Municipal mergers and special provisions of local council members in Japan

Transcription

1 MPRA Munch Personal RePEc Archve Muncpal mergers and specal provsons of local councl members n Japan Haruak Hrota and Hdeo Yunoue Nagoya Unversty of Commerce and Busness, Chba Unversty of Commerce 29 Aprl 2011 Onlne at MPRA Paper No , posted 21 June :23 UTC

2 Muncpal mergers and specal provsons of local councl members n Japan Haruak Hrota, Hdeo Yunoue Abstract The number of muncpaltes n Japan has decreased from 3,232 n 1999 to 1,820 n 2006 because of muncpal mergers, called Hese-no-Dagappe. Ths paper estmates the poltcal choces of local councl members n Japan s muncpal mergers. In Japan, beng a local councl member s a full-tme job. The local councl has veto powers over local admnstraton. Snce the wage for a local councl member s qute hgh, councl members lke to keep ther seats. The jobs of local councl members are affected by muncpal mergers, as preferental treatment and penaltes are delvered by the central government to the local government n muncpal mergers. In our results, merged muncpaltes apply Specal Provsons for local councl members because of the sze of the muncpalty. The choce of muncpalty s also affected by the natonal government s poltcal power. In addton, Specal Provsons lead to addtonal fscal burdens. These fscal burdens wll transfer to the whole country because the Local Allocaton Tax grants system (abbrevated as LAT grants), a natonal grants system, works well n Japan. The muncpaltes that choose the Specal Provsons explot the benefts from other muncpaltes wthout any addtonal costs. Our results show that the central government nduces the free-rder problem n Japan. Keywords: muncpal mergers, local councl sze, ntergovernmental relatons, free rdng JEL Classfcatons: H77, H11, D72 We gratefully acknowledge Tuukka Saarmaa (Government Insttute for Economc Research (VATT)) and Randall Holcombe (Florda State Unversty) for ther comments at The 2011 Meetng of the European Publc Choce Socety. We thank Hdek Konsh (Waseda Unversty), Junch Nagamne (Kwanse Gakun Unversty), Shn Sato (Osaka Gakun Unversty), Akra Yokoyama (Chuo Unversty), Junchro Wada (Yokohama Cty Unversty) and semnar partcpants n Tohoku Gakun Unversty, Yokohama Cty Unversty and Chuo Unversty for ther helpful comments and suggestons. Addtonally, Hrota acknowledges the fnancal support of a Grant-n-Ad for Young Scentsts (B) (Prmary Researcher) Research No from the Japan Socety for the Promoton of Scence, and Yunoue acknowledges the fnancal support of a Grant-n-Ad for Young Scentsts (B) (Prmary Researcher) Research No from the Japan Socety for the Promoton of Scence. Correspondng author. Haruak Hrota, Assstant Professor, Faculty of Commerce, Nagoya Unversty of Commerce and Busness; 4-4 Sagamne, Komenok, Nsshn,Ach, , Japan E-mal:hrota@nucba.ac.jp Hdeo Yunoue, Assstant Professor, School of Servce Innovaton, Chba Unversty of Commerce; Konoda, Ichkawa, Chba, , Japan. E-mal:yunoue@cuc.ac.jp 0

