RIETI Dscusson Paper Seres 11-E-065 Effect of Penson Reform on Penson-Beneft Expectatons and Savngs Decsons n Japan OKUMURA Tsunao Yokohama Natonal Unversty USUI Emko Nagoya Unversty and IZA The Research Insttute of Economy, Trade and Industry http://www.ret.go.jp/en/
RIETI Dscusson Paper Seres 11-E-065 September 2011 Effect of Penson Reform on Penson-Beneft Expectatons and Savngs Decsons n Japan OKUMURA Tsunao (Yokohama Natonal Unversty) USUI Emko (Nagoya Unversty and IZA) Abstract Usng the Japanese Study of Agng and Retrement (JSTAR), a Japanese panel survey of people age 50 or older, we fnd that many Japanese n ther early 50s compared to those n ther late 50s and early 60s expect ther level of publc penson benefts to declne. We fnd that recent penson reform (whch rased the pensonable age) affected people by ncreasng the age when they expect to clam ther benefts by almost the exact amount for all. Nevertheless, the effect of reform on ther expectatons for future benefts remaned nsgnfcant. We also fnd evdence that anxety about the publc penson program s future nduces people to save more. Keywords: subjectve expectatons, uncertanty, savngs, and penson reform. JEL Classfcaton: E21; H55 RIETI Dscusson Papers Seres ams at wdely dssemnatng research results n the form of professonal papers, thereby stmulatng lvely dscusson. The vews expressed n the papers are solely those of the author(s), and do not represent those of the Research Insttute of Economy, Trade and Industry. For helpful comments receved, we thank Mtsuhde Hoshno, Hdehko Ichmura, Masayuk Morkawa, Jm Smth, and partcpants that attended the meetng held at RIETI on March 23, 2011 and the RIETI-RAND Workshop held on July 30, 2011. 1
1. Introducton In Japan s rapdly agng socety, as a large wave of the baby-boomer generaton s reachng retrement age, the publc penson program faces fnancal challenges that threaten the program s solvency and benefts for future generatons. To cope wth the overall ncrease n benefts and the deteroratng fscal balances, the Japanese government pursued penson reforms that rase the pensonable age from 60 to 65, and ntroduced an automatc adjustment of ther beneft levels to allow the penson to adapt flexbly to demographc and macroeconomc changes. These reforms may affect non-retrees expectatons about the qualty of future publc penson arrangements and, as a consequence, alter ther savngs, asset holdngs, and labor supply decsons. For example, non-retrees may expect further delays n ther pensonable age and/or reductons n ther future beneft level. 1 They may, therefore, decde to delay retrement and save more for retrement. The purpose of ths paper s frst to nvestgate the effect of the penson reforms on the non-retrees expectatons about ther future publc penson. Ths paper also examne whether the loss of confdence n the future publc penson system nduces ndvduals to save more for retrement. In other words, we examne the substtuton effect of changes n the publc penson on prvate savngs. Due to lack of data on penson expectatons, there s almost no emprcal study n Japan that examnes the relatonshps between ndvduals subjectve expectatons of the future publc penson and ther decsons on savngs and how long they plan to work. We utlze the Japanese Study of Agng and Retrement (JSTAR), a new Japanese panel survey that collects nformaton on the economc, socal, and health condtons of people age 50 or older. One of the unque characterstcs of the JSTAR s that the survey questons respondents about ther subjectve 1 Snce the penson beneft and contrbuton schemes are requred by law to be revewed at least every fve years from the vewpont of fnancal balances and ther sustanablty, non-retrees may fear that the scheme may worsen. 2
probablstc expectatons of publc penson benefts. By utlzng ths nformaton, we can measure the degree of uncertanty about publc penson benefts for each respondent. 2 Many studes report that mddle-aged and older adults save excessvely when compared to the amounts gven n the lfe cycle hypothess. Kotlkoff and Summers (1989) and De Nard, French, and Jones (2009, 2010) show that mddle-aged and older adults save a large amount for the followng reasons: to leave a bequest to ther chldren, to pay for expensve medcal care, and to prepare for a longer lfe expectancy. In Japan, n addton to these motves, t has been ponted out by Horoka, Fujsak, Watanabe, and Kouno (2000) and Horoka, Suzuk, and Hatta (2007) that the ndvduals perceptons that publc pensons could become more unrelable are an mportant factors n explanng the hgh savng n Japan. The Economc Survey of Japan (2009) has used cross-country data to fnd a negatve relatonshp between the household savng rate and the proporton of ndvduals reportng confdence about the future of ther publc penson. However, ths predcton has not been tested based on ndvdual data that have as rch demographc and economc nformaton as the JSTAR. Furthermore, recent publc penson reform may weaken the ndvduals confdence n the future publc penson system and affect ther savngs decsons. However, ths reform effect has not been explored yet. 3 The Japanese penson reforms n 1994 and 2000 rased the pensonable age of the Employee Penson Insurance (EPI) gradually from 60 to 65 years of age (see Table 1). The EPI benefts consst of a flat-rate part and a wage-proportonal part. For men, the pensonable age of the flat-rate beneft of the EPI s beng rased by one year every three years from 2001 to 2013 untl the age of 65 s reached. The pensonable age for the wage-proportonal beneft of the EPI 2 Domntz and Mansk (2006) elct the subjectve probablty dstrbuton of benefts from the respondents to measure the uncertanty of penson benefts. 3 Domntz, Mansk, and Henz (2002) smulate how socal securty polcy may affect retrement savngs. Delavande and Rohwedder (2011) estmate the relatonshp between the respondents uncertanty about ther socal securty polcy and ther portfolo choce of rsky and safe assets. 3
