Fnancal Instablty and Lfe Insurance Demand + Mahto Okura * Norhro Kasuga ** Abstract Ths paper estmates prvate lfe nsurance and Kampo demand functons usng household-level data provded by the Postal Servces Research Insttute. The results show that ncome, chldren, penson and knowledge factors have a sgnfcant effect on the decson as to whether each household purchases lfe nsurance products. The amount of ncome and fnancal assets also appear to have sgnfcant effect on the purchase of prvate lfe nsurance and Kampo. However, penson and bankruptcy experence appear only to have an mpact on Kampo, whle aged (less than 40) and occupaton (cvl servant) factors affect only prvate lfe nsurance. Dummy varables representng comparson, knowledge, and bankruptcy experence dd not have any sgnfcant effect on decsons concernng prvate lfe nsurance. Smultaneous estmatons are also used to examne why households that already have one type of lfe nsurance product (e.g. prvate lfe nsurance) purchase the other type of lfe nsurance product (e.g. Kampo). The results ndcate that ncome, chldren, and bankruptcy experence varables are not sgnfcant factors when households wth prvate lfe nsurance product decde to purchase addtonal Kampo. The results also show that a knowledge dummy has a negatve mpact on addtonal purchases. 1. Introducton Accordng to the Japan Insttute of Lfe Insurance s Annual Report, total nsurance money held by all Japanese nsurance companes, ncludng ndvdual nsurance and annutes, reached approxmately 1,200 trllon yen (about 11,702 bllon US dollars) n FY 2003. Ths amounts to nearly 24,520 thousand yen per household (about 234,798 US dollars) 1. Snce ths represents a sgnfcant share of household total expendture, lfe nsurance s an mportant tem when we address problems wth Japanese household expendture. A number of researchers have nvestgated the relatonshps between lfe nsurance demand and factors such as household ncome, age, workng style, the number of chldren, and so forth n order to shed lght on the demand structure of lfe nsurance products n Japan. For example, Tachbanak and Shmono (1994) found that the total amount of fnancal assets, whether a spouse was workng, + The order of author names was determned by the toss of a con. The authors gratefully acknowledge the fnancal support of the Postal Lfe Insurance Foundaton of Japan and the Postal Servces Research Insttute for provdng the requste data. * Assocate Professor, Faculty of Economcs, Nagasak Unversty. Emal:okura@net.nagsak-u.ac.jp ** Assocate Professor, Faculty of Economcs, Nagasak Unversty. Emal:n-kasuga@net.nagasak-u.ac.jp 1 The number of households s from Japan s 2004 Basc Regster. 1
and the number of dependents (exceptng the spouse) have a postve mpact on nsurance expendture. On the other hand, ncome level and academc background have a negatve mpact on nsurance expendture. Alternatvely, Urata et al. (1999) argued that household ncome, home ownershp, persons aged less than 40, and chldren have a postve mpact on both decsons, and that the amount of fnancal assets ( ) and spouse workng full-tme ( ) are sgnfcant factors affectng nsurance expendture. However, Masu (2000) has argued that rsky asset and savngs-orented nsurance are complements n that a household may hold both assets smultaneously. Further, the amount of savngs-orented nsurance depends on the academc background (graduatng from hgh school), household secondary ncome, and ndvdual annutes. Fnally, Komamura et al. (2000) proved that publc penson and protecton-orented nsurance are substtutes, whle publc penson and savngs-orented nsurance are complements. The purpose of ths research s to estmate the demand for lfe nsurance n Japan. There are two ponts of departure of ths study from prevous work n ths area. Frst, the study examnes not only prvate lfe nsurance products, but also Kampo. We can predct that households whch have a greater knowledge of fnancal nsttutons, the law and markets usually purchase Kampo. There s no exstng research that makes such a predcton. In addton, some households purchase not only Kampo, but also prvate lfe nsurance products, even though both are strongly substtutable 2. In ths case, what are the dfferences between households that have already purchased Kampo and do not purchase any nsurance products and households, whch purchase a prvate lfe nsurance product? Second, we use