3 1 Introducton The number of muncpaltes n Japan decreased from 3,232 n 1999 to 1,820 n 2006 by muncpal mergers, called Hese-no-Dagappe. Muncpal mergers ntend to strengthen the admnstratve capacty of muncpaltes for fscal decentralzaton. Furthermore, n terms of the economcs-of-scale effect, muncpal mergers can decrease the total expendtures of local governments. The purpose of ths paper s to estmate the effects of the poltcal choces of local councl members n Japan s muncpal mergers. In Japan, beng a local councl member s a full-tme job. The local councl has veto powers over local admnstraton. Snce the wage for a local councl member s qute hgh, councl members lke to keep ther seats. The jobs of local councl members are affected by muncpal mergers, as preferental treatment and penaltes are delvered by the central government to the local government n muncpal mergers. Research on muncpal mergers s manly focused on the decson makng process and the economes of scale of expendture. For example, Sørensen (2006) examned poltcal factors of merger decsons as well as expected effcency gans wth data on Norwegan muncpaltes. Of the emprcal results n Japan, several confrmed ncentves for mergers. Hrota (2007), Kawaura (2009) and Kawamura (2010) studed the ncentves for muncpal mergers n local publc fnance n Japan usng a multnomal and nested logt model. Hrota (2007) reported that factors for the merger of muncpaltes are bad fscal condtons, muncpaltes wth depopulated areas and muncpaltes wth small areas. In partcular, the results confrm that preferental treatment by the central government has effects on mergers of muncpaltes. In addton, the behavor of neghborng muncpaltes affects decson makng for mergers. Wth respect to the effect of economes of scale on expendture, Hanes (2003) studed amalgamaton mpact on the local publc expendture n Sweden. Dolley et al. (2007) reported the effect of mergers on total expendture n Australa, and Geys et al. (2007) consdered Germany. 1 Recently, Hrota and Yunoue (2009) nvestgated the scale effect usng panel data on Japanese muncpaltes and found that muncpaltes acheved reductons n ther total expendtures wth mergers. In ths paper, we focus on the relatonshp between the choces of local councls and muncpal mergers. There s a possblty that poltcal and socogeographc factors nfluence the decsons of muncpaltes n Japan to merge. There are polces such that local councls of merged muncpaltes can choose Zann Tokure (Specal Provsons of Holdng Seats (SPHS) and Tesu Tokure (Specal Provsons of Relaxaton of Legal Sze (RLS)). The former deals wth the protecton of the jobs of local councl members and the latter deals wth the structural change of local councls. These polces provde preferental treatment for councl mergers by the central government, whch of course lead to addtonal 1 See, for example, Bodkn and Conkln (1971), Hrsh (1959, 1965), and Walzer (1972), who studed European countres and the U.S. 1

4 fscal burdens. In addton, our research relates to the common pool problem. If some muncpaltes choose the Specal Provsons, they free rde on other muncpaltes. Ths s because the structure of Japanese local publc fnance s that t s receved as large grants from the central government (general grants system, abbrevated as LAT grants). Whle a sgnfcant porton of local government expendture s fnanced by transfers from the central government, the fscal resource of the LAT grants s natonal taxes. In addton, the amount of LAT grants of each muncpalty s determned based on the muncpalty s fscal shortage (Ihor 2009; Sato and Yunoue 2009). Therefore, an ncrease of a muncpalty s expendture leads to an ncrease n the LAT grant. Ths causes an addtonal fscal burden natonwde. In Specal Provsons for muncpal mergers, a merged muncpalty also receves preferental treatment (for example, expanson of LAT grants) from the central government. In other words, the central government manpulates the merged muncpalty to take free rde on other local governments to the merged muncpalty. The recent lterature relatng to the free-rder problem of muncpal mergers s dscussed n Nelson (1992), Bradbury and Cran (2001), Baqr (2002), Bradbury and Stephenson (2003), Tyrefors Hnnerch (2009) and Jordahl and Lang (2010). Accordng to Hrota and Yunoue (2011), an addtonal expense of a local councl was observed when a muncpalty appled the SPHS. The rest of the paper s organzed as follows. Secton 2 dscusses muncpal mergers n Japan. Secton 3 defnes the emprcal model and provdes the results of the logt estmaton. Secton 4 concludes the paper. 2 Muncpal mergers n Japan 2.1 Background of muncpal mergers To start wth, we brefly explan Japanese fscal decentralzaton. There are three layers of government structures n Japan. There are the central, prefectural and muncpal governments. The local government nvolves prefectures and muncpaltes. Japan has tended toward centralzaton n admnstraton and publc fnance for the last several decades. The central government has a strong nfluence on the prefectural or muncpal governments. In local publc fnance, local government has largely depended on ntergovernmental transfers from the central government for a long tme. For example, the Local Allocaton Tax grants system (abbrevated as LAT grants) has extremely strong effects on fscal adjustment. The LAT grants system nvolves about 17 trllon yen every year and contrbutes about 20% of local government revenues. In fscal 2003, over 3,000 muncpaltes receved LAT grants from the central government (Ihor 2009; Sato and Yunoue 2009). In recent years, both central and local governments have suffered from fscal problems and a declnng and agng populaton, partcularly n local government. Gven these problems, the central 2