wll then be ncreased by one year every three years from 2013 to 2025 untl 65 years of age s reached. From 2005 to 2030, the pensonable age for women wll ncrease n a parallel manner to that of men, but wll lag by 5 years. Snce the gradual ncrease n the pensonable age s based on brth cohorts (two or four consecutve cohorts havng the dentcal pensonable age), we can explot ths polcy varaton to dentfy the effects of penson expectatons on work and savngs decsons. In Secton 2, we revew recent publc penson reform and elderly employment polcy n Japan. Secton 3 descrbes the JSTAR data. Secton 4 presents the dstrbuton of publc penson expectatons condtonal on the brth cohort. Secton 5 examnes the effect of the soco-economc characterstcs on publc penson expectatons and also studes the effect of publc penson reform on penson expectatons. Secton 6 provdes estmates on the effects of publc penson expectatons on savngs decsons. Secton 7 concludes the paper. 2. Overvew of Reform on Publc Penson and Elderly Employment Polcy n Japan The publc old-age penson scheme n Japan s composed of three plans: (1) Natonal Penson Insurance (NPI, Kokumn Nenkn) for self-employed workers and non-employed people; (2) Employees Penson Insurance (EPI, Kose Nenkn) for those employed by prvate busness corporatons; and (3) Mutual Ad Insurance (MAI, Kyosa Nenkn) for those employed n the publc sector and prvate schools. The NPI, EPI, and MAI cover 45.5, 48.0, and 6.5 percent of the populaton nsured by publc penson programs n 2007, respectvely (Osho, Shmzutan, Osh; 2010, 2011). Snce the MAI has almost the same beneft scheme as the EPI, the MAI and the EPI are combned n the JSTAR questonnare. The NPI conssts only of a flat-rate beneft (the so called Old-Age Basc Penson, Rore Kso Nenkn), but the EPI conssts of a two-ter beneft scheme: flat-rate and wage-proportonal benefts. To be elgble to receve the publc penson n Japan, one must pay a monthly premum nto the plan for a mnmum of 25 years. Dependent 4
spouses of the EPI and MAI benefcares are enttled to the flat-rate beneft wthout payng contrbutons f ther ncome s below the mnmum taxable ncome. 4 The pensonable age for the flat-rate NPI s 65 n prncple. 5 On the other hand, the pensonable age for the flat-rate EPI had been set at age 60 snce 1975 for men and snce 2000 for women. However, penson reforms n 1994 and 2000 rased the pensonable age for the EPI s flat-rate and wage-proportonal benefts n stages from age 60 to 65 as presented n Table 1, Panel A. For male EPI benefcares, the pensonable age s rased n the followng two stages: In the frst stage, the pensonable age for the flat-rate EPI has been rased by one year for every three years from 2001 untl 2013 to reach age 65. In the second stage, the pensonable age for the wage-proportonal EPI wll be rased by one year every three years begnnng from 2013, and reachng 65 n 2025. For female EPI benefcares, the pensonable age wll be ncreased wth a lag of 5 years relatve to men:.e., the pensonable age for female EPI benefcares wll be rased from 60 to 65, startng from 2006 untl 2013 for the flat-rate beneft, and from 2018 untl 2030 for the wage-proportonal beneft. To provde stable employment for adults n ther early 60s, who wll no longer be elgble for the flat-rate EPI, the government passed the Employment Measures Law n 2004. Ths Law requres companes to ensure employment up to the pensonable age, and thus oblgates companes to gradually rase the mandatory retrement age or to ntroduce a contnued employment system from age 60 to age 65 by 2013 (as n Table 1, Panel B), or to completely abolsh mandatory retrement. 4 Untl 1986, the employees dependent spouses ether contrbuted voluntarly to the NPI or were smply left unnsured (except for a survvors beneft). The 1986 reform put the dependent spouses under the publc penson umbrella, although they were exempt from contrbutng to publc penson plans. 5 The NPI benefcares can opt to receve penson at any tme after age 60. The penson amount vares by the clamng age; t s reduced when receved before 65 and ncreased when receved after 65. 5
In the followng sectons, we examne whether these polcy reforms have affected the non-retrees expectatons about the publc penson clamng age, beneft level, retrement, and savngs decsons. 3. Data and Descrptve Statstcs Data used n ths study are the Japanese Study of Agng and Retrement (JSTAR), whch s desgned and carred out jontly by the Research Insttute of Economy, Trade, and Industry (RIETI), Htotsubash Unversty, and the Unversty of Tokyo. The JSTAR s Japan s frst globally comparable panel data survey of the elderly. Its desgn s smlar to the U.S. Health and Retrement Study (HRS), the Survey of Health, Ageng and Retrement n Europe (SHARE), and the Englsh Longtudnal Study of Ageng (ELSA). The JSTAR covers a wde range of nformaton ncludng: economc, socal, and health condtons of mddle-aged and older adults. The ndvduals n the baselne sample of JSTAR were between the ages of 50 and 75 n 2007 and lved n fve muncpaltes n eastern Japan: Takkawa Cty n Hokkado; Senda Cty n the Tohoku area; Adach Ward, whch s a specal cty n the center of the Tokyo metropols; Kanazawa Cty n the Hokurku area; and Shrakawa town, whch s located n the mountans of Chubu area. Naha Cty, located n Oknawa, was added to the sample n 2008; and Tosu Cty, located n the Kyushu area, was added n 2009. 6 The sample had 5,800 partcpants n the sample partcpated n the survey and the response rate was close to 60 percent. The second wave of surveys was conducted n 2009 for the baselne sample. A more detaled descrpton of the survey s desgn and sample s contaned n Ichmura, Hashmoto, and Shmzutan (2009). We have restrcted the JSTAR data to respondents between the ages of 50 and 65 who were not currently recevng publc penson benefts. At the tme of the ntervew, 2,355 6 The JSTAR s not a probablstc natonal samplng, but seven ctes selected for probablstc sample for each ste. 6
