve surveys they often name hgh admnstraton costs and cost of search for or nformaton about external supplers and ther offered range of products as an obstacle for provdng an occupatonal penson. Economc research stll msses to explan econometrcally the ncdence of occupatonal pensons n Germany. The purpose of ths paper therefore s to evaluate 1 BetrAVG = Law on the Improvement of Company Pensons. Penson plan types are: book reserve schemes (Drektzusage), drect nsurance (Drektverscherung), Pensonskasse, support fund (Unterstützungskasse), and penson fund (Pensonsfonds).

5 2 the determnants of the observed gap n occupatonal penson provson between German frms. We frst formulate hypotheses on supply and demand for occupatonal pensons and provde an overvew of recent lterature n secton 2. In secton 3, the hypotheses wll be examned econometrcally on the bass of the German Soco- Economc Panel (GSOEP) and the German SAVE survey. Fnally, secton 4 concludes and gves an outlook on future research. 2. Overvew of Prevous Lterature 2.1 Stylzed Facts Several analyses based on surveys showed that a gap n penson provson between SMEs and large frms was an apparent phenomenon n Germany, the US, UK and Norway. Some of these found sgnfcant dfferences across lnes of busness. In Germany, most of the ntervewed people state that they do not beleve the statutory penson system to be effectve enough to provde money for old age. Indvdual retrement ncome by statutory penson paybacks won t be suffcent to hold the standard of lvng. However, many of them do not seem to be adequately nformed about ths ssue. The same apples to employers. More than 75% of them are not aware of the fact that they are oblged to provde ther employees a system of deferred compensaton (R+V 2004, p.5). Most of the authors dentfy a gap between SMEs and large frms n penson provson, as well as between Eastern and Western Germany (Kortmann and Haghr 2005, Müller 2004, R+V 2004). An amazng result of the surveys s people s unsatsfactory wllngness to save addtonal money for retrement. It can be explaned by the fact that the lower people s ncomes the lower are ther savngs for retrement (Refner et al. 2003, p.19/20). The German socal securty system offers poor and unemployed people the possblty to receve a fx transfer ncome (Sozalhlfe) whch does not enable people to make payments to an occupatonal system dependent on ncome. New fndngs n those surveys are provded by answers to the questons why people want to partcpate n a penson plan and why they reject partcpaton. The most often named arguments to enhance demand are advce by and competence of the consultant or the employer. Employees are aware that they are not able to nform themselves about recent developments or statutory advancements. Therefore they expect consultants to act as nformaton ntermedares. If ths crteron s fulflled properly employees demand an occupatonal penson plan (HDI PM 2005, p.20/21, R+V 2004, p.6).

6 3 People are more lkely to demand a penson plan f they realze that ther employer feels socal responsblty 2 or wants to ncrease motvaton and corporate dentfcaton. In most cases of penson rejecton, employees are not nterested n retrement provson, because they thnk the offered system s too rsky, the benefts from the statutory penson system are overestmated, or they wll not have enough money left to save. Sometmes, employees also fear to ask ther employer for a plan (HDI PM 2005, Müller 2004, Delta Lloyd 2004). 3 Smlar results are found by other authors who evaluated whether moblty of employees served as an addtonal factor of nfluence (see table 1). Table 1: Survey of smlar studes on job moblty and occupatonal pensons Hernæs et al. (2006) Rabe (2004) Andrett (2001) Taylor (2000) Even/ Macpherson (1994) Allen/ Clark/ MacDermed (1993) Gustman/Stenmeer (1993) Norway young employees have a shorter job tenure, do rarely adopt an occupatonal penson due to the wll to get ahead Germany moblty vares n terms of sectors and wage dfferentals between employers Denmark, Ireland, UK, Netherlands occupatonal pensons are not repressng moblty nternatonally, need to dfferentate between DB and DC plans UK USA USA USA not able to young male prove that people wsh employees do to be more not change flexble and jobs due to prevent less occupatonal pensons Source: own creaton Note: DB = defned beneft plan, DC = defned contrbuton plan portablty serves as a mechansm of selfselecton and s moblty enhancng lack of penson portablty does not affect job moblty, backloadng effect s very low Addtonally, Gustman et al. (1994) menton that moblty s affected by employee s current wage and ther potental captal loss. Rabe (2004) and Andrett (2001) prove ths argument for several European countres. As Germany, countres lke the US, UK, and Norway face the followng problems: low partcpaton rates n prvate or occupatonal penson systems and less retrement plan provson n SMEs than n large companes. A commonly named reason for ths s: young employees and those wth low ncome reject a partcpaton n an occupatonal plan. Indvdual reasons are dfferent (Ghlarducc and Lee 2005, Hnz and Turner 1998, Even and Macpherson 1994). Moreover, t has been shown that long job tenures are the result of offered penson plans (Meall and Pudney 1996). 2 Ths s a modern substtute for employee retenton. 3 These surveys nclude only answers from employees, unemployed are left out as they are not able to make payments to an occupatonal penson system durng ther perod of unemployment. Self-employed people are excluded n most cases because ther savngs for old age can be added to prvate penson provson.