5 government has been aggressvely promotng fscal and poltcal decentralzaton snce the Decentralzaton Act n In addton, the Kozum Cabnet mplemented the plan on the Trple Reform from 2003 to These reforms are the shftng of the tax source from the central to the local governments, reducng subsdes and revewng LAT grants. The ntent of the Trple Reform s to replace specfc and general-purpose subsdes wth local taxes (see Sato and Yunoue 2009). As a result, government subsdes were reduced by about 5.2 trllon yen, the tax source was shfted from the central government to the local governments by about 3 trllon yen, and LAT grants were reduced by about 3.6 trllon yen (Ihor 2009). The central government encourages muncpal mergers n tandem wth work on economc and fscal structural reforms. Through the muncpal merger process, known as Hese-no-Dagappe, the number of muncpaltes n Japan decreased from 3,232 n 1999 to 1,820 n The prncpal objectves of muncpal mergers are to acheve a strengthenng of admnstraton and an mprovement n ts fscal condton. Mergers are expected to reduce total expendture because of economes of scale. In order to promote muncpal mergers, the central government varously supports the local government by Sh Tyo Son no Gappe no Tokure n kansuru Hortu, whch s the Specal Muncpal Mergers Law from the Mnstry of Internal Affars and Communcatons (MIC). The Specal Muncpal Mergers Law, the so-called Ame to Much no Sesaku (the Carrot-and-Stck Polcy), was promsed to merged muncpaltes for varous preferental treatments and punshments for a few years. Ths law was temporary legslaton untl March Durng ths perod, merged muncpaltes receved preferental treatment (carrot polcy) from the central government: expanson of LAT grants and local tax measures, specal bonds on the mergers, and specal provsons of local councl members. On the contrary, nonmerged muncpaltes, n partcular small vllages, receved punshment (stck polcy), such as shrnkng of LAT grants. In addton, there was a relaxaton of the populaton s requrements for upgradng to Sere Shte Tosh (the government-desgnated ctes). A government-desgnated cty s able to have ncreased authorty as well as prefecture government. The central government relaxed requrements to the extent of settng the populaton requrements from one mllon to 0.7 mllon persons. In addton, when towns are upgraded to ctes, the requrements are temporarly relaxed to the extent of settng the populaton lmt at 30,000 or more. In the polces above, the central government manpulated muncpaltes wth the carrot-and-stck approach. In ths paper, we focus on prorty local councl members and the poltcal choces of merged muncpaltes. 3