ndvduals respondng to the frst wave of the JSTAR (5 muncpaltes, Naha, and Tosu) reported that they were not recevng publc penson benefts at the tme of the ntervew, and 2,002 ndvduals ndcated that they would be utlzng publc penson benefts n the future. Of those, 8.41 percent reported that the amount of ther expected future publc penson benefts would not be reduced; 35.05 percent reported that ther expected benefts would be reduced; and the remanng 56.55 percent reported that they dd not know. It appears that many non-retrees are uncertan about the amount of ther expected future publc penson benefts. Table 2 gves the characterstcs of our sample. Among those who plan to receve publc penson benefts n the future, the average age s 56.7, 48.6 percent are female, 85.2 percent are marred, and 84.3 percent are workng for pay. In terms of educaton, 15.5 percent receved less than a hgh-school degree, 45.7 percent receved a hgh-school degree, 18.0 percent attaned a junor college degree, and 20.6 percent earned a unversty degree or greater. 4. Expectatons on Publc Penson Benefts: Emprcal Fndngs In ths secton, we provde our fndngs on the expectatons about the publc penson clamng age, retrement age, future publc penson beneft level, and the drop n the future publc penson beneft level, condtonal on brth cohorts. We fnd that the belefs about the future of the publc penson program demonstrate substantal uncertanty and heterogenety. 4.1. Expectatons about the Publc Penson Clamng Age and the Retrement Age We start by dsplayng the probablty of the expected penson clamng age, condtonal on fscal year brth cohort, separately by the NPI and the EPI benefcares. We examne (1) whether there are notceable dfferences n expected penson clamng age between the younger and the older cohorts, and (2) whether the expected penson clamng age dffers across cohorts wth dfferent pensonable age for the EPI benefcares. We conduct the same analyss for the expected retrement age. 7
Fgure 1 dsplays the probablty of the expected publc penson-clamng age, condtonal on fscal year brth cohorts, separately by gender and by type of penson ndvduals are plannng to receve. 7 The sample ncludes fve muncpaltes n 2007, Naha n 2008, and Tosu n 2009. For men and women plannng to receve the NPI, 26.8 percent of those born after 1952 dd not know about when they would clam ther penson, whle 55.8 percent expected to begn recevng benefts at age 65. For those born before 1951, 14.6 percent dd not know when they would clam ther penson, whle those expectng to receve benefts at age 65 ncreased to 68.6 percent. Therefore, n examnng the older cohorts, we fnd that the percent of those who are undecded declnes and that more ndvduals are plannng to clam ther penson at age 65. On the other hand, for men who plan to receve the EPI (and thus are lkely to be affected by the penson reform), the 1945-1946 cohorts (where age 63 s the elgble age to clam the flat-rate beneft) are most lkely to expect to clam ther penson at age 63. The 1947-1948 cohorts (where age 64 s the elgble age to clam the flat-rate beneft) are most lkely to clam at age 64. Much younger cohorts (whose elgblty for the flat-rate beneft has ncreased to age 65) are a mxture of those who expect to clam ther penson at age 65 and those who do not yet know what ther expected clamng age wll be, whch s much the same as the younger cohorts n the NPI sample. Smlar to the EPI male benefcares, the EPI female benefcares are more lkely to delay clamng ther penson based on the ncrease n the elgblty age for the flat-rate beneft. Thus, the reform appears to have an mpact on the penson-clamng age of both male and female EPI benefcares. We next examne whether ndvduals coordnate ther decsons about when to begn clamng publc penson benefts and when to retre (.e., stop workng). Table 3 looks at the relatonshp between the expected penson-clamng age and the expected retrement age for fve 7 In Japan, the government s fnancal year runs from Aprl to March of the followng year. 8
muncpaltes n 2009, Naha n 2008, and Tosu n 2009. 8 Among men who are currently workng, only 15.0 percent expect to clam ther future penson benefts at about the same tme that they retre. Ths low percentage s partly due to the fact that n Japan 38.9 percent expect to never stop workng. Fgure 2 dsplays the probablty of the age at whch workers expect to retre, condtonal on fscal year brth cohorts for the same sample as n Table 3. For those who plan to receve the NPI, ndvduals n younger cohorts expect to never stop workng, but those who expect to retre after age 65 ncreases n the older cohorts. For those who plan to receve the EPI, ndvduals n younger cohorts are mxed wth those who expect to never stop workng and those who expect to retre age 60. In the older cohorts, the percent of ndvduals who expect to retre at a later age gradually ncreases, although a fracton of those who expect to never retre remans about the same. Hence, the retrement decson appears to be made ndependently of an ndvduals penson elgblty age, and also to be unaffected by penson reform. Ths pattern s clearly seen when we focus on male EPI benefcares who are workng n jobs that have a mandatory retrement polcy. Among ths group, although 54.69 percent expect to clam ther penson at age 65, only 24.6 percent expect to retre at age 65 and 33.8 percent expect to never stop workng. 9 In concluson, the ndvdual s expected age of clamng publc penson benefts appears to be unrelated to ther own and spousal retrement decsons and spousal penson-clamng age. Ths s n contrast to the U.S. and European countres where the penson-clamng age usually concdes wth the clamant s or ther spouse s retrement decsons. Proposals to reform the 8 When asked about the tmng of retrement n the 2007 wave, 39.4 percent reported undecded when to retre. In the 2009 survey, the respondents were gven an addtonal opton: never expect to stop workng, then there was a sgnfcant drop n those who reported undecded. Specfcally, among those who reported undecded n the 2007 wave, 54.85 percent reported never expect to stop workng and 22.3 percent reported undecded n the 2009 wave. Therefore, we restrct the sample to those who responded to the revsed JSTAR questonnare for expected retrement age. 9 As for the relatonshp between the expected penson-clamng ages reported by both husbands and wves of the JSTAR sample, about half of husbands and wves expect to clam ther penson when they are age 65. Only 14.0 percent of husbands and wves expect to clam ther pensons at the same tme. 9