7 4 Fnally, despte 401(k) plans 4 whch are popular n the US and do not have excessve admnstraton cost, partcularly SMEs reject penson plan provson. All these results show that penson gaps between SMEs and large frms are an mportant polcy pont at present n varous countres. The studes for Germany named above reman on a descrptve level to show the development of occupatonal penson plan supply and demand. It s therefore necessary to show econometrcally whch determnants have a statstcally sgnfcant mpact on penson plan provson. To fll ths gap, we derve several hypotheses from economc theory n the followng chapter. 2.2 Theoretcal Background and Hypotheses Gven the gaps n the provson of penson plans by the statutory penson system, reasons as to why people demand a certan category of penson need to be evaluated. Mtchell (1988) notes that employees make up suboptmal decsons concernng ther penson plans as they are mperfectly nformed about tax deference, dfferent channels to choose and other characterstcs. She mentons a dvergence of efforts on employee and employer sde due to the fact that employees have dfferent ncentves to ask for a plan. Employers are more lkely to offer a plan f the frm s proftable. An equlbrum between demand by employees and supply by employers could be acheved only f we lved n a world wthout nformaton cost. Several authors dentfed parameters nfluencng a person s ndvdual decson to partcpate n an occupatonal penson plan subject to the assumpton that employers offer some. What reasons do arse for the demand for and supply of occupatonal pensons? Economc lterature developed dfferent hypotheses to explan nfluences on both employee sde and employer sde. The frst three arguments refer manly to supply sde factors, although a clear separaton between supply sde and demand sde s not possble. Frst, penson systems could act as a mechansm of employee retenton. Ths may be n the nterest of both employees and employers. Partcpaton n a plan mples that the employer saves a certan amount of money for the employee s retrement perod. Ether the same amount of money s saved as a wage deferral (defned contrbuton) every perod or the employee s beng pad out a fxed penson amount at the 4 401(k) plans are used n the US as a hybrd fnanced nstrument of deferred compensaton wth an opton for the employee to opt out of ths system. Ths way employees are permtted to beneft from current tax deferrals on current wages (Kusko et al among others). A short dscusson of 401(k) plans can be found n Munnell and Sundén (2006).

8 5 begnnng of the retrement perod (defned beneft). Gustman and Stenmeer (1993) show that defned contrbuton plans are more portable and therefore assocated wth hgher employee turnover than defned beneft plans as a new employer only needs to pay the defned contrbutons to the employee wthout havng to guarantee the fnal amount of pensons from a defned beneft plan. Taylor (2000) supports ths hypothess for the UK. Several authors call beneft plans a sponsorng mechansm of employers due to the fact that the employees contrbutons are save for retrement and therefore act as a backloadng mechansm. By fundng a penson, the employer receves a credt from the employee whch wll be pad back when the latter retres. However, the employee s lkely to loose the compound nterest rate f he or she quts the job before retrement and changes to another employer. Then a new penson wth a lower captal stock has to be bult up there (Dorsey et al. 1998, Ippolto 1997, Kotlkoff and Wse 1985, Lazear and Moore 1984). 5 A further argument of matchng the nterests of employer and employee s made by Backes-Gellner and Pull (1999) who clam that the provson of a beneft n retrement perods s effcent f employees consder benefts to be better than the monetary equvalent at present. Ths supports the fndng of Mtchell (1988) that t s better to nform employees about penson benefts and thus ncrease ndvdual acton to nsure themselves for old age. A second argument that s closely related to the hypothess of employee retenton s based on Lazear (1979). He shows that employees are able to realze the ndvdual beneft amount of a penson plan. Therefore they avod shrkng before retrement. Ths s due to effcency wages: employees earn lower wages than ther margnal product of labor at younger age and hgher wages when they get older. So employer and employee buld up an mplct contract. An antcpated possble captal loss evoked by job loss because of shrkng leads to a reducton of employee moblty. Otherwse people would look for a job n whch they are not offered a penson. Moreover, the avodance of shrkng may have the addtonal effect of decreasng tranng costs and the costs of fndng and recrutng new staff as employees have an ncentve to work harder and have a longer tenure (McCarthy 2006, Allen et al. 1993). An mpact on motvaton and loyalty by frnge benefts or occupatonal pensons could be proved for the US (Allen et al. 1993) but not for Germany (Frck et al. 1999). Our frst hypothess therefore s 5 In Germany ths mechansm depends on the arrangement between employer and employee. The German law ( 1b BetrAVG) guarantees the repayment of contrbutons to a penson system f the employee s older than 30 years or the enttlement exsts for at least fve years when the employee changes the job.