6 2.2 Specal provsons of the local councl members We brefly explan the sze of local councls wthn muncpaltes. Table 1 shows an upper lmt of the local councl sze dependng on the extent of the muncpalty s populaton (Local Autonomy Act, Artcle 91). The number of local councl members usually reaches upper lmt. Notce that there are exceptons for large ctes. In a cty whose populaton s over 900,000, an upper lmt of the local councl sze s decded as follows. The local councl sze of a large cty s ncreased by eght persons for each addtonal 500,000 persons n the populaton. For example, f the populaton sze s 1,400,000 (equals 900,000 plus 500,000), an upper lmt of local councl sze s 64 (equals 56 plus 8). In ths secton, we explan the Act on Specal Provsons of the Merger of Muncpaltes about local councl members (abbrevated as Specal Provsons). There are two artcles of Specal Provsons for local councls. 2 Snce the central government wshes to promote muncpal mergers, the Specal Provsons for local councls are preferental treatment (carrot polcy) under muncpal mergers. The frst s Zann Tokure, whch s the Specal Provsons of Holdngs Seats (SPHS). Ths law provdes job protecton for local councl members. When a merged muncpalty selects the SPHS, a local councl member of the former muncpalty can stay n the new muncpalty for a few years. Ths s a preferental treatment to address poltcal opposton to the merger by local councl members. For example, f two ctes (Cty A and Cty B) and one town (Town C) have a plan to merge and the numbers of the former councl members are 34, 30 and 14, respectvely, then the local councl sze of the new muncpalty wll be 78 when they apply the SPHS. On the contrary, the upper lmt of councl members would be only 46 wthout the SPHS. In ths example, 32 councl members receve the preferable treatment of retanng ther seats. The second s Tesu Tokure, whch s the Specal Provsons of Relaxaton of Legal Sze (RLS). Ths law allows the structural change of a local councl. When a merged muncpalty selects the RLS, t s possble to relax the upper lmt of ts local councl s legal sze up to twce the upper lmt of normal muncpaltes for a few years. However, when a merged muncpalty does not use both the SPHS and the RLS, the merged muncpalty elects a new local councl wth an upper lmt as reported n Table 1. 2 Another Specal Provson s a retrement penson plan. When former local councl members automatcally lose ther job under muncpal mergers, the central government regards retred councl members as regular members of the local councl and pays them a penson. 4

7 3 Emprcal analyss 3.1 Data and summary statstcs We estmate a bnary choce model usng Japanese muncpalty data. We have cross-sectonal data of the 549 merged muncpaltes for three years ( ). Because we focus on the poltcal choce of the merged muncpalty, the nonmerged muncpaltes are excluded from our sample. The muncpal government s data are manly derved from the Sh Tyo Son Kessan Card (Statstcs of Fnal Account of Muncpal Governments) and the Gappe Dgtal Archve (Dgtal Archve of Muncpal Mergers). 3 We manly focus on two knds of factors, socogeographc and poltcal. Frst, we use socogeographc factors that are represented by populaton, area, number of former muncpaltes ncluded n a merged muncpalty, coalfeld areas, remote slands, and depopulated areas. The coalfelds, remote slands and depopulated areas have been desgnated by the MIC as requrng specal support. For example, many coalfeld areas reman n bad fnancal crcumstances. Ths socogeographc factor, whch ncludes the fscal condton, nfluences the poltcal choce of the merged muncpalty. Ths s the reason why we consder the socogeographc factors of muncpaltes as dependent varables. Second, we are nterested n whether the poltcal factors affect the merged muncpaltes or not. We show the Japanese government structure n Fgure 1. In Japan, because not only fnancal supports but also personnel exchanges have strong nfluences, the central government controls the prefectures. Moreover, the prefectures nfluence muncpaltes wth fnancal and personnel power. For reference, ndependent varables of each poltcal factor nclude the followng. Prefecture Instructon Pattern: The MIC requested prefectures to set up Merger Support Measures to help muncpaltes to progress muncpal mergers. Thus, the prefectures nstructed a combnaton of muncpaltes. Because of the long hstory of a centralzed system n Japan, ths nstructon to a combnaton of muncpaltes s not conclusve but t has a strong nfluence on the muncpaltes. If the observed combnaton of muncpal mergers concded wth the nstructed combnaton, then the dummy varable equals 1 and s zero otherwse. Ths dummy varable shows the degree of upper-level government s nfluence on the poltcal choce of the lower-level government. In other words, f the coeffcent of the dummy varable s sgnfcant, then the upper-level government has nfluence on the poltcal choce of the lower-level government. Detals of the observed combnatons of muncpal mergers and the nstructed combnaton set dummy varable nvolve the followng four patterns: 3 Gappe Dgtal Archve ( MIC). 5