publc penson e.g., rase the full retrement age have been found to encourage workers to delay ther retrement n the U.S. (De Nard, French, and Jones, 2010), but ths polcy reform may not have as large an effect on the retrement decsons n Japan as n the U.S., due to the fact that older Japanese workers already have qute a hgh work ncentves. 4.2. The Expected Publc Penson Beneft Level Next, we next dsplay the probablty of the expected publc penson beneft level. For ths, we have nformaton on the expected amount of publc penson beneft and the expected rato of the publc penson beneft to pre-retrement earnngs (replacement rate). Fgure 3 dsplays the probablty of the expected amount of the publc penson beneft, condtonal on fscal year brth cohorts, graphed separately by gender and by type of penson that s planned to be receved. When we compare ndvduals who report don t know wth those who report knowng the expected amount, those who report don t know are, on average: 2.25 years younger, 3.8 percent more lkely to be female, and 4.3 percent more lkely to be labor force partcpants. However, there s no notceable dfference n the educatonal level. In Fgure 3, the medan of the expected future publc penson beneft for men (women) s 720,000 (630,000) yen for the NPI benefcares and 1,800,000 (840,000) yen for the EPI benefcares. The nterquartle range of the dstrbuton for men (women) s 400,000 (330,000) yen for the NPI benefcares and 1,400,000 (852,000) yen for the EPI benefcares. Therefore, the expected beneft level s greater and more wdely dstrbuted for EPI benefcares than for NPI benefcares. For both the NPI and EPI benefcares, the medan of the dstrbuton s about the same for younger and older cohorts, but the nterquartle range of the dstrbuton s larger for younger cohorts than for older cohorts. Ths suggests a greater uncertanty n the amount of the expected beneft level n the younger cohorts. Through the penson reform, there s no sgnfcant dfference n the expected 10
beneft level for the EPI benefcares among the cohorts to whch the unform pensonable ages apply. Snce the wage-proportonal part of the publc penson benefts s proportonal to the pre-retrement earnngs, the heterogenety n the expected publc penson benefts for the EPI benefcares presumably reflects not only the heterogenety n the ndvduals expectatons about the future structure of the penson system (whch s subject to poltcal reform), but also reflects heterogenety n the ndvduals labor earnngs to whch the benefts are ted. To extract the former heterogenety, we utlze the questonnare n JSTAR, whch asks about the expected rato of the publc penson beneft to pre-retrement earnngs (replacement rate, see E-033 n the Appendx). Fgure 4 dsplays the probablty of the expected rato of the publc penson beneft to pre-retrement earnngs, condtonal on fscal year brth cohorts, separately by gender and by type of penson an ndvdual s plannng to receve. The majorty of the younger male EPI benefcares do not know ther expected replacement rate, but the don t know responses gradually declne n the older cohorts. The percent of ndvduals who report ther expected replacement rate to be only 10-30 percent ncreases by more than 10 percent for the 1947-1948 cohorts of male EPI benefcares (whose pensonable age for the wage-proportonal beneft s 60) when compared wth the 1949-1952 cohorts (whose correspondng pensonable age s 61). 4.3. The Expected Drop n Future Publc Penson Benefts The JSTAR has a unque queston that drectly asks the respondents about the subjectve uncertanty of ther future publc penson beneft levels. Specfcally, JSTAR nqures about the lkelhood that the beneft level expected to be receved could be reduced by 10 percent or more n the future (see E034 n Appendx). Fgure 5 dsplays the responses to the expected probablty of a more-than-10 percent decrease n the publc penson beneft level, condtonal on fscal year brth cohorts, and lsted separately by gender and by type of penson to be receved. Overall, 56.1 percent report that they do not know whether the publc penson beneft wll drop; and those who 11
report don t know are 3.5 percent more lkely to be female and slghtly less lkely to be educated, but dffer lttle n other demographc characterstcs. The fracton of the respondents who report 0 percent chance of reducton n future penson beneft level s hgher n the older cohorts than n the younger cohorts, whle that of the respondents who report more-than-a 50 percent chance of a 10 percent reducton s lower. Therefore, older cohorts tend to be less pessmstc than younger cohorts about the contnuaton of the structure of the publc penson system. There appear to be no sgnfcant dfferences n the expected beneft level for the EPI benefcares among the consecutve cohorts to whch the same pensonable ages apply through the penson reform. In concluson, there s substantal uncertanty and heterogenety n belefs about the future of the publc penson program. Younger cohorts are less confdent than older cohorts about future publc penson benefts. They report greater probablty of declne n future publc penson benefts and more provde don t know responses about ther future publc penson expectatons. There are also dfferences between the NPI and the EPI benefcares n ther expectatons on publc pensons. The EPI benefcares are affected by the penson reform and they have more of a pcture of the future of the publc penson program. 5. Publc Penson Expectatons and Penson Reform In ths secton, we use regresson framework to (1) analyze the determnants of the publc penson expectatons and retrement expectatons and (2) assess the effect of the penson reform on the publc penson expectatons and retrement expectatons. 5.1. Expectatons on Publc Penson Benefts: Estmaton Results In Secton 5.1, we analyze the relatonshp between publc penson expectatons and soco-economc varables. Table 4 presents estmates of the effect of soco-economc characterstcs on varous measures of publc penson expectatons. These characterstcs are: age, 12
gender, martal status, educaton, labor market experence, work status, health status, asset, ncome, chance to survve untl age 75 dvded by lfe table probablty, cty, and calendar years. When the dependent varable s contnuous, the OLS estmates are shown. When the dependent varable s bnary, the probt estmates of margnal effects evaluated at the mean value of the varable are shown. The sample ncludes fve muncpaltes n 2007 and 2009, Naha n 2008, and Tosu n 2009, f not stated otherwse. In Table 4, Column 1, we provde the estmated effects on the plan to receve publc pensons. We fnd that age s not an mportant predctor of the lkelhood of recevng a publc penson n the future. To be elgble to receve the publc penson n Japan, one must pay a monthly premum nto the plan for a mnmum of 25 years, so the labor market partcpants and those who have more than a hgh school educaton (and thus have better employment potental) are more lkely to have a greater probablty of recevng a publc penson n the future. There s no sgnfcant dfference between men and women n the probablty of recevng a publc penson. Ths may be because as dependent spouses, women have the rght to receve an allowance added to