knowledge and experence factors to estmate lfe nsurance demand. Ferce competton has made fnancal nsttutons, the legal system and markets very complex. We predct that knowledge and expermentaton wth fnancal matters depends upon decsons about asset allocaton. However, exstng research has not examned such factors. It s therefore not possble to know whether the bankruptcy experence, for example, affects asset allocaton 3. There are several prncpal fndngs. Frst, ncome, chld, penson and knowledge factors have a sgnfcant mpact on the decson by ndvduals to purchase ether lfe nsurance product. The bankruptcy experence varable seemed to have mpact only n the case of Kampo. Second, the level of ncome and fnancal assets appears to have sgnfcant effect on the purchase of both prvate lfe nsurance and Kampo. Beng aged less than 40 and cvl servant occupaton factors are shown to only affect the purchase of prvate lfe nsurance. Thrd, ncome, chld, and bankruptcy experence factors are not sgnfcant when households wth prvate lfe nsurance product purchase Kampo. 2 Accordng to publc comments by the Lfe Insurance Assocaton of Japan n 2004, Kampo was founded to complement prvate lfe nsurance products. At that tme, Japanese prvate lfe nsurance frms were very fnancally poor so that they could not sell proper lfe nsurance products, especally to low-ncome ndvduals. Kampo and prvate lfe nsurance products are now substtutable because prvate lfe nsurance frms sell many other knds of lfe nsurance products. 3 Prevous research has not examned these problems because they only arose n conjuncton wth the dramatc changes n Japanese fnancal markets n recent years. 2
Knowledge also has a negatve mpact on addtonal purchases. Fourth, the exstence of fnancal nstablty has lttle to do wth the mpact on nsurance expendture, but knowledge factors have some mpact on the purchasng decson. However, households that have experenced bankruptcy prevously tend to purchase Kampo nstead of prvate lfe nsurance. The remander of ths paper s organzed as follows. Secton 2 brefly dscusses the background of the Japanese lfe nsurance market. Secton 3 ntroduces the data, estmaton method, and explanatory varables. The results are presented n Secton 4. In Secton 5 we estmate and examne the smultaneous equaton models. Some concludng remarks are presented n the fnal secton. 2. Background There are two man reasons today why each Japanese household has to choose an nsurance frm and/or nsurance product more carefully than n the past. Frst, nsurance frms compete more fercely to gan market share snce the change n Japanese nsurance law n 1996. Wth these changes, nsurance frms could sell more consumer-orented and dfferentated products, so households need to consder the most sutable product from the wde varety avalable. Second, snce 1996 some nsurance frms have been bankrupted through ths ferce competton. By May 2005, seven lfe nsurance and two non-lfe nsurance frms had been bankrupted. Accordngly, every household has to assess whether the nsurance frm that they contract wth wll reman solvent n the future. In addton, all Japanese households have the opton to purchase lfe nsurance products outsde of prvate lfe nsurance frms. Japan Post sells publc lfe nsurance products named Kampo. Japan Post s not a prvate sector entty, but a form of publc sector entty. In 2003 Kampo sold amounted to about 185 trllon yen (about 1,769 bllon US dollars) 4, a lttle more than Sumtomo, Japan s thrd bggest prvate lfe nsurance frm 5. Thus, t s mpossble to gnore Kampo when we examne the asset allocaton of Japanese households. Japan Post tends to sell lfe nsurance products that are very smlar to those of other prvate lfe nsurance frms. Snce January 2004, Japan Post sells blended lfe nsurance products wth whole lfe and term nsurance. Ths product s sometmes postoned as the man product by the prvate nsurance frms. However, there are several dfferences between Kampo and prvate lfe nsurance products. Frst, Kampo s nsurance money s restrcted to under 10 mllon yen per person. Second, the Japanese government guarantees all nsurance money f Japan Post were to go nto bankruptcy, whereas prvate lfe nsurance contracts sometmes fal when an nsurance frm bankrupts. Thrd, Japan Post does not lmt jonng on the bass of professons. Fourth, Kampo s sold n post offces located across Japan 6. 4 See Postal Servces n Japan 2004 (annual report), p.14. 5 Sumtomo lfe nsurance frm had about 176 trllon lfe nsurance amounts n fscal 2003. 6 Accordng to Postal Servces n Japan 2004 (annual report) (pp.125 126), as of March 31, 2004, there are 24,715 post offces n Japan. All Japanese prefectures (47 prefectures) have more than 200 post 3