9 H 1 : A longer tenure n a frm goes along wth hgher motvaton. As motvaton s enhanced by occupatonal benefts there s a postve relatonshp between the probablty of adoptng a penson and tenure. 6 The thrd argument refers to economes of scale acheved by large frms. If a frm employed a large number of employees t could offer a penson plan that requred possbly hgher fxed costs of admnstratve expenses. 6 So, the costs per employee decrease wth frm sze. Furthermore, large frms have a hgher probablty to survve busness cycle fluctuatons and are more able to earn extra money to pay for hgher pensons. A small frm would not be able to pay a penson f t went bankrupt. The rsk of becomng nsolvent s hgher for small frms than for large frms. Furthermore, a case of death or nvaldty could result n hgh payments, whch reduce cash-flow and may cause nsolvency, f the employer dd not save enough for an event lke that. Ths rsk s hgher for small frms than for large frms because large frms are more able to calculate the probablty of a sudden case of death. Fnally, economes of scale and bankruptcy rsk dffer across lnes of busness (Horba and Yoshda 2002, Aok 1984, Mtchell and Andrews 1981). Ths leads us to two further hypotheses: H 2 : Large companes are more lkely to offer a penson plan than small companes. H 3 : The demand for occupatonal penson plans dffers across lnes of busness and regons. From the vew of human captal theory we have to add that people who are well nformed and stay longer at a frm tend to be more productve than others. The reason for ths s that employees acqure frm-specfc human captal that as a result ncreases frm productvty (Becker 1964). Ths could not be rejected for the US (Cornwell and Dorsey 2000). As Mtchell (1988) ponts out educated employees and women are better nformed about detals of penson plans as they are more lkely to be nvolved n the process of gettng nformaton about penson determnants, therefore we develop hypothess 4. H 4 : As female and well educated people are better nformed, they know better about the advantages of occupatonal pensons and therefore have a hgher probablty of partcpatng n a plan. 6 Admnstraton costs dffer across ways of penson provson.

10 7 Mncer (1976) tested several socodemographc factors (lke schoolng, hours worked, gender) and ther nfluence on ndvdual ncome. We suggest that these factors nfluence retrement ncome, too, as Blundell et al. (1994) found out that demographc attrbutes and labor market varables do sgnfcantly nfluence ntertemporal behavor. Freeman (1985) addtonally uses nformaton about labor unon membershp based on the model of unon monopoly power and ts collectve voce, showng both theoretcally and emprcally that t has a postve nfluence on penson plan demand. Besdes, Gustman et al. (1994) argue that moblty s affected not only by the employee s current wage, but also by the captal loss occurred from a job change. Thus, we expect: H 5 : The demand for occupatonal penson plans depends on socodemographc attrbutes (e.g. ncome, age, martal status, poston, and number of chldren). H 6 : Savngs motves based on ndvdual crcumstances nfluence the wllngness to provde for retrement. These hypotheses shall be tested n the followng chapter.

11 3. Emprcal analyss 3.1 Data The data used n ths study have been taken from the German Soco Economc Panel (GSOEP) and the German SAVE survey. The GSOEP s an annual survey among German ndvduals wth questons concernng several socoeconomc factors wth every wave focusng on a specal topc. From 1985 to 1995 questons concernng occupatonal pensons were asked. Thus, the most recent data avalable refer to The dataset of our analyss contans people who were asked to partcpate n the survey. 8 The answers to the questons about occupatonal penson coverage are dsplayed n table 2. As shown n the frst column, n 1995 almost 2000 people n our dataset worked n a frm that offered a penson plan, but only 1392 of them (73%) were enttled n a plan (see second column). It s an astonshng fact that 16% of those who answered the queston In some companes the employees are offered a company penson n addton to socal securty. Does your company offer ths for all or some of ts employees? dd not know whether ther frm offered a penson plan. Ths s a hnt to an undersupply of nformaton nsde frms as a lot of people are not aware of ther possblty to make retrement provsons. Moreover, 87.8% of the people consdered n ths study are not enttled n plans from an earler employer. Ths could possbly be due to the fact that the mean age of people ownng an occupatonal penson s 41.7 years. People of ths age are more lkely to be bound to the frm than younger employees who are more career-mnded. 35.7% of the total number of employees enttled n an occupatonal penson plan are female, and 64.3% of them are male. Table 2: Answers to questons about supply and clams of occupatonal pensons (percentages n parentheses) enttled n an occupatonal penson clams from earler employer supply yes 1941 (24.16) 1392 (73.15) 467 (5.85) no 4778 (59.47) 305 (16.03) 7009 (87.78) don t know 1315 (16.37) 206 (10.82) 509 (6.37) total 8034 (100) 1903 (100) 7985 (100) Note: Ths table ncludes self-employed people. Later on, they wll be left out of our analyss, because most other studes attrbute occupatonal pensons of self-employed people to the prvate pllar. Source: GSOEP, wave 12 (1995) 8 7 Due to a small number of answers n earler waves and therefore resultng small varance n the dataset we concentrate our analyss on the year 1995 and compare the results to those of the year 2005 based on the German SAVE survey. As the Rester-penson had been ntroduced n 2001 we are not able to evaluate a relatonshp between Rester penson and occupatonal pensons n The number of answers dffers between the dfferent parts of the questonnare. For a detaled descrpton of the survey see Hasken-DeNew and Frck (2005).