ther husbands pensons. Those who expect a greater chance of survvng untl age 75 are lkely to be prepared to receve publc penson benefts n the future. In Table 4, Column 2, we present the estmated effect on the expected penson-clamng age. Indvduals who are more educated (junor college or unversty and over) are more lkely to expect to clam ther publc penson at an older age, whle those who are n ther late 50s (relatve to those n early 50s), female, and those who are less healthy are more lkely to expect to clam penson at a younger age. In Table 4, Column 3, we present the estmated effect on the expected retrement age. Indvduals n ther late 50s and early 60s (relatve to those n early 50s) are more lkely to expect to retre at a later age. Indvduals n ther late 50s plan to clam ther publc penson at a younger age, but they expect to stop workng at a later age, ndcatng that for ths group the publc penson benefts are lkely to be nsuffcent to mantan ther lvng standards. 13
Those who have less employment potental (less ncome and less than hgh-school educaton) expect to retre at a later age. Also, those who expect a greater chance of survvng untl age 75 expect to retre at a later age. In Table 4, Columns 4 and 5, the estmated effects on the expected amount of the publc penson beneft and the expected replacement rato are presented, respectvely. We fnd that ndvduals n ther late 60s (relatve to those n early 50s) expect to receve 69,740 yen more n publc penson benefts, but the expected replacement rato does not vary much by age. In Table 4, Column 6, we present the estmated effect on the expected probablty of more-than-10 percent drop n the future publc penson benefts. As the queston s only asked of the respondents and not ther spouses, the sample sze drops to 1148. The coeffcent on those ages 60-65 s negatve, suggestng that (1) ndvduals n ther early 60s are much more confdent of the contnuaton of the publc penson system than those n ther early 50s, and (2) ndvduals who are closer to clamng ther publc penson are able to perform more accurate penson calculatons (Gustman and Stenmeer, 2005). Labor market partcpants are more lkely to expect a drop n ther future publc penson benefts, as they may choose to work to guard aganst the declne n ther future publc penson. 5.2. The Effect of Penson Reforms on Expectatons of Publc Penson Benefts In ths secton, we evaluate the effect of penson reforms on ndvdual expectatons by utlzng a regresson dscontnuty desgn that explots month of brth cutoff to determne the ndvduals pensonable ages. We examne whether there was a dscrete change n publc penson expectatons on ether sde of the fxed threshold whch s Aprl 2 of those brth years affected by the reform (see Table 1, Panel A). That s, n a partcular reform year, the pensonable age for the EPI benefcares whose brth date s before Aprl 1 s shorter by one year than those whose brth date s later than Aprl 2. In contrast, n the year when there s no reform, the pensonable age wll reman the same 14
for all ndvduals, regardless of ther brth date. Thus, we can conduct a regresson dscontnuty analyss to estmate the effect of penson reform on publc penson expectatons, by nteractng a dummy for plannng to receve Employee Penson Insurance ( EPI ), a dummy for the reform year ( REFORM ), and a dummy for brth date later than Aprl 2 ( BIRTH estmate the followng regresson: 2 ). Specfcally, we y X EPI BIRTH 2 BIRTH 2 EPI 1 2 3 4 REFORM REFORM EPI REFORM BIRTH 2 5 6 7 REFORM BIRTH 2 EPI 1 (1) where y s the measures of publc penson expectatons, and X s the covarates (soco-economc varables) as used n Table 4. The coeffcent 1 reflects the reform effect on the publc penson expectatons. 10 Note that n ths analyss we assume that people do not change benefcares between the EPI and the NPI as a consequence of the reform. Table 5 presents the estmaton results. When we estmate Equaton (1), we do not smply compare those born before Aprl 2 wth those born after Aprl 2. Instead, we keep those born before Aprl 2 as the reference category, but separate those born after Aprl 2 to: (1) those born between Aprl 2 to June 30 (2 nd quarter), (2) those born between July 1 to September 31 (3 rd quarter), and (3) those born between October 1 to December 31 (4 th quarter). Thus, BIRTH 2 s composed of 2 nd quarter, 3 rd quarter, and 4 th quarter dummes. In Table 5, Column 1, we fnd that penson reform nduces people to rase ther expected penson-clamng age by 1.170 (.457) years for those born n the 2 nd quarter (relatve to those born n the 1 st quarter who are unaffected by the reform), by 1.178 (.451) years for those born n the 3 rd quarter, and by 1.068 (.445) years for those born n the 4 th quarter. Therefore, the EPI 10 Ths framework s smlar to Bottazz et al. (2006) whch studes how expected retrement age and expected replacement rate have been affected by the Italan penson reform. 15
benefcares are fully nformed about the reform, and those affected by the reform respond by expectng to delay ther penson clamng age by exactly one year. In Table 5, Column 2, the effect of penson reform on expected retrement age s reported. The penson reform rases the retrement age by.430 (1.759) years for those born n the 2 nd quarter, -.996 (1.520) years for those born n the 3 rd quarter, and.0145 (1.346) years for those born n the 4 th quarter. The reform effect on retrement age s small and nsgnfcant, and t appears that retrement decson s ndependent of penson reform. Accordng to the estmaton results n Table 5 Column 3, people expect that due to the penson reform ther future publc penson beneft wll drop by 280,000 yen per year for those born n the 2 nd quarter, drop by 620,000 yen per year for those born n the 3 rd quarter, but ncrease by 240,000 yen per year for those born n the 4 th quarter. The reform effect on those born n the 3 rd quarter s especally large n magntude, amountng to a nearly 50 percent drop n ther expected publc penson beneft (as the medan of the expected publc penson beneft for the EPI benefcares s 1,200,000 yen). However, these estmates are nosy, and the sgn for those born n the 4 th quarter goes the opposte drecton. Next, the reform effect on the expected replacement rato s reported n Table 5, Column 4. We fnd a postve but small and nsgnfcant reform effect, whch s nconsstent wth the negatve effect on penson beneft level n Table 5, Column 3. We conclude that the reform has notceable effect on rasng penson clamng age, but the results are nconclusve regardng the effect of penson reform on the expected publc penson beneft levels. 6. The Effects of Expectatons of Penson Benefts on Savngs We examne how ndvduals expectatons about the future of the publc penson beneft affect ther savngs decsons. 16