3. Estmaton for Lfe Insurance Demand Functon 3.1. The data Ths research shed lght on some open questons descrbed above usng the data offered by the Postal Servces Research Insttute. At the same tme, we also consder Kampo s man role and what Kampo should be n the future. The data were collected by way of a questonnare. The questonnare contaned very detaled ndvdual nformaton so that we could obtan approprate data from the database. Ths database represents nsurance money and nsurance premums of prvate lfe nsurance, Kampo, and cooperatve nsurance. The number of vald responses was 4,182; a response rate of about 70%. Some 3,273 responses were from household wth two members or more, and the remanng 909 from sngle-person household. In order to protect the prvacy of respondents, we only get about 90% out of total, 3,762. Data from households that dd not respond were excluded. In addton, we restrcted the data n order to satsfy the followng condtons: (1) the age of the householder s less than 60, (2) the householder s workng. We restrcted the data because we would lke to focus on households whose requrements for lfe nsurance were relatvely hgh. Through these restrctons, the sample fell to 2,004 vald responses. 3. 2. Equaton for Estmaton In ths secton, we explan the estmaton method used for the lfe nsurance demand functon. We can dvde a household s lfe nsurance purchasng behavor nto the followng two stages. Frst, households decde whether to purchase lfe nsurance products. Second, they decde the amount of nsurance they decde to purchase. We must be careful that the observed amount of nsurance shows only the demand of households who decde to purchase lfe nsurance products. We express ths by the followng mathematcal form: y = x β + u * '. 1,2, K, n 2 = u N(0, σ ) y = y = 0 * f f y y * * > 0 0 In ths stuaton, there exsts the estmaton bas ( σλ( ' β / σ ) ) shown below f we apply ordnary regresson method to y > 0 observaton. x ' ' ' ' E ( y y > 0) = x β + E( u u > x β) = x β + σλ( x β / σ ) offces. 4
Here, λ ( ) s often called as nverse Mlls rato and s shown as ' ' ' 7 λ( xβ / σ ) = φ ( xβ / σ ) / Φ ( xβ / σ ) If we defne α β / σ, Heckman s two-step method (Heckman (1976), Amemya (1985), Wooldrdge (2002)) s used to estmate the two-stage procedure, as follows: Step (1): Decson on whether to purchase lfe nsurance products: Calculate αˆ, maxmum lkelhood estmator of α, based on Probt model. Step (2): (If household decde to purchase lfe nsurance products) Decson on purchasng amount of nsurance purchased: Usng postve observaton, regress y onto x and λ( x ' ˆ α). In ths paper, we estmate the above two stages smultaneously usng the maxmum lkelhood method. 3. 3 Explanatory Varables Based on prevous sectons, we specfy the followng explanatory varables. Step (1): Wth respect to decson on whether to purchase lfe nsurance products: Yen amount of ncome, yen amount of fnancal assets, occupaton of non-household head dummy (1 f some of the member n famly except for household head has a job, 0 otherwse), chldren dummy (1 f household has more than one chld, 0 otherwse), metropoltan dummy (1 f household s located n metropoltan area, 0 otherwse), penson dummy (1 f household has more than one pensoner, 0 otherwse), comparson of nsurance companes dummy (1 f household compares more than three companes when they purchase lfe nsurance product, 0 otherwse), knowledge on amendment of nsurance busness law dummy (1 f household knows amendment of nsurance busness law to allow nsurance company to cut guaranteed yelds, 0 otherwse), bankruptcy experence dummy (1 f household has experenced bankruptcy of fnancal nsttuton whch they often use, 0 otherwse). Step (2): Wth respect to the decson on purchasng an amount of nsurance: Amount of ncome, amount of fnancal assets, age of household head, owner-occuped house dummy (1 f household has ther house wth no loan, 0 otherwse), publc offcals dummy (1 f occupaton of household head s publc offcals, 0 otherwse), large-scale frm dummy (1 f household head works n large-scale frm wth more than 500 employees, 0 otherwse), penson dummy, comparson of nsurance companes dummy, knowledge on amendment of nsurance busness law dummy, bankruptcy experence dummy. 7 ), Φ ( ) φ ( mean densty functon and dstrbuton functon, respectvely. 5