8 (1) f the nstructed combnaton s (Cty A, Cty B and Town C), and the observed combnaton s (A, B and C), then the dummy varable equals 1; (2) f the nstructed combnaton s (A, B and C), and the observed combnaton s (A and B), then the dummy varable equals zero; (3) f the nstructed combnaton s (A, B and C), and the observed combnaton s (A, B, C and D), then the dummy varable equals zero; (4) f the nstructed combnaton has two patterns (A, B and C) (D, E and F), and the observed combnaton s (A, B, C, D, E and F), then the dummy varable equals 1; where A, B, C, D, E, F represent the muncpaltes. Prefectural Governor from the Bureaucracy: In Japanese local admnstraton and fnance, the upper-level government often has a strong nfluence on the lower-level government because of personnel relatonshps. 4 MIC bureaucrats have become prefectural governors n some Japanese local admnstratons for nearly 100 years. They mght be actvely mplementng a polcy towards the realzaton of muncpal mergers. If a prefectural governor was prevously an MIC bureaucrat, then the dummy varable equals 1 and s zero otherwse. The numbers of the RLS and the SPHS applcatons by prefecture are reported n Fgure 2. Accordng to Fgure 2, Hroshma and Kagoshma acheved the hghest number of applcatons of the RLS of 8. On the other hand, dfferences were observed for the SPHS. The hghest number of applcatons of the SPHS was 24 for Ibarak. Usng Fgure 2, we confrmed the dfferences among prefecture governments as regards the choce of Specal Provsons of Local Councl Members. Summary statstcs are reported n Table 2. The mean value of the populaton data s 87,975. Whle the maxmum value of the populaton s about 1.4 mllon, the mnmum value s 1,572. In the data of merged muncpaltes, sample varablty of the populaton data s large. Smlarly, area and number of muncpaltes ncluded n muncpal mergers vary greatly n these data. Moreover, summary statstcs of muncpaltes that apply the SPHS or the RLS, are reported n Tables 3 and 4, respectvely. There are 313 muncpaltes that appled the SPHS. On the other hand, 88 muncpaltes appled the RLS. It s of nterest to pont out the statstcal dfference between the SPHS and the RLS. Comparng the mean value of the populaton data of the RLS wth that of the SPHS, t s clear that the mean value of the populaton of the RLS s large. The mean value of populaton data of the RLS s 199,692. On the contrary, that of the SPHS s 72,724. It s lkewse clear that area and number of muncpaltes exhbt statstcal dfferences between the RLS and the SPHS. The mean value of area of the RLS s 581 km 2, but that of the SPHS s 305 km 2. The maxmum value of number of muncpaltes 4 Sum (2000) and Yunoue (2005) reported that LAT grants derved by personnel relatonshps are observed n Japan. 6

9 ncluded n merged muncpaltes s 14 for the RLS and t s 8 for the SPHS. 3.2 Emprcal model and results We estmate what factors of muncpaltes affect ther choce of the Specal Provsons n the mergers usng the emprcal model, defned by equaton (1): y * = α + β where CF Pop p _ G ln( Pop ) + β Pop2 + β CoalFeld + β remote + β Depopulated + β + β pre _ Governor + ε r ln( Pop ) *2 D + β Area ln( Area ) + β p _ I Num Num. ofmun pre _ Instructon, (1) y y 1 = f y * 0 = f y * < 0 0 and refers to the th muncpalty; β represents the coeffcents of ndependent varables; and ε s an error term composed by dentcally ndependent random varables wth standard logstc dstrbuton. * y s latent varable. The dependent varable y s consdered to be two patterns of muncpalty choce as follows. (1) Specal Provsons of Holdngs Seats (SPHS) : the am of our regresson for the SPHS s to examne the choce of job protecton for local councl members. (2) Specal Provsons of Relaxaton of Legal Sze (RLS) : the objectves of our regresson for the RLS are to examne the choce of structural changes of local councls. The ndependent varables nclude socodemographc characterstcs such as populaton, squared populaton, area, dummy varable for coalfeld areas, dummy varable for remote slands and dummy varable for depopulated areas. The varable Num. ofmum represents the number of former muncpaltes ncluded n a merged muncpalty. We also consder the poltcal factors: pre _ Instructon s a dummy varable ndcatng the prefecture s poltcal power n muncpal mergers; and pre _ Governor s a dummy varable representng the presence of MIC s bureaucrat-turned-prefectural governor. The emprcal results of the choce of the SPHS are reported n Table 5. 5 The coeffcent of the populaton term s statstcally sgnfcant at the 5% level. Because the coeffcent of populaton s postve and the coeffcent on squared populaton s negatve, ths mples that addtonal populaton 5 See also the margnal effect of the probablty of each varable n Fgure 3. 7