Our emprcal specfcaton s: PS y X, (2) where PS s the prvate savng varable, y s the measures on publc penson expectatons, X s the covarates as used n Table 4, and s the error term. The key coeffcent of nterest s, whch reflects the effect of publc penson expectatons on prvate savngs. The above specfcaton s based on the model whch reflects the relatonshp between publc penson-beneft wealth (PBW ) and prvate savngs (PS ): PS PBW X. (3) where 0 because of the substtuton effect of publc penson wealth (PBW ) on prvate savngs (PS ). We defne the publc penson wealth (PBW ) as follows: the present value of the expected future publc penson benefts that wll be receved from the expected penson clamng age to the maxmum length of lfe where the dscount factor s composed of survval probablty and nterest rate. That s: N T 1 PBW E,, N 1 r (4) where s the planned amount of benefts, N s the expected penson clamng age, T s the maxmum length of lfe, s the survval probablty, and r s the real nterest rate. The varables, N, and can be obtaned from the JSTAR, as can be substtuted as the expected amount of publc penson beneft (or the expected replacement rato), N as the expected penson clamng age, and as the expected chance to survve. 17
The ndvduals expectatons of benefts amounts E can be calculated as follows. Let p denote the subjectve probablty that the publc penson beneft could be reduced by more-than-10 percent n the future, whch s also avalable n the JSTAR questonnare (E034 n Appendx). We then can show that the expected amount of benefts E has a lower bound of 0.9 1 p and an upper bound of 1 0.1p, because: 0, 0.9 0.9, 1 0.9 1 p, 1 0.1 p. E p p (5) In Equatons (4) and (5), t s straghtforward to argue that the planned publc penson postvely affects PBW, but the expected penson clamng age N negatvely affects PBW. In addton, the upper and lower bounds of the expected amount of benefts E s negatvely affected by p. Therefore, for smplcty, let us lnearly approxmate Equatons (4) and (5) as: PBW N p u (6) 0 1 2 3 X. where s ncluded n X. We predct that 1 0, 2 0, and 3 0. Next, plug Equaton (6) nto Equaton (3) and obtan: PS N p X (7) 1 2 3. Note that snce 0, the expected sgns of the coeffcents are: 1 0, 2 0, and 3 0. Equaton (7) equals to Equaton (2) when y, N, p. In Table 6, we estmate the reduced form prvate savngs equaton (Equaton (7)). The dependent varable PS s a logarthm of the respondents fnal savng goals whch can be 18
obtaned from the JSTAR. 11 The covarates X are the same as those n Table 4. The OLS estmates are reported n the frst three columns of Table 6. In the specfcaton n Column 1, whch ncludes only p as the measure on publc penson expectatons, the effect of the probablty of expectng a more-than-10 percent drop n future beneft s sgnfcantly postve. Ths mples that that those who expect a declne n future publc penson benefts plan to save more. The specfcaton n Column 3 ncludes the varables (expected replacement rato), N, and p as the measures of publc penson expectatons. A 10 percentage declne n the expected probablty of more-than-10 percent declne n penson beneft rases the prvate savngs goal by 7.9 percent, although the effect s nsgnfcant. Those who report that they do not know whether ther future benefts would declne save sgnfcantly more. Although effects are nsgnfcant, those who plan to have a larger replacement rato save less, as do those who expect to clam ther penson later. Therefore, the sgns on the coeffcents p and are as we hypotheszed n Equaton (7), but the sgn on the coeffcent N s not n the expected drecton. The OLS estmates may be based f the expectaton about ther publc penson clamng age and the error term of prvate savng equaton are correlated. In partcular, ndvduals who opt to receve pensons after ther pensonable ages may plan to retre later and thus save less. 12 To solve such an endogenety problem, we use penson reform as a source of exogenous varaton of the expected penson clamng age. Specfcally, we use as nstruments: the nteracton between a 11 There are two reasons as to why we use the respondents fnal savngs goals to represent ther prvate savngs (PS) n Equaton (7). Frst, both the fnal savngs goal and the prvate penson wealth (PBW) are the expected values. Second, whle the PBW s the present value at the penson clamng age, the fnal savngs goal s usually consdered to be the amount the respondents are plannng to save by ther penson clamng age; thus, n Equaton (3), we assume that the PS and the PBW are measured at about the same tme. 12 Snce the beneft formula for the wage-proportonal part s based on the average total earnngs and the length of nsured (workng) perod, the respondents who expect a larger amount of future benefts may have a larger savngs because they obtaned greater earnngs. Therefore, we control earnngs as covarates n Equaton (7). 19
dummy for plannng to receve the Employer Penson Insurance ( EPI ), a dummy for the reform year ( REFORM ), and a dummy for brth date later than Aprl 1 ( BIRTH 2 ), whch are used as ndependent varables n Equaton (1). 13 Note that penson reform has a postve mpact n delayng the expected penson clamng age, and thus can be used as nstruments for the expected penson clamng age. In the last two columns n Table 6, we report the IV estmates. In Column 5, the sgn of coeffcent of penson clamng age s now the same as that predcted n Equaton (7), although all the effects are nsgnfcant except among those ndvduals who report don t know, where they have a postve and sgnfcant effect on savngs. We obtan smlar results when we use the amount of expected publc penson beneft to represent. Therefore, we conclude that those who don t know whether ther future benefts would declne would sgnfcantly ncrease ther fnal savngs goal. We only fnd weak evdence that (1) a reported drop n expectatons about the future publc penson beneft has a postve effect on the ndvdual s prvate savngs goal, (2) a decrease n penson beneft levels has postve effect on prvate savngs goal, and (3) an ncrease n penson clamng age has postve effect on prvate savngs goal. 7. Concluson Usng the Japanese Study of Agng and Retrement (JSTAR), we fnd that many Japanese n ther early 50s do not beleve that publc penson benefts wll be there for them when they retre. We also fnd that the anxety about the publc penson program s future mght nduce people to save more. Restorng balance to the publc penson could reduce the long-term defct and make 13 We assume that the actual penson beneft changes exogenously, and that the EPI benefcares dd not swtch jobs nor become NPI benefcares to offset the mpact of the penson reform on ther savngs. Bottazz et al. (2008) estmate the substtuton effect of penson wealth on prvate wealth by usng dummes for post Italan penson reform perod and employment groups as nstruments. 20