Descrptve statstcs are shown n Table 1 and the correlaton matrx n Table 2. Note that the orgnal questonnare survey adopts class value alternatves nstead of askng the amount tself. Therefore, we transform class value alternatves nto an amount based on class average, and estmate the demand functon. In addton, amount of fnancal assets ncludes amount of fundng savng-based nsurance n the questonnare. Strctly speakng, we must exclude amount of fundng savng-based nsurance from amount of fnancal assets. Unfortunately, such nformaton was not made avalable. Accordngly, cauton must be taken when nterpretng the results. ------------------------------Insert Table 1 and Table 2 about here ------------------------------ 4. The Results In ths secton, we consder the lfe nsurance demand functon results shown n Table 3. The left-hand sde s the results of the estmaton for domestc prvate lfe nsurance whle the rght-hand sde shows the estmaton results for Kampo 8. In addton, the lower part shows the result for the decson on whether purchasng lfe nsurance products whle the upper part shows the result for the decson on purchasng the amount of nsurance. ------------------------------Insert Table 3 about here ------------------------------ Frst, let us examne the estmaton results for the lfe nsurance holdng functon (lower part) and consder the statstcally sgnfcant varables wth respect to both domestc lfe nsurance and Kampo. Both ncome and chldren dummy varables have a postve sgn and are statstcally sgnfcant. Ths s consstent wth our expectatons. Next, t s ratonal for a sgnfcant postve sgn of penson dummy because a pensoner s an aged person and s consdered to have a greater demand for lfe nsurance. Knowledge of the nsurance busness law, whch can be regarded as a proxy for fnancal lteracy, has a sgnfcant postve sgn. Ths seems to reflect the fact that the household who holds lfe nsurance products tends to have more ample knowledge about fnancal condtons than a household wth no demand. On the other hand, one dstnctve varable that shows a dfferent sgn between domestc lfe nsurance and Kampo s the bankruptcy experence dummy. Ths has a postve sgnfcant sgn only 8 In the questonnare survey, prvate lfe nsurance companes are classfed nto two categores, domestc and foregn-fnanced, accordng to the followng descrpton. Wth respect to prvate lfe nsurance company, please take t as foregn-fnanced f all of the company name s expressed by katakana or Roman alphabet n prncple. In other case, f company name s expressed by mxture of Chnese character, hragana and katakana, please take t as domestc. Note that the followng four companes are exceptons: Aoba (foregn-fnanced), Orx (domestc), Sony (domestc), T&D fnancal (domestc). 6
n the case of Kampo. Ths result reflects that Kampo provdes safe products wth a government guarantee 9. Wth respect to amount of fnancal assets, t shows sgnfcant postve sgn only n the case of Kampo. However, as stated before, ths varable also ncludes amount of fundng savng-based nsurance, so t s dffcult to provde a clear nterpretaton. Next consder the estmaton result for demand functon on amount of nsurance (upper part). Both the amount of ncome and amount of fnancal assets have a sgnfcant postve sgn. Ths means that households wth hgher ncomes and more fnancal assets need larger amounts of nsurance 10. Wth the demand functon for the prvate domestc prvate nsurance company, the age of the household head has a sgnfcant negatve sgn. Generally, people tend to purchase more securty-orented nsurance when they are young, so ths result s understandable. Urata et al. (1999) also report a smlar result, wth a sgnfcant postve sgn for the less than 40 age dummy. Further, the publc offcals dummy has a sgnfcant negatve sgn. Ths may reflect that publc offcals do not have to buy much securty-orented nsurance because of better job securty. Publc offcals also jon n mutual-ad penson plan. Ths penson plan has some advantages n comparson wth other penson plans joned by employees and the self-employed 11. In other words, publc offcals can get more pensons and so they tend to buy smaller amounts of nsurance. On the other hand, the