10 sze has a postve effect on choce probablty of the SPHS for low-populaton muncpaltes. On the other hand, the populaton effect becomes negatve for hgh-populaton ctes. The coeffcent of the number of prevous muncpaltes s negatve and statstcally sgnfcant at the 1% level. These results are qute ntutve. The SPHS s the most preferable polcy for local councl members, and, thus, local councl members hope to apply ths specal provson. However, the physcal capacty of the councl floor s lmted, and, therefore, the small muncpaltes tend to choose the SPHS. The coeffcent of the depopulated area dummy varable s estmated to be negatve and statstcally sgnfcant at the 5% level. Ths result seems to conflct wth the former results. Accordng to Hrota and Yunoue (2011), addtonal expenses of local councls were observed when a muncpalty apples the SPHS. 6 Because depopulated areas have poor fscal condtons, they are unable to afford the addtonal fscal burden to ncrease the number of local councl members. The coeffcent of the Prefecture Instructon Pattern dummy s postve and statstcally sgnfcant. A muncpalty that s affected by the prefecture s poltcal power tends to choose the SPHS. Ths mples that some merged muncpaltes were ndrectly nduced by the carrot polcy of the MIC through prefectural nstructon. The emprcal results for the RLS are reported n Table 6. The estmated values of the populaton terms are contrary to those for the SPHS. A large populaton cty has a hgh probablty of choosng the RLS. Smlarly, the coeffcent of the area term s postve and statstcally sgnfcant; muncpaltes whose areas are large tend to choose the RLS. The coeffcent of the number of former muncpaltes s postve and sgnfcant at the 1% level. The coeffcent of the poltcal term s not estmated to be statstcally sgnfcant. Prefecture nstructon and prefectural governor from the Bureaucracy do not sgnfcantly affect the choce of the RLS. These results mply that large ctes have an ncentve to upgrade ther status by mergers. Snce upgradng a cty status leads to more power of regstraton, ctes that face the promoton of status tend to merge wth other ctes. Therefore, the nstructon of prefecture or the leadershp of a governor does not sgnfcantly affect the choce of the Specal Provsons. 4 Concluson Ths paper has nvestgated the poltcal choce of muncpal mergers n Japan for both the SPHS and the RLS for merged muncpaltes. Frst, a small-szed muncpalty has a hgh probablty of choosng the SPHS. The carrot polcy of the central government leads to the promoton of muncpal mergers. The choce of the SPHS s also affected by the prefecture s poltcal power. Ths mples that some merged muncpaltes were ndrectly nduced by the carrot polcy of the MIC through prefectural nstructon. Second, a large-szed muncpalty has a hgh probablty of choosng the RLS. Ther prvate ncentves tend to merge wth 6 Egger and Koethenbuerger (2010) showed that the sze of government spendng s postvely related to the number of legslators usng German muncpal panel data. 8