12 Another possble reason for the apparent low moblty mght be that before 2002 the portablty and vestng rules had been more restrctve and more complex. The rules concernng vestng perods are a popular example 9 : An employee partcpatng n an occupatonal penson plan keeps hs enttlements f he leaves the frm before occurrence of an nsured event only f he s at least 35 years old and ether the penson promse exsts for at least 10 years or the employee s employed at that frm for at least 12 years and hs penson promse exsts for at least 3 years (BetrAVG 1 I 1, as amended on (BGBl. I p. 2911)). After excluson of self-employed people from our analyss 31% of those who are subordnates n a small frm do not own an occupatonal penson whereas almost 85% of those employed at a large frm (2000 and more) have clams n a penson system. 10 Ths can be seen from table Table 3: Percentage of clams n terms of frm sze clams n a penson system frm sze no (n %) yes (n %) < 5 to < to < to < and more Source: GSOEP, wave 12 (1995), weghted, frm sze measured by number of employees Concernng hypothess 5 derved above, our dataset suggests that people wth hgher wages 12 have a hgher probablty to partcpate n an occupatonal penson scheme. To show ths we took the ncome quartles and related them to frm sze wth respect to the fact that the person owns a clam n occupatonal pensons. Table 4 ponts out that n our dataset most people wth hgh ncome are employed n large frms, whch refers to the thess that large frms acheve economes of scale n admnstraton costs. Table 4: Number of penson partcpants wth respect to ncome and frm sze frm sze ncome quartles below 5 5 to <20 20 to < to < 2000 > st quartle nd quartle rd quartle th quartle total Note: frm sze measured by number of employees Source: own calculatons based on GSOEP, wave 12 (1995) 9 9 The BetrAVG has been revsed several tmes and legslators tred to smplfy the rules but n some cases lke vestng and portablty t s stll complex due to several exceptons self-employed people are excluded from ths part of our analyss. 11 SMEs can be defned dfferently: Some use yearly turn-over or total assets. The EU defnes SMEs up to a number of 250 employees. Devatng from ths we defne n our dataset SMEs due to the data structure from less than 5 employees to an upper lmt of 200 employees. 12 We approxmate wages by usng average monthly ncome.

13 On a descrptve bass no clear answer can be gven on the queston about the relaton between educaton and penson coverage. Table 5 shows that 81% of those wth a unversty degree are enttled n a penson plan but almost 83% of those wth no unversty degree as qualfcaton for ther current job are enttled n a penson plan, too. 13 Table 5: Relaton between unversty degree and penson coverage (percentages n parentheses) no penson penson no unversty degree unversty degree Source: own calculatons based on GSOEP, wave 12 (1995), weghted. In order to fnd out whether people s atttude towards penson plan adopton changed durng the past ten years we use the German SAVE survey that besdes socoeconomc questons ncludes questons on people s savngs behavor as well as atttudes towards assets and ncome. The survey started n 2001 and the second wave followed n The thrd wave we use, ncludng 2305 households, was collected n summer The weghts used on the multvarate level are based on the dstrbuton of ncome and age of the German sample census of the respectve year. After excludng unemployed people (because they are not able to acqure clams from occupatonal penson plans that year) we are able to do estmatons based on the answers of 2102 heads of the households. 15 In tmes of demographc change wth a decreasng amount of money pad to retrees by the statutory penson system t could be useful for those who beleve to have an above-average lfe expectancy to provde for old age va occupatonal pensons. These may care for themselves as they would be extremely nfluenced by lower payoffs and could possbly face old age poverty. The SAVE data show on a descrptve bass that only 15.0 (8.5) percent of those expectng to lve longer than the average use occupatonal (prvate) penson plans to fll ther ndvdual gap n fnancal benefts (see table 6). But there s stll the queston whether supposed lfe expectancy has a statstcally sgnfcant nfluence on ndvdual penson plan demand On the multvarate level weghts based on a person level projecton factor defned as the nverse of the corrected probablty to be drawn (korrgerte Zehungswahrschenlchket) are used. 14 More detaled nformaton on the composton of the dataset and ts methodologcal aspects can be found n Schunk (2006a) and Schunk (2006b). 15 In contrast to the GSOEP the SAVE dataset contans nformaton on savngs behavor whch makes t easer to count for certan savngs motves but we are no longer able to count for lnes of busness or frm sze. Therefore, we are not able to estmate whether substantal dfferences n penson coverage exst across lnes of busness or frm szes.

14 Table 6: Lfe expectancy and penson plans (percentages n parentheses) 11 occupatonal penson prvate penson lfe expectancy longer than average no yes no 1432 (87.0) 380 (85.0) yes 223 (13.0) 67 (15.0) lfe expectancy longer than average no yes no 1487 (89.85) 409 (91.5) yes 168 (10.15) 38 (8.5) Source: own calculatons based on SAVE dataset 2005 Furthermore, t can be argued that people partcpatng n prvate penson plans may value these as a substtute for occupatonal pensons. The SAVE dataset shows that 19.3 percent of those partcpatng n an occupatonal penson plan have a prvate penson as well and 8.3 percent of those not havng an occupatonal penson have a prvate penson nstead (see table 7). It s amazng that more than 90 percent of those not partcpatng n occupatonal plans do not provde for old age on ther own. 16 Calculated wth reference to the whole survey we see that 79% of all people partcpatng n the SAVE survey care nether prvately nor occupatonally for old age. Therefore the descrptve statstcs do not support the argument of occupatonal and prvate pensons beng substtutes. Table 7: Prvate versus occupatonal pensons (percentages n parentheses) occupatonal penson prvate penson no yes no 1662 (91.7) 150 (8.3) yes 234 (80.7) 56 (19.3) Source: own calculatons based on SAVE dataset 2005 A closer look has to be taken on the multvarate level. To test the above hypotheses, we wll do the followng estmatons Estmaton method We use data consstng of all male and female employees workng full-tme or parttme 18 above age 16. People can hold clams n occupatonal pensons ether at the 16 A possble explanaton mght be a msunderstandng n the questonnare: Many people who partcpate n ths survey mght not have a usual product whch could be counted as a prvate or occupatonal way of penson provson. But n 1995 almost 9 mllon households (almost 2.3 mllon households at the age of years) n Germany were owners of a flat or a house whch can be counted as a way to provde for old age (source: GSEOP 1995). Ths mght be due to the fact that real estate (or reverse mortgage) as a mechansm to provde for retrement becomes more and more popular now but had not been regarded that way n The same argument apples to lfe nsurances. 17 The mportance of mcro-level data for demand analyss compared to aggregate data has been evaluated and dscussed by Blundell et al. (1993).