people feel more secure about ther retrement. Therefore, t s possble that elmnatng the publc penson shortfall could reduce the msplaced fear that causes Japanese to over-save. 21
Appendx: Questonnare regardng Penson Expectatons Frst wave, 2007, 5 Ctes (1) E-028. In the future do you expect to receve any further publc penson? 1. Yes 2. No 3. Don t know 4. Refused to answer If answer to E-028 was 1, go to E-030; otherwse, go to E-029-1 (Benchmark: spouse present). (2) E-030. What type of penson s t? Please select the most approprate choce from the followng. 1. Natonal Penson Plan (basc old-age penson) 2. Old age welfare annuty or retrement mutual penson (ncludng basc penson) 3. Survvor s penson 4. Dsablty penson 5. Don t know 6. Refused to answer Regardless of answer, go to E-031. (3) E-031. At what age do you expect to begn recevng that penson? 1. Age: 2. Don t know 3. Refused to answer Regardless of answer, go to E-032. (4) E-032. About how much do you expect to receve per year before taxes? If you don t mnd, please tell me the total amount that wll be deposted to your bank or postal account. 1. Approxmately yen 2. Don t know 3. Refused to answer Regardless of answer, go to E-033. (5) E-033. About what percent of your last workng salary does that amount to? If you dd not work before, please say so. 1. % 2. Dd not work 3. Don t know 4. Refused to answer Regardless of answer, go to E-034. (6) E-034. Do you thnk t s lkely that the amount you expect to receve could be reduced by 10% or more n the future? If you don t thnk ths s possble, please say so. 22
1. % 2. No possblty 3. Don t know 4. Refused to answer Regardless of answer, go to E-035-1. The respondents are questoned: (1), (2), (3), and (4) regardng ther spouses penson expectatons. Reference Bernhem, B. Douglas. 1998. Fnancal Illteracy, Educaton, and Retrement Savng. In Lvng wth Defned Contrbuton Pensons, eds. Olva S. Mtchell and Sylvester J. Scheber, 38-68., Phladelpha: Unversty of Pennsylvana Press. Bottazz, Renata; Tullo, Jappell;, and Maro Padula. 2006. Retrement Expectatons, Penson Reforms, and Ther Effect on Prvate Wealth Accumulaton. Journal of Publc Economcs 90(12): 2187-212. Campbell, John Y., João F. Cocco, Francsco J. Gomes, and Pascal J. Maenhout. 2001. Investng Retrement Wealth: A Lfe-Cycle Model. In Rsk Aspects of Investment-Based Socal Securty Reform, eds. John Y. Campbell and Martn Feldsten. Chcago: Unversty of Chcago Press, NBER Books: 439-82. Delavande, Adelne, and Susann Rohwedder. 2011. Indvduals uncertanty about future socal securty benefts and portfolo choce. Journal of Appled Econometrcs, 26(3): 489-519. De Nard, Maracrstna, Erc French, and John B. Jones. 2009. Lfe Expectancy and Old Age Savngs. Amercan Economc Revew Paper and Proceedngs 99(2): 110-115. De Nard, Maracrstna, Erc French, and John B. Jones. 2010. Why Do the Elderly Save? The Role of Medcal Expenses. Journal of Poltcal Economy 118(1): 39-75. Domntz, Jeff., Charles F. Mansk, and Jordan Henz. 2002. Socal Securty Expectatons and Retrement Savngs Decsons, NBER Workng Paper No. 8718. Domntz, Jeff., and Charles F. Mansk. 2006. Measurng Penson-beneft Expectatons Probablstcally. Labour 20(2): 201-236. Feldsten, Martn., Elena Ranguelova, and Andrew Samwck. 2001. The Transton to Investment-Based Socal Securty When Portfolo Returns and Captal Proftablty Are Uncertan. In Rsk Aspects of Investment-Based Socal Securty Reform, eds. John Y. Campbell and Martn Feldsten. Chcago: Unversty of Chcago Press, NBER Books:41-90. French, Erc, and John B. Jones. 2011. The Effects of Health Insurance and Self-Insurance on Retrement Behavor. Econometrca 79 (3): 693-732. 23
Gustman Alan L. and Thomas Stenmeer. 1999. Effects of Pensons on Savngs: Analyss wth Data From the Health and Retrement Study. Carnege-Rochester Conference Seres 50: 271-326. Gustman Alan L. and Thomas Stenmeer. 2001. How Effectve Is Redstrbuton Under The Socal Securty Beneft Formula? Journal of Publc Economcs 82(1): 1-28. Gustman Alan L. and Thomas Stenmeer. 2004. What People Don t Know About Ther Pensons and Socal Securty. In Prvate Pensons and Publc Polces, eds. Wllam G. Gale, John B. Shoven and Mark J. Warshawsky. Washngton, D.C., Brookngs Insttuton: 57-125. Horoka, Carles Y., Wataru Suzuk, and Tatsuo Hatta. 2007. Agng, Savng, and Publc Pensons n Japan, NBER Workng Paper No. 13273. Horoka, Charles Y., Hdek Fujsak, Wako Watanabe and Takatsugu Kouno. 2000. Are Amercans More Altrustc Than the Japanese? A U.S. Japan Comparson of Savng and Bequest Motves. Internatonal Economc Journal 14(1): 1-31. Ichmura Hdehko, Hdek Hashmoto, and Satosh Shmzutan. 2009. Japanese Study of Agng and Retrement (JSTAR) Frst Results 2009 Report. RIETI Dscusson Paper Seres 09-E-047. McHale, John. 2001. The Rsk of Socal Securty Beneft-Rule Changes: Some Internatonal Evdence. In Rsk Aspects of Investment-Based Socal Securty Reform, eds. John Y. Campbell and Martn Feldsten. Chcago: Unversty of Chcago Press. NBER Books: 247-90. Mansk, Charles F. 2004. Measurng Expectatons. Econometrca 72(5): 1392-76. Osho, Takash, Satosh Shmzutan, and Akko Sato Osh. 2010. Does Socal Securty Induce Wthdrawal of the Old from the Labor Force? In Socal Securty Programs and Retrement Around the World: The Relatonshp to Youth Employment, eds. Jonathan Gruber and Davd A. Wse.Chcago: Unversty of Chcago Press: 217-41. Osho, Takash, Akko Osh and Satosh Shmzutan. 2011. Socal Securty Reforms and Labor Force Partcpaton of the Elderly n Japan. Japanese Economc Revew 62(2): 248 71. 24