11 those of other ctes, and they the RLS to upgrade to larger ctes. In other words, poltcal factors do not affect the choce of the RLS. These Specal Provsons lead to an addtonal fscal burden. Ths fscal burden wll transfer to the whole country snce the LAT grants system, a natonwde grants system, works well n Japan. The muncpaltes that choose the Specal Provsons explot the benefts from other muncpaltes wthout any addtonal costs. Our results show that the central government nduces the free-rder problem n Japan. References Baqr, R. (2002). Dstrctng and government overspendng. Journal of Poltcal Economy, 110, Bodkn, R. G., Conkln, D. W. (1971). Scale and other determnants of muncpal government expendtures n Ontaro: A quanttatve analyss. Internatonal Economc Revew, 12, Bradbury, J. C., Cran, W.M. (2001). Legslatve organzaton and government spendng: Cross-country evdence. Journal of Publc Economcs, 82, Bradbury, J. C., Stephenson, F. (2003). Local government structure and publc expendtures. Publc Choce, 115, Dolley, B., Byrnes, J., Crase, L. (2007). Is bgger better? Local government amalgamaton and the south Australan rsng to the challenge nqury, Economc Analyss & Polcy, 37(1), Egger, P., Koethenbuerger, M. (2010). Government spendng and legslatve organzaton: Quas-expermental evdence from Germany. Amercan Economc Journal: Appled Economcs, 2(4), Geys, B., Henemann, F., Kalb, A., (2007). Local governments n the wake of demographc change: effcency and economcs of scale n German muncpaltes, ZEW dscusson paper No Hanes, N. (2003). Amalgamaton mpact on local publc expendtures n Sweden. Workng Paper. Umeå Unversty. Hrota, H. (2007). Verfcaton of muncpal ncorporaton wth dscrete choce model. Plannng Admnstraton, 30(4), (In Japanese). Hrota, H., Yunoue, H. (2009). Muncpal mergers and local government expendture. Evdence from Hese muncpal mergers n Japan. Mmeo (In Japanese). Hrota, H., Yunoue, H. (2011). Do muncpal mergers affect local councl expendtures? Evdence from muncpal mergers n Japan. Annual Journal of Japanese Local Publc Fnance, 18, (In Japanese). Hrsch, W. (1959). Expendture mplcatons of metropoltan growth and consoldaton. Revew of Economcs and Statstcs, 41,

12 Hrsch, W. (1965). Cost functons of an urban government servce: Refuse collecton. Revew of Economcs and Statstcs, 47, Ihor, T. (2009). Poltcal decentralzaton and fscal reconstructon n Japan. In S. Ichmura & R. Bahl (Eds), Decentralzaton polces n Asan development (Ch. 3, pp ). Sngapore: World Scentfc. Jordahl, H., Lang, C-Y. (2010). Merged muncpaltes, hgher debt: On free-rdng and the common pool problem n poltcs. Publc Choce, 143, Kawamura, K. (2010). Poltcal atttudes, local electons and muncpals mergers n Japan. Tokyo: Bokutakusha. (In Japanese). Kawaura, A. (2009). Self-servng mayors and local government consoldatons n Hokkado, mmeo. Nelson, M. A. (1992). Muncpal amalgamaton and the growth of the local publc sector n Sweden. Journal of Regonal Scence, 32, Sato, S., Yunoue, H. (2009). The reform of Japanese local governments. In Ichmura, S., & R. Bahl (Eds), Decentralzaton Polces n Asan Development (Ch. 11, pp ). Sngapore: World Scentfc. Sørensen, R. L. (2006). Local government consoldatons: The mpact of poltcal transacton costs. Publc Choce, 127, Sum, E., (2000). Hojokn no Chk Habun n Okeru Sej Kanryo Youn no Kensyo, Mta Gakka Zassh 93(1), (In Japanese). Tyrefors Hnnerch, B. (2009). Do mergng local governments free rde on ther counterparts when facng boundary reform? Journal of Publc Economcs, 93, Walzer, N. (1972). Economes of scale and muncpal polce servces: The Illnos experence. Revew of Economcs and Statstcs, 60, Yunoue, H. (2005). Influences of bureaucrats on specal local allocaton tax grants, Publc Choce Studes, 45, (In Japanese). 10

13 Fg 1. Japanese Government structure Central Government (Natonal Government) Grants, Subsdes Personnel Exchanges Prefecture (Local Government) Prefecture Prefecture Grants, Subsdes Personnel Exchanges Cty Town Vllage Cty Town Vllage 11