penson dummy shows sgnfcant postve sgn only n the case of Kampo. Ths mght reflect the fact that aged people tend to have a greater amount of endowment nsurance. The three varables (comparson of nsurance frms dummy varable, knowledge on amendment of nsurance busness law dummy, bankruptcy experence dummy) that are prepared as proxes of fnancal nstablty, they seem not to affect on the decson for purchasng amount of nsurance. On the other hand, n decdng whether to purchase lfe nsurance products, the knowledge dummy shows a sgnfcant postve sgn n the cases of both domestc prvate lfe nsurance companes and Kampo. In addton, bankruptcy experence dummy shows a sgnfcant negatve sgn only n the case of Kampo. In other words, households that have experenced bankruptcy wthn a decade tend to ncrease the probablty of purchase nsurance products suppled by Kampo, whch provde safer products. From ths result, we can say that recent fnancal nstablty and a sense of future 9 Accordng to the questonnare conducted by Mnstry of Internal Affars and Communcatons n 2004 (URL http://www.soumu.go.jp/s-news/2004/040817_1.html), the two top reasons for purchasng Kampo are feel safety because Kampo s sold by Japanese government and Japanese government guarantees to pay the nsurance money. 10 In the questonnare, there s a queston concernng the amount of debt. After transformng the class value alternatve nto a yen amount, we estmate the demand functon but obtan a sgnfcant postve sgn. Ths mght be because the amount of debt ncludes housng loan, so household wth hgher ncome tend to have a smlar amount of debt and ths leads to a postve sgn. Urata et al. (1999) also pont out ths possblty. However, ths type of debt dffers from the ordnary meanng of debt, so we report the estmaton result excludng the amount of debt. 11 For example, a mutual-ad penson plan has addtonal payments scheme. Moreover, the range of relatves that can obtan the survvor s penson from a mutual-ad penson s wder than that of other penson plans. 7
uncertanty surely affect nsurance purchasng behavor of Japanese households. Thus, the household needs to buy more nsurance products f t s faced wth bankruptcy because ts nsurance money may lower. However, accordng to the data, more than 50% of household heads who experenced bankruptcy are more than 50 years n age so that they may not be able to buy more nsurance products. Even f they want to buy the nsurance products, the nsurance frms may refuse to contract or requre payng very hgh nsurance premums. On the other hand, t s relatvely easy for them to buy Kampo because Japan Post does not montor each ndvdual. 12 To sum up, the households that have experenced bankruptcy tend to purchase Kampo because the age of the household head s relatvely hgh. 5. Smultaneous Estmaton 5. 1 Portfolo of Lfe Insurance Products In the prevous secton, we estmated the demand functons of lfe nsurance products offered by prvate domestc nsurance companes and Kampo. However, household tends to hold multple lfe nsurance products smultaneously. In consderaton of ths fact, we attempt to devse an mprovement for the more precse estmaton usng the same data. Table 4 shows household portfolo pattern of lfe nsurance products based on the entre questonnare sample. Each pattern s ranked n descendng order. We see that household whch has only product offered by prvate domestc nsurance company s ranked as No. 1 and household whch has only product offered by Kampo s ranked as No. 3. It proves that both products surely play an mportant role n Japanese households. On the other hand, there exsts household whch has both products offered by prvate domestc nsurance company and Kampo and s ranked as No. 2. In addton, there are many other patterns for household to have multple products offered by dfferent nsttuton. Therefore, n ths secton, we estmate and examne smultaneous equaton wth respect to decson on whether purchasng nsurance products offered by prvate domestc nsurance company and Kampo. ------------------------------Insert Table 4 about here ------------------------------ 5. 2 Method for Smultaneous Estmaton Followng the above formulaton, there are two possble ways to estmate smultaneous equatons. Case 1: In the case of allowng correlaton between two lfe nsurance holdng functon We use a smultaneous Probt model for estmatng lfe nsurance holdng functon. In short, we 12 Of course, t does not mean that Japan Post contracts wthout any condtons. 8