15 12 actual employer or at a former employer or not at all. 19 As the resultng dependent varable s a bnary one we have to estmate a bnary probt model underlyng the assumpton of a cumulatve standard normal dstrbuton functon. Frst we estmate the probablty P of an ndvdual person (or a frm) ownng (offerng) an occupatonal penson (Y = 1) n terms of a vector of socoeconomc attrbutes X whch can be wrtten as 1 β * P = Pr( Y = 1) = Pr( I I ) = F( I ) = + 2π 1 β2x 2 t wth I* ncorporatng an unobservable utlty ndex 20 of ndvdual and t beng a standardzed normal varable t ~ N (0,1). e dt (1) As the utlty ndex I s beng defned as 1 1 I = F ( I ) = F ( P ) = β 1 + β 2 X t can be estmated econometrcally as (2) wth u beng the error term I β X + u. (2) = 1 + β 2 Socoeconomc attrbutes ncluded n vector OP TEN FIRMSIZE INDUSTRY REGION GENDER EDU INC MARITAL AGE X of ndvdual are: dummy varable equal to 1 f ndvdual acqures clams of an occupatonal penson or (n a second set up) f a frm offers ndvdual a penson plan tenure (years) of employee at hs current employer vector of dummy varables for the number of employees of the frm ndvdual s employed n 21 vector of dummy varables for ndustry at one-dgt level dummy varable equal to 1 f ndvdual lves n Western Germany dummy varable equal to 1 f person s male dummy varable equal to 1 f ndvdual needed a unversty degree to get hs or her job 22 average gross monthly ncome of person (n 2005 we use average net ncome of person ) dummy varable equal to 1 f ndvdual s marred age (years) of person 18 Some studes only nclude employees workng more than 37.5 hours a week. We ncluded also people workng part-tme nto our analyss, snce both groups are offered occupatonal pensons. Unemployed people n occupatonal tranng or socal programs (lke year spent by young people dong work n the area of socal servces or mltary servce) are left out of the analyss. 19 Those who answered don t know wll be treated as mssng. 20 Ths s what Gujarat (1995) names a crtcal value n decson makng. 21 D1: less than 5 employees, D2: 5-19, D3: , D4: , D5: 2000 and more 22 For 1995 we used EDU as proxy for educaton of ndvdual because data are not avalable. Ths way we are able to catch up tme preferences as those who do longer beneft from educatonal servces n order to earn hgher wages n future have a small tme preference. Therefore they are less lkely to demand pensons or other forms of savngs (see for example Börsch-Supan (2001) and Rodepeter (1997) among others).

16 FOP POSITION CHILD MOTIVE dummy varable equal to 1 f person owns an occupatonal penson from a former employer dummy varable equal to 1 f ndvdual s a blue-collar worker dummy varable equal to 1 f there are chldren under age 16 lvng n the household of ndvdual, n 2005 t s the number of chldren under 16 lvng n the household of ndvdual vector of varables ncludng dummy varables on whether person expects to lve longer than average, whether owns a Rester penson (supported by the state) and whether owns a prvate penson (except Rester), as well as questons on the degree of rsk to become unemployed, on satsfacton wth the ndvdual economc stuaton, on the standard of lvng and on whether person gets by wth ndvdual ncome. 13 Therefore the structural part of the regresson can be wrtten as (3) OP = α + α TEN + α GENDER 5 + α FOP + α POSITION α FIRMSIZE + α EDU α INC 7 + α CHILD 12 + α INDUSTRY 3 + α MARITAL 8 + α MOTIVE 13 + α REGION + α AGE u (3) Afterwards we evaluated what attrbutes sgnfcantly nfluenced the duraton of clams n occupatonal pensons. Therefore, we dd OLS estmatons usng the same socoeconomc attrbutes lke n equaton (3) but wth the logarthm of duraton (years) of clams n occupatonal pensons as the dependent varable and the log of ncome and age as ndependent varables. We also tred to estmate determnants of the supply of an occupatonal penson on frm sde. 23 However, due to nsuffcent attrbutes of the frms that supply an occupatonal penson to ther employees these regressons could be mproved f more frm specfc attrbutes were avalable. 24 All observatons have been weghted n our multvarate analyses. 3.3 Results Our regressons provde nsghts nto the German occupatonal penson system and people s penson plan adopton. 25 To start wth the supply sde of our analyss (see table 9) we dd regressons on the mpact of frm-specfc varables nfluencng the supply of a penson plan by a 23 To do ths we used frm-specfc nformaton provded by the ndvdual answers of those partcpatng n the GSOEP. 24 An econometrc study on the supply of an occupatonal penson n Lower Saxony has been carred out by Schnabel and Wagner (1999). 25 The results of our senstvty analyses on supply and demand sde arguments are dsplayed by tables 9 to 12 n the appendx.