Panel A Sample: Men who Plan to Receve Natonal Penson Insurance (NPI), 5 Ctes 2007, Naha 2008, and Tosu 2009, Panel B Sample: Men who Plan to Receve Employee Penson Insurance (EPI), 5 Ctes 2007, Naha 2008, and Tosu 2009 Fgure 1: Probablty of Expected Penson Clamng Age, Condtonal on Brth Cohort 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 Probablty 66 70 65 64 63 61 62 60 65 60 63 Don't Know 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 Probablty 66 72 65 64 63 61 62 60 63 64 65 Don't Know Brth Cohort 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 n 2007 (65,63) (65,62) (65,61) (65, 60) (64,60) (63,60) (62,60) Polcy Reform (flat rate, wage proportonal beneft) Brth Cohort 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 n 2007 Panel C Sample: Women who Plan to Receve Natonal Penson Insurance (NPI), 5 Ctes 2007, Naha 2008, and Tosu 2009 Panel D Sample: Women who Plan to Receve Employee Penson Insurance (EPI), 5 Ctes 2007, Naha 2008, and Tosu 2009 Fgure 1: Probablty of Expected Penson Clamng Age, Condtonal on Brth Cohort 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 Probablty 66 70 65 64 63 61 62 60 Don't Know 65 60 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 Probablty 66 70 65 64 63 61 62 60 60 61 62 65 Don't Know Brth Cohort 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 n 2007 (65,60) (64,60) (63, 60) (62,60) (61,60) (60,60) Polcy Reform (flat rate, wage proportonal beneft) Brth Cohort 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 n 2007
Fgure 2: Probablty of Expected Retrement Age, Condtonal on Brth Cohort Panel A Sample: Men, Plan to Receve Natonal Penson Insurance: 5 Ctes 2009, Naha 2008, and Tosu 2009 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1957 1956 1955 1954 1953 1952 1951 1950 1949 1948 1947 1946 Brth Cohort 52 53 54 55 56 57 58 59 60 61 62 63 Age n 2009 Panel B Sample: Men, Plan to Receve Employee Penson Insurance: 5 Ctes 2009, Naha 2008, and Tosu 2009 Probablty Undecded Never Stop Workng 65 70 100 Never 70-100 66-69 65 61-64 60 50-59 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1957 1956 1955 1954 1953 1952 1951 1950 1949 1948 1947 1946 Probablty Undecded 60 Never Stop Workng 65 61 64 Never 70 100 66 69 65 61 64 60 50 59 Brth Cohort 52 53 54 55 56 57 58 59 60 61 62 63 Age n 2009 (65,63) (65,62) (65,61) (65, 60) (64,60) (63,60) Polcy Reform Fgure 2: Probablty of Expected Retrement Age, Condtonal on Brth Cohort Panel C Sample: Women, Plan to Receve Natonal Penson Insurance: 5 Ctes 2009, Naha 2008, and Tosu 2009 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 61 64 60 0 1957 1956 1955 1954 1953 1952 1951 1950 1949 1948 1947 1946 Brth Cohort 52 53 54 55 56 57 58 59 60 61 62 63 Age n 2009 Panel D Sample: Women, Plan to Receve Employee Penson Insurance: 5 Ctes 2009, Naha 2008, and Tosu 2009 Probablty Undecded Never Stop Workng 65 Never 70-100 66-69 65 61-64 60 50-59 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1957 1956 1955 1954 1953 1952 1951 1950 1949 1948 1947 1946 Probablty Undecded Never Stop Workng 65 60 50 59 61 64 Never 71-100 66-69 65 61-64 60 50-59 Brth Cohort 52 53 54 55 56 57 58 59 60 61 62 63 Age n 2009 (65,60) (64,60) (63,60) (62, 60) (61,60) Polcy Reform
Fgure 3: Quantles of Expected Amount of Publc Penson Beneft, Condtonal on Brth Cohort Panel A Sample: Men who Plan to Receve Natonal Penson Insurance, 5 Ctes 2007, Naha 2008, and Tosu 2009 300 250 200 150 100 50 0 1954 1953 1952 1951 1950 1949 1948 1947 1946 1945 1944 1943 1942 Brth Cohort 53 54 55 56 57 58 59 60 61 62 63 64 65 n 2007 Panel B Sample: Men who Plan to Receve Employee Penson Insurance, 5 Ctes 2007, Naha 2008, and Tosu 2009 10,000 Yen.75 quantle.50 quantle.25 quantle 300 250 200 150 100 50 0 1956 1955 1954 1953 1952 1951 1950 1949 1948 1947 1946 1945 1944 1943 1942 Brth Cohort 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 n 2007 (65,62) (65,61) (65, 60) (64,60) (63,60) (62,60) (61,60) Polcy Reform (flat rate, wage proportonal beneft) 10,000 Yen.75 quantle.50 quantle.25 quantle Fgure 3: Quantles of Expected Amount of Publc Penson Beneft, Condtonal on Brth Cohort Panel C Sample: Women who Plan to Receve Natonal Penson Insurance, 5 Ctes 2007, Naha 2008, and Tosu 2009 200 180 160 140 120 100 80 60 40 20 0 1956 1955 1954 1953 1952 1951 1950 1949 1948 1947 1946 1945 1944 1943 1942 Brth Cohort 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 n 2007 Panel D Sample: Women who Plan to Receve Employee Penson Insurance, 5 Ctes 2007, Naha 2008, and Tosu 2009 10,000Yen.75 quantle.50 quantle.25 quantle 200 180 160 140 120 100 80 60 40 20 0 1956 1955 1954 1953 1952 1951 1950 1949 1948 1947 1946 1945 Brth Cohort 10,000 Yen.75 quantle.50 quantle.25 quantle 51 52 53 54 55 56 57 58 59 60 61 62 n 2007 (65,60) (64,60) (63, 60) (62,60) (61,60) (60,60) Polcy Reform (flat rate, wage proportonal beneft)
Fgure 4: Probablty of Expected Rato of Publc Penson Benefts to Pre Retrement Earnngs, Condtonal on Brth Cohort Panel A Sample: Men who Plan to Receve Natonal Penson Insurance, 5 Ctes 2007, Naha 2008, and Tosu 2009 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 10% 0 1956 1955 1954 1953 1952 1951 1950 1949 1948 1947 1946 1945 1944 1943 1942 Brth Cohort 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 n 2007 Panel B Sample: Men who Plan to Receve Employee Penson Insurance, 5 Ctes 2007, Naha 2008, and Tosu 2009 Probablty Don't Know 50% 100% 50% 20% 10% 1 0.9 0.8 0.7 0.6 0.5 0.4 50% 0.3 0.2 0.1 0 1956 1955 1954 1953 1952 1951 1950 1949 1948 1947 1946 1945 1944 1943 1942 Brth Cohort 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 n 2007 (65,62) (65,61) (65,60) (64, 60) (63,60) (62,60) (61,60)Polcy Reform (flat rate, wage proportonal beneft) Probablty Don't Know 100% 50% 30% 10% 30% Fgure 4: Probablty of Expected Rato of Publc Penson Benefts to Pre Retrement Earnngs, Condtonal on Brth Cohort Panel C Sample: Women who Plan to Receve Natonal Penson Insurance, 5 Ctes 2007, Naha 2008, and Tosu 2009 Panel D Sample: Women who Plan to Receve Employee Penson Insurance, 5 Ctes 2007, Naha 2008, and Tosu 2009 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1957 1956 1955 1954 1953 1952 1951 1950 1949 1948 1947 1946 Probablty 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1957 1956 1955 1954 1953 1952 1951 1950 1949 1948 1947 1946 Brth Cohort 50 51 52 53 54 55 56 57 58 59 60 61 n 2007 Don't Know 100% 50% 30% 10% Probablty Don't Know 100% 50% 30% 10% Brth Cohort 50 51 52 53 54 55 56 57 58 59 60 61 n 2007 (65,60) (64,60) (63, 60) (62,60) (61,60) Polcy Reform (flat rate, wage proportonal beneft)