use the same settng as the prevous secton n estmatng holdng functon, but here we allow for correlaton of the error term between two equatons wth respect to prvate domestc nsurance company and Kampo. We assume the dstrbuton of error terms s expressed by the BVN (Bvarate Standard Normal Dstrbuton). The formulaton s as follows. y = β, y 0, otherwse 0 * ' * 1 x1 1 + ε1 1 = 1 f y1 > * ' * 2 = x2β 2 + ε 2 2 = 1 f y2 > [ ε1, ε 2 ] BVN(0,0,1,1, ρ y, y 0, otherwse 0 Note that ), ρ : correlaton coeffcent. Case 2: In the case of holdng two knds of products smultaneously offered by prvate domestc Insurance company and Kampo The base settngs are almost the same as case 1, but we employ an addtonal assumpton that we can observe y, ) n the frst equaton only when household purchases lfe nsurance ( 1 x1 product n the second equaton (n other words, y 1). In the context of our analyss, ths 2 = means that we examne the factors for holdng lfe nsurance product offered by prvate domestc nsurance company n the frst equaton, and examne the factors affectng why purchasng households also purchase a Kampo product n the second equaton. Ths factor analyss corresponds to the largest smultaneous holdng pattern n Table 4. 5. 3 Estmaton Results and Interpretaton Our estmaton result s shown n Table 5. Case 1 shows the result of estmatng the two holdng functons n Table 3 smultaneously, whle allowng for correlaton between the error terms. The correlaton coeffcent ρ s 0.172. The sgn and sgnfcance of coeffcent shows almost the same tendency as lower part n Table 3, reflectng a low correlaton. On the other hand, case 2 shows factors ndcatng that household wth prvate lfe nsurance product also purchase Kampo product. The rato of households holdng prvate lfe nsurance product s about 62%, and the rato of household holdng Kampo product smultaneously s about 37%. We can say that ncome or the chldren dummy s not as mportant factor for addtonal holdng of Kampo product although prvate lfe nsurance holdng functon n the frst stage shows almost the same tendency. In addton, the bankruptcy experence dummy shows postve sgn, but s nsgnfcant n the second equaton. Further, knowledge dummy shows rather sgnfcant negatve sgn. Ths result can be nterpreted as follows. The queston n the questonnare s that the lfe nsurance frms could change assumed nterest rate before bankruptcy n accordance wth amendment of nsurance law snce 24 August, 9
2003. Do you know that? 13 Ths questonnare conducted from 29 November to 21 December 2003. 14 There were no nsurance frms to apply such lowerng. Thus, t s rather dffcult to get that knowledge unless they are famlar wth fnancal nformaton n ther daly lfe. In other words, knowledge dummy represents not only the knowledge about amendment of nsurance law, but also more general knowledge about fnancal nsttutons and markets. We can obtan two reasons households that have the knowledge about amendment of nsurance law do not purchase Kampo addtonally. Frst, Kampo contans more savng factors than prvate lfe nsurance products. Thus, these households want to nvest not n Kampo, but n other fnancal assets that do not fx ther money. Second, t s natural to thnk that these households purchased prvate lfe nsurance products more properly because they have fnancal knowledge 15. Thus, they may not need to purchase more nsurance product to adjust to ther needs. ------------------------------Insert Table 5 about here ------------------------------ 6. Concludng Remarks Ths paper estmated prvate lfe nsurance and Kampo demand functons usng household-level data offered by Postal Servces Research Insttute. Our results show that that the ncome, chld, penson, knowledge factors have a sgnfcant effect on the decson whether each household purchases lfe nsurance products. The bankruptcy experence varable appears to have mpact only n the case of Kampo. However, penson and bankruptcy experence appears only to have an mpact on Kampo, whle aged (less than 40) and occupatonal (cvl servant) factors affect only prvate lfe nsurance. Dummy varables representng comparson, knowledge, and bankruptcy experence dd not have any sgnfcant effect on decsons concernng prvate lfe nsurance. Smultaneous estmatons are also used to examne why households that already have one type of lfe nsurance product (e.g. prvate lfe nsurance) purchase the other type of lfe nsurance product (e.g. Kampo). The results ndcate that ncome, chld, and bankruptcy experence varables are not a sgnfcant factor when households wth prvate lfe nsurance product decde to purchase addtonal Kampo. The results also show that a knowledge dummy has a negatve mpact on addtonal purchases. In relaton to recent fnancal nstablty, ths analyss shows the followng results. Frst, the exstence of fnancal nstablty has not mpacted on nsurance funds, but the knowledge factor has some mpact on the purchasng decson. Second, households whch experenced bankruptcy before have a tendency to Kampo rather than prvate lfe nsurance. These results mply that fnancal 13 Ths queston was orgnally wrtten n Japanese. 14 Ths amended nsurance law passed n the Det on 18 July 2003. 15 Tanaka (1999) has nvestgated ths pont usng an emprcal approach. 10
nstablty n Japan affects households decsons. References Amemya, T. (1985), Advanced Econometrcs, Harvard Unversty Press. Heckman, J. (1976), The Common Structure of Statstcal Models of Truncaton, Sample Selecton, and Lmted Dependent Varables and a Smple Estmator for Such Models, Annals of Economc and Socal Measurement 5, pp. 475 492. Komamura, K., T. Shbuya, and F. Urata (2000), The Effect of Publc Penson on Purchasng Behavor for Lfe Insurance Product, Chapter 10, Economc Analyss of Penson and Household, Toyo Keza Shnpo-sya, pp. 135 145. [In Japanese: Ko-tek Nenkn ga Seme Hoken Kanyu Ko-do n Ataeru Ekyo ]. Masu M. (2000), The Estmaton for Asset Management Polcy n Japanese Household, Management of Lfe Insurance Company, Vol. 68, No. 1., pp. 85 97. [In Japanese: Kake no Shsan Unyo-Ho-shn no Sute ]. Postal Servces Research Insttute (2004), Survey of Lfe Style and Fnancal Insttuton Utlzaton (mplemented n November, FY2003), [In Japanese: Kurash to Knyu Kkan Ryo n Kansuru Ishk Cho-sa ]. Tachbanak, T. and K. Shmono (1994), Demand Analyss of Lfe Insurance Asset Selecton for Safe Asset, Rsk Asset and Lfe Insurance, Chapter 9, Indvdual Savng and Lfe Cycle, Nkke Shmbun Press, pp. 220 243. [In Japanese: Seme Hoken no Jyuyo Bunsek Anzen Shsan, Kken Shsan oyob Hoken n Kansuru Shsan Sentaku ]. Tanaka, H. (1999), A Study of Customer s Satsfactons wth Lfe Insurance Products, Lfe Insurance Management, Vol. 67, No. 1, pp. 17 39. [In Japanese: Semehokensyouhn no Kokyakumanzokukouzou n Kansuru Ichkousatsu ]. Urata, F., K. Komamura, and T. Shbuya (1999), Purchasng Behavor for Lfe Insurance Product, Management of Lfe Insurance Company, Vol. 67, No. 1., pp. 3 16. [In Japanese: Kake no Seme Hoken Kanyu Ko-do ]. Wooldrdge, J. (2002), Econometrc Analyss of Cross Secton and Panel Data, MIT Press. 11
Table Descrptve Statstcs Dependent Varables Mean S.D. Amount of Insurance (Domestc Prvate) 2531 3542 0 40200 Holdng nsurance dummy (Domestc Prvate) 0.6203 0.4854 0 1 Amount of Insurance (Kampo) 296 891 0 19998 Holdng nsurance dummy (Kampo) 0.3164 0.4652 0 1 Holdng nsurance dummy (All) 0.7725 0.4194 0 1 Explanatory Varables Amount of ncome 626.75 458.87 100 5000 Amount of fnancal asset 804.04 1309.65 100 10000 Age of household head 43.54 10.32 20 59 Owner occuped house dummy 0.2480 0.4320 0 1 Publc offcals dummy 0.0908 0.2874 0 1 Large-scale frm dummy 0.1786 0.3831 0 1 Penson dummy 0.1811 0.3852 0 1 Comparson of nsurance companes dummy 0.0210 0.1433 0 1 Knowledge dummy 0.4716 0.4993 0 1 Bankrupt experence dummy 0.1652 0.3714 0 1 Occupaton of non-household head dummy 0.9666 0.1798 0 1 Chldren dummy 0.6437 0.4790 0 1 Metropoltan dummy 0.6113 0.4876 0 1 12
Table 2: Correlaton Matrx Explanatory Varables (number of observatons 2004) Amount of ncome Amount of fnancal asset Age of household head Owner occuped house dummy Publc offcals dummy Large-scale frm dummy Penson dummy Comparson dummy Knowledge dummy Bankrupt experence dummy Occupaton of non-household head dummy Chldren dummy Metropoltan dummy Amount of ncome 1.0000 Amount of fnancal asset 0.4895 1.0000 Age of household head 0.3205 0.2918 1.0000 Owner occuped house dummy 0.2088 0.3287 0.3772 1.0000 Publc offcals dummy 0.1973 0.0780 0.0555 0.0316 1.0000 Large-scale frm dummy 0.1537 0.0680-0.0576-0.0748-0.1474 1.0000 Penson dummy 0.1447 0.1665 0.2263 0.2639 0.0362-0.0705 1.0000 Comparson of nsurance companes dummy -0.0192-0.0314-0.0387-0.0356-0.0099 0.0318-0.0326 1.0000 Knowledge dummy 0.1955 0.2046 0.1483 0.0570 0.0945 0.1179 0.0411 0.0223 1.0000 Bankrupt experence dummy 0.1034 0.0960 0.0976 0.0775 0.0465 0.0030 0.0490-0.0088 0.1021 1.0000 Occupaton of non-household head dummy 0.0514-0.0109 0.0643 0.0104-0.0088 0.0215 0.0154 0.0078-0.0078 0.0154 1.0000 Chldren dummy 0.2429 0.0764 0.2023 0.0967 0.0248 0.0532 0.1199 0.0434 0.0495 0.0475 0.0761 1.0000 Metropoltan dummy -0.0678-0.0474 0.0798 0.1000 0.1095-0.1091 0.1066 0.0023-0.0793 0.0487-0.0173 0.0288 1.0000 13