14 Table 1. Local Councl-Sze Law: Japanese muncpaltes Cty Town and vllage Populaton sze Number of councl members Populaton sze Number of councl members ~50, ~2, ,000 ~100, ,000~5, ,000 ~200, ,000~10, ,000~300, ,000~20, ,000~500, ,000~ ,000~900, ,000~ 56~96 Table 2. Summary statstcs Varable Obs Mean Std. Dev. Mn Max Specal Provsons of holdngs seats Dummy Relaxaton of legal-sze Dummy Populaton Area Num. of muncpaltes Coal Feld Area Dummy Remote Islands Dummy Depopulated Area Dummy Prefecture Instructon Pattern Dummy Prefectural Governor from MIC Dummy Table 3. Summary statstcs of Specal Provsons of Holdngs Seats (SPHS) Varable Obs Mean Std. Dev. Mn Max Populaton Area Num. of muncpaltes Coal Feld Area Dummy Remote Islands Dummy Depopulated Area Dummy Prefecture Instructon Pattern Dummy Prefectural Governor from MIC Dummy Table 4. Summary statstcs of Relaxaton of Legal Sze (RLS) Varable Obs Mean Std. Dev. Mn Max Populaton Area Num. of muncpaltes Coal Feld Area Dummy Remote Islands Dummy Depopulated Area Dummy Prefecture Instructon Pattern Dummy Prefectural Governor from MIC Dummy

15 Fg 2. Applcaton numbers of Specal Provsons :SPHS

16 Table 5. Bnary logt model estmaton results: Specal Provsons of Holdngs Seats (1) Varables Specal Provsons Margnal effects of holdngs seats ln(populaton) 3.308** 0.811** (1.341) (0.329) ln(populaton) ** ** (0.061) (0.015) ln(area) (0.148) (0.036) Num. of muncpaltes *** *** (0.070) (0.017) Coal Feld Area Dummy (0.706) (0.147) Remote Islands Dummy (0.323) (0.081) Depopulated Area Dummy ** ** (0.255) (0.060) Prefecture Instructon Pattern Dummy 0.499* 0.118* (0.269) (0.061) Prefectural Governor from Bureaucrat Dummy (0.207) (0.051) Constant ** (7.352) Observatons Log lkelhood Pseudo R Margnal effects after logt Notes: Standard errors are n parentheses. Varables whose coeffcents are sgnfcant at the 10%, 5%, and 1% levels are ndcated by *, **, and ***, respectvely. 14

17 Table 6. Emprcal results for Relaxaton of Legal Sze (2) Varables Relaxaton of legal-sze Margnal effects ln(populaton) * * (2.242) (0.206) ln(populaton) ** 0.019** (0.099) (0.009) ln(area) 0.468* 0.041* (0.246) (0.021) Num. of muncpaltes 0.324*** 0.028*** (0.082) (0.008) Coal Feld Area Dummy ** (1.361) (0.034) Remote Islands Dummy (0.416) (0.042) Depopulated Area Dummy (0.408) (0.032) Prefecture Instructon Pattern Dummy (0.371) (0.031) Prefectural Governor from Bureaucrat Dummy (0.332) (0.028) Constant (12.605) Observatons Log lkelhood Pseudo R Margnal effects after logt Notes: Standard errors are n parentheses. Varables whose coeffcents are sgnfcant at the 10%, 5%, and 1% levels are ndcated by *, **, and ***, respectvely. 15

18 Fg 3. Margnal effects Estmated Probablty ln(pop) 95% CI Specal Provsons of holdngs seats Relaxaton of legal-sze Estmated Probablty Num. of muncpaltes 95% CI Specal Provsons of holdngs seats Relaxaton of legal-sze 16

19 Fg 3. (Contnued.) Estmated Probablty ln(area) 95% CI Specal Provsons of holdngs seats Relaxaton of legal-sze 17