17 14 company. Column I shows the mpact of gross ncome and frm sze on penson plan provson. We fnd that ncome (as a proxy for wages pad) has a postve nfluence on penson plan supply. 26 Furthermore, n the regressons tabulated n columns I through III we dentfy statstcally sgnfcant dfferences between frm szes. Frms wth a small number of employees (no more than 20) do less often offer an occupatonal penson plan than frms wth more than 2000 employees compared to the reference group. 27 Across lnes of busness (columns II and III) we dentfed statstcally sgnfcant dfferences. Fnancal ntermedares and manufacturng frms are more lkely to provde a penson system than frms n the constructon ndustry and the reference group. Wth reference to the queston whether people demand a penson plan and how long they acqure clams from t we are able to pont out the followng. Table 10 and table 11 can be used to compare the penson plan demand n 1995 wth that of They show that n 1995 provson behavor s less motvated by personal attrbutes but caused by frm characterstcs. In 2005 ndvdual factors become more mportant. The sgn of personal attrbutes remans the same n all cases. Personal ncome has the expected postve and sgnfcant nfluence on the dependent varables as proposed by H 5. People who earn hgher wages are more able to provde for the future. Furthermore, the mpact of age s statstcally hghly sgnfcant whch ndcates that older people are aware of the fact that they should prepare for ther retrement age. The nfluence of martal status s only predctable for Therefore t cannot be concluded that sngle adults seem to be less accountable for others and therefore care less for themselves. Marred have a statstcally sgnfcant hgher probablty to partcpate n a plan n The nfluence of beng a man s not sgnfcant so that t s not possble to reject (or not) the hypothess of women beng better nformed than men and therefore havng a hgher probablty to adopt a penson. Gven that our analyses nclude only workng men and women ths argument needs to be taken carefully. On the contrary, n 1995 those who are supposed to be hghly educated (people havng a unversty degree) provde less for old age by the second pllar. In 2005, we are not able to fnd evdence. A possble reason for ths result mght be that young, hgh-sklled employees, who are more career-mnded, neglect old age penson 26 An ncluson of other frm-specfc factors (lke turnover or number of employees) could enable us to draw more concrete conclusons on penson plan supply from employer sde but due to data shortenngs we are not able to do that. 27 It would be nterestng to evaluate whether the age of a company has a statstcally sgnfcant mpact on the supply of an occupatonal penson plan.

18 15 provson because they want to be more moble, and are not wllng to bnd themselves to a company facng reductons n the captal stock f they change jobs. Besdes ths, hgh sklled employees mght possbly use the thrd pllar (prvate pensons) as a substtute for occupatonal pensons because fewer problems lke restrcted portablty occur there. 28 To control for ths wth the data of 2005 we used two varables (Rester penson 29 and prvate penson) that nclude nformaton about the ndvduals atttude towards prvate penson provson (table 11, columns VI and VII). In 2005, both the ownershp of a Rester penson or another prvate penson have a statstcally sgnfcant postve nfluence whch s aganst the opnon that prvate pensons are substtutes for occupatonal pensons. It seems as f the chance to care occupatonally s hgher for those who provde for old age prvately as well because they are better nformed about the need to save. As both employed men and employed women are ncluded n our analyss t cannot be stated that typcally men act as the head of the household and therefore care more for ther famly. To have chldren under age 16 lvng n the household based on the argument of care does not have a statstcally sgnfcant mpact on adopton or duraton of clams. But the larger the number of chldren an ndvdual has n 2005 the less he or she provdes for old age. A popular argument for ths s that famles need to spend money to care for ther famly and therefore have less money left to save. Furthermore, f people own a plan from a former employer they are more lkely to have a plan at ther current employer as they seem to be nformed about ther possbltes and show the wllngness to care for retrement ncome (see table 10, columns II to VI). Therefore, H 4 can nether be rejected nor confrmed. Beng employed as a blue-collar worker does not have a clearly dentfed nfluence on both penson plan adopton and duraton of plans n all cases. Addtonally, wth tenure beng hghly statstcally sgnfcant we are not able to reject H 1. People who plan to care for retrement ncome seem to be aware of the fact that they need to proft from the compound nterest effect and therefore are wllng and motvated to stay longer at an employer and hold a penson. 28 In our study we are not able to control for ths possblty as the people partcpatng n the GSOEP have only been asked questons about prvate provson n The SAVE dataset we use to control for changes durng the past ten years ncludes answers on prvate penson provson. 29 The Rester penson had been ntroduced n 2001 by the German mnstry of labor and socal affars wth the am to support especally people of lower ncome levels. Therefore people demandng ths product are offered support by tax deferrals.