Table : Estmaton Results of Lfe Insurance Demand Functon Prvate Domestc Kampo Explanatory Var. Coef. z-value Coef. z-value Dependent: Amount of nsurance Dependent: Amount of nsurance Amount of ncome 2.225 8.698 *** 0.348 2.793 *** Amount of fnancal asset 0.278 3.139 *** 0.111 2.546 ** Age of household head -34.709-2.898 *** -2.018-0.296 Owner occuped house dummy -102.531-0.412-124.083-0.988 Publc offcals dummy -622.381-1.760 * 139.473 0.813 Large-scale frm dummy 44.168 0.174-186.869-1.231 Penson dummy 209.624 0.806 328.158 2.479 ** Comparson of nsurance companes dummy 109.930 0.147 302.205 0.881 Knowledge dummy -289.397-1.350-183.701-1.621 Bankrupt experence dummy 46.566 0.176-38.905-0.289 Constant 4606.360 8.040 *** 907.473 2.283 ** Dependent Holdng nsurance Dependent Holdng nsurance Amount of ncome 0.001 5.708 *** 0.000 2.329 ** Amount of fnancal asset 0.000-0.267 0.000 2.780 *** Occupaton of non-household head dummy -0.056-0.349 0.067 0.386 Chldren dummy 0.429 6.903 *** 0.579 8.475 *** Metropoltan dummy -0.038-0.620 0.090 1.425 Penson dummy 0.142 1.771 * 0.257 3.341 *** Comparson of nsurance companes dummy -0.266-1.342 0.163 0.803 Knowledge dummy 0.336 5.550 *** 0.162 2.603 *** Bankrupt experence dummy 0.037 0.454 0.186 2.338 ** Constant -0.363-2.145 *** -1.336-7.256 *** ρ -0.293-0.128 σ 3571 1342 λ -1047-171 Rato of selected household (%) 62.03% (=1243/2004) 31.64% (=634/2004) Log-lkelhood -14133-6625 *** : Sgnfcant at 1% level ** : Sgnfcant at 5% level * : Sgnfcant at 10% level 14
Table : Combnaton of Lfe Insurance Products Combnaton Number of Household Rato (%) 1 Prvate (domestc) only 969 23.17% 2 Prvate (domestc) and Kampo 767 18.34% 3 Kampo only 295 7.05% 4 Prvate (domestc & foregn) and Kampo 174 4.16% 5 Prvate (domestc) and Kampo and JA mutual 164 3.92% 6 Prvate (domestc)& foregn) p 159 3.80% 7 mutual 118 2.82% 8 Prvate (domestc) and Other mutual 97 2.32% 9 Prvate (domestc) and JA mutual 95 2.27% 10 Prvate (foregn) only 91 2.18% 11 JA mutual only 90 2.15% 12 Kampo and JA mutual 87 2.08% 13 Other mutual only 66 1.58% 14 Prvate (foregn) and Kampo 58 1.39% 15
Table 5: Smultaneous Estmaton Results of Lfe Insurance Holdng Functon Case 1: Smultaneous Estmaton of Holdng Prvate Domestc and Kampo Prvate Domestc Kampo Explanatory Var. Coef. z-value Coef. z-value Dependent: Holdng nsurance Dependent: Holdng nsurance Amount of ncome 0.000 6.356 *** 0.000 2.436 ** Amount of fnancal asset 0.000-0.220 0.000 2.910 *** Occupaton of non-household head dummy -0.038-0.224 0.062 0.366 Chldren dummy 0.403 6.405 *** 0.580 8.420 *** Metropoltan dummy -0.054-0.867 0.095 1.496 Penson dummy 0.143 1.782 * 0.256 3.400 *** Comparson of nsurance companes dummy -0.253-1.293 0.169 0.845 Knowledge dummy 0.336 5.590 *** 0.164 2.617 *** Bankrupt experence dummy 0.043 0.537 0.185 2.300 ** Constant -0.337-1.893 * -1.337-7.408 *** ρ 0.172 4.406 *** Rato of selected household (%) 62.03% (=1243/2004) 31.64% (=634/2004) Log-lkelhood -2396.355 *** : Sgnfcant at 1% level ** : Sgnfcant at 5% level * : Sgnfcant at 10% level Case 2: Estmaton for Household wth Prvate Domestc to Purchase Kampo Addtonally Prvate Domestc Kampo Explanatory Var. Coef. z-value Coef. z-value Dependent: Holdng nsurance Dependent: Holdng nsurance Amount of ncome 0.000 6.175 *** 0.000-0.933 Amount of fnancal asset 0.000 0.319 0.000 2.541 ** Occupaton of non-household head dummy -0.045-0.265 0.048 0.289 Chldren dummy 0.403 6.420 *** 0.064 0.732 Metropoltan dummy -0.051-0.841 0.060 0.975 Penson dummy 0.151 1.959 * 0.143 1.852 * Comparson of nsurance companes dummy -0.260-1.346 0.138 0.662 Knowledge dummy 0.322 5.393 *** -0.148-2.325 * Bankrupt experence dummy 0.033 0.417 0.115 1.448 Constant -0.319-1.809 * 0.152 0.784 ρ -0.965-23.586 *** Rato of selected household (%) 62.03% (=1243/2004) 37.41% (=465/1243) Log-lkelhood -2010.777 *** : Sgnfcant at 1% level ** : Sgnfcant at 5% level * : Sgnfcant at 10% level 16