19 Lke Mtchell and Andrews (1981), who were able to show dfferences across lnes of busness and regons for the US we fnd both ndustry and regonal effects for Germany although they cannot be dentfed clearly (see table 10 columns II to VI and table 11 columns II to VII). 30 People employed n Western Germany are statstcally more probable to be ncluded n a plan due to hstorcal effects. Besdes, people from Eastern Germany mght have a dfferent atttude towards fnancal old age provson as they could rely more on state actvty. The economc development of each regon and ts nfluence on frms may play an mportant role because those frms that face a postve economc development are more lkely to provde a penson plan due to a hgher ablty to fund them. Dfferences n sector-specfc demand and supply for occupatonal pensons may be related to the hstorcal background, too. In constructon, sales, hotels and restaurants t has not been common n hstory to provde penson plans. Another possble reason mght be that people employed n sectors lke farmng and constructon have a lower mean tenure than employees n other sectors (see table 8). Therefore they are less lkely to provde for old age. An argument for the postve sgn of fnancal ntermedares mght be that ths lne of busness s one of the external ntermedares to offer penson products. They may use ther advantages n nformaton and should therefore be aware of the need to prevent. Table 8: Mean tenure by lne of busness tenure st.d. farmng/agraran 10.3 (9.7) mnng 11.4 (8.3) manufacturng of goods 15.8 (10.0) electrcty 15.8 (11.6) constructon 11.6 (9.6) sale, hotels, restaurants 13.0 (9.5) fnancal ntermedaton 11.8 (9.8) others 11.8 (9.4) Source: own calculatons based on GSOEP, wave 12 (1995) Note: table 8 ncludes the mean tenure of people who are enttled n an occupatonal penson plan, numbers are gven n years, standard devaton n parentheses Furthermore, as columns V and VI of table 10 pont out frm sze s an mportant factor. We fnd out that a small supply n penson plans occurs n small frms. People employed n frms wth 5 to 19 employees have a sgnfcantly lower probablty to partcpate n an occupatonal penson plan and a lower duraton of plans (table 10 and 12) than frms employng 2000 and more employees. Ths can be explaned by scale economes and the fact that large frms are more lkely to have a hstorcal Ths may be determned by the number of observatons n each category and the resultng varance.

20 17 background n plan provson. Furthermore, large frms are more lkely to survve busness cycle fluctuatons and therefore are more wllng to provde a penson plan. The dmenson of payments depends on the duraton of the plan. Thus, we cannot reject hypotheses H 2 and H 3. To control for dfferent atttudes towards savng we ncluded varables that are supposed to capture the current stuaton of lvng, rsks and other savngs motves (see table 11). The estmaton results for 2005 show that H 6 cannot be rejected. People make up dfferent mnds on ther current standard of lvng. The more they are satsfed wth ther economc stuaton and ther standard of lvng the more lkely they are to save for retrement. Contrary to our assumptons the rsk to become unemployed does not have a statstcally sgnfcant mpact. If people expect to have longer lfe expectancy they are more lkely to partcpate n a penson plan. To sum up the results for the duraton of an enttlement (see table 12) we fnd that ndvdual ncome, martal status, age, unversty degree and beng occuped n a small or large frm have a huge mpact on the decson of a person to stay n a penson plan for a longer perod. Dfferences across lnes of busness could not be dentfed clearly. 4. Concluson Currently penson savngs decrease n large companes. Others underestmate the costs arsng from offerng a penson plan to employees. In most cases a decrease s caused by a large poston of pensons n the balance sheet whch causes worse ratngs wth the result of worse condtons for recevng credts. Therefore, ratng agences ask companes to use penson funds or other nstruments to reduce nternal penson savngs. Besdes, companes offer pensons to ther workforce by several ams (e.g. economc reasons, motve of foresght, responsblty for the workforce). Therefore t s mportant to evaluate what people look lke who adopt an occupatonal penson. In ths analyss we have shown that dfferences exst both across SMEs and large frms as well as across regons usng econometrc methods. Our results are supported by other descrptve studes carred out for Germany whch ndcate that a gap between dfferent frm szes occurs. We pont out that penson plan provson dffers across lnes of busness. These results are n lne wth theoretcal and emprcal lterature whch arguments that employees of hgher educaton and those wth low ncome as well as young employees seem to face nformaton problems

21 18 concernng the necessty to provde for old age. Besdes nformaton problems young people mght have a hgher dscount rate than older people and prefer to spend more money on consumpton and do not care for old age. Less provson by sngle adults and people wth chldren could not be dentfed n 1995 but n 2005 people s provson for old age decreases wth the number of chldren. Concernng nformaton dfferences across ndvduals t would be nterestng to nvestgate the effects of ndvdual labor unon membershp. Labor unons are able to collect and pool nformaton and afterwards pass t to ther members. Furthermore they have a larger mpact on wage dscussons as well as frnge benefts. Therefore, labor unons are supposed to have a greater mpact on penson provson as has been shown by other authors for the US. Besdes, the wllngness to partcpate n an occupatonal penson plan could be dependent on ndvdual payments to the statutory penson system. It could thus be expected that those payng small amounts to the statutory system are aware of not recevng large paybacks at retrement age and do hence demand occupatonal plans. Important determnants of savng for retrement are ndvdual atttudes and satsfacton wth the standard of lvng. Ths has been measured by several varables. The more satsfed people are at present the more they want to keep t n future and therefore they care more for retrement. To elmnate the named gaps further research needs to be done on questons concernng the possblty for the state to set ncentves for both employers and employees to adopt a penson scheme and mprove nformaton polcy. Acknowledgements: The author thanks Dors Neuberger, Lucnda Trgo Gamarra, Hans Eggert Remers, Alexander Conrad, and the partcpants n the Economc Research Semnar at the Unversty of Rostock for helpful comments on ths paper. The usual dsclamer apples. The data used n ths publcaton were made avalable to us by the German Soco-Economc Panel Study (GSOEP) at the German Insttute for Economc Research (DIW), Berln and the German SAVE survey by the Mannhem Research Insttute for the Economcs of Agng.

22 I Appendx Table 9: Estmaton results: Determnants of probablty of penson plan supply (1995) company offers an dependent varable occupatonal penson I II III observatons Wald Test ntercept *** *** *** gross ncome *** *** *** (8.63) (9.35) (9.37) lne of busness manufacturng of goods ***0.316 ***0.308 (5.11) (4.86) constructon (-0.78) fnancal ntermedares ***0.902 ***0.895 (7.10) (7.02) frm sze 5 to <20 employees *** *** *** (-7.76) (-6.96) (-6.83) 2000 and more employees ***0.621 ***0.568 ***0.563 (9.67) (8.84) (8.68) Note: ***= sgnfcant at 1%, **= sgnfcant at 5%, *= sgnfcant at 10%, estmaton method: bnary probt, dependent varable: company offers an occupatonal penson to ts employees, z-values n parentheses, results weghted by a person level projecton factor Reference groups: lne of busness: people employed n agraran, mnng, electrcty, sale, hotels, restaurants, and others; frm sze: less than 5 employees + 20 to < 2000 employees Source: own estmaton based on GSOEP (1995), wave 12

23 II Table 10: Estmaton results: Determnants of penson plan provson (1995) I II III IV V VI observatons Wald Test ***15.04 ***23.50 ***26.07 ***24.04 ***30.71 ***33.26 ntercept personal attrbutes gross ncome * * *10-6 (0.76) (0.55) (0.41) (0.17) (0.12) (0.30) male (0.63) (0.60) (0.39) (0.60) (0.50) (0.24) marred (0.05) (0.13) (-0.15) (-0.11) (-0.17) (-0.19) age ***0.018 *** (2.70) (2.62) (0.83) (0.97) (1.02) (0.94) unversty degree * (*) (-1.68) (-1.51) (-0.37) (-0.05) (-0.15) (-0.47) western Germany **0.371 *0.298 * (2.38) (1.80) (1.73) (1.32) (1.38) clams at earler employer **0.631 **0.709 ***0.759 ***0.686 (2.29) (2.41) (2.72) (2.61) tenure **0.025 **0.024 **0.023 **0.025 (2.19) (2.03) (2.03) (2.13) tenure * earler clams ** *-0,041 ** ** (-1.99) (-1.87) (-2.01) (-2.12) lne of busness manufacturng of goods (0.28) (0.10) fnancal ntermedares *0.453 *0.406 (1.86) (1.66) frm sze 5 to <20 employees * * (-1.69) (-1.81) 2000 and more employees (0.40) (0.62) Note: ***= sgnfcant at 1%, **= sgnfcant at 5%, *= sgnfcant at 10%, estmaton method: bnary probt, dependent varable: clams n an occupatonal penson, z-values n parentheses, results weghted by a person level projecton factor Reference groups: lne of busness: people employed n agraran, mnng, electrcty, constructon, sale, hotels, restaurants, and others; frm sze: less than 5 employees and 20 to < 2000 employees Source: own estmaton based on GSOEP (1995), wave 12

24 III Table 11: Estmaton results: Determnants of penson plan provson (2005) I II III IV V VI VII observatons Wald Test ***71.91 ***72.98 ***93.75 ***94.16 *** *** *** ntercept *** *** *** *** *** *** *** personal attrbutes net ncome *** *** *** *** ** ** ** (3.39) (3.09) (3.05) (3.05) (2.55) (2.40) (2.39) male (0.53) (0.44) (0.26) (0.28) (0.24) (0.32) (0.27) marred ***0.437 ***0.413 ***0.432 ***0.435 ***0.389 ***0.385 ***0.385 (4.58) (4.27) (4.52) (4.71) (4.20) (4.15) (4.14) age ***0.098 ***0.101 ***0.107 ***0.108 ***0.112 ***0.110 ***0.109 (4.32) (4.28) (4.64) (4.66) (4.78) (4.75) (4.67) age2 *** *** *** *** *** *** *** (-4.83) (-4.77) (-5.21) (-5.20) (-5.31) (-5.26) (-5.18) unversty degree (0.17) (0.48) (0.47) (0.43) (-0.10) (-0.13) (-0.21) number of chldren ** ** ** ** * * (*) (-2.11) (-1.95) (-2.13) (-2.32) (-1.69) (-1.85) (-1.58) western Germany ***0.256 ***0.254 ***0.247 **0.192 **0.196 **0.202 (2.80) (2.78) (2.68) (2.08) (2.10) (2.18) motves expects to lve longer than average ***0.319 ***0.313 ***0.257 ***0.258 ***0.255 (3.48) (3.42) (2.76) (2.76) (2.74) rsk to become unemployed (-0.96) (-0.26) (-0.31) (-0.21) acceptable standard of lvng **0.042 **0.042 **0.042 (2.05) (2.04) (2.05) satsfacton wth economc stuaton ***0.069 ***0.069 ***0.067 (3.42) (3.37) (3.26) owns a Rester penson (*) (1.54) prvate penson nsurance **0.216 (1.92) Note: ***= sgnfcant at 1%, **= sgnfcant at 5%, *= sgnfcant at 10%, estmaton method: bnary probt, dependent varable: clams n an occupatonal penson, z-values n parentheses, weghted results are based on the dstrbuton of ncome and age of the German sample census Source: own estmaton based on SAVE dataset, wave 2005