The Demise of the Swiss Interest Rate Puzzle. March WWZ Working Paper 04/09 (B-093)

Transcription

1 Wrschafswssenschaflches Zenrum (WWZ) der Unversä Basel March 2009 The Demse of he Swss Ineres Rae Puzzle WWZ Workng Paper 04/09 (B-093) Peer Kugler, Bearce Weder

2 The Auhor(s): Prof. Dr. Bearce Weder d Mauro Johannes-Guenberg Unversä DE-5528 Manz bearce.weder@un-manz.de Prof. Dr. Peer Kugler Deparmen of Moneary Macro Economcs (WWZ), Unversy of Basel Peer Meran-Weg 6 CH Basel peer.kugler@unbas.ch A publcaon of he Cener of Busness and Economcs (WWZ), Unversy of Basel. WWZ Forum 2009 and he auhor(s). Reproducon for oher purposes han he personal use needs he permsson of he auhor(s). Conac: WWZ Forum Peer Meran-Weg 6 CH-4002 Basel forum-wwz@unbas.ch

3 Paper prepared for he SNB-IMF Conference on Exchange Raes November 2008, Zürch The Demse of he Swss Ineres Rae Puzzle Peer Kugler and Bearce Weder Ths verson: January Absrac In hs paper we analyzed he volaons of UIP for he Swss Franc agans he Dollar, he Euro, he Yen, he Pound and he Canadan Dollar usng recen daa up o fall Ths exercse provdes he followng man resuls : frs he Swss neres rae puzzle dsappeared,.e. mean reurns on Swss Franc asses are no longer sysemacally and sgnfcanly lower (as n he sample) when compared jonly wh oher major currences n he 999 o 2008 perod. Second and n conras o earler evdence we faled o fnd evdence ha geopolcal crses lead o an apprecaon of he Swss Franc n he las en years. Thrd even he shor run valdy of UIP canno be rejeced for he Swss Franc agans he fve currences for he sample. Unforunaely our aemp o locae he me of change and he currences nvolved by he applcaon of ess for srucural breaks wh unknown break pon dd no lead o clear cu concluson for he Euro and he Dollar, he wo mos mporan foregn currences from a Swss perspecve. JEL Classfcaon: E43, E44, G5 Auhors E-mal: Peer.Kugler@unbas.ch, Bearce.Weder@un-manz.de ) Deparmen of Economcs, Unversy of Basle, Swzerland and Deparmen of Economcs Unversy of Manz and CEPR.

4 . Inroducon Almos exacly en years ago, we sared researchng wha we called he Swss Ineres Rae Puzzle wh he nenon of evenually solvng. Ths aemp led, among ohers, o hree papers (Kugler and Weder, 2002, 2004, 2005) n whch we analyzed he long and shor run neres rae pary of 0 mayor OECD currences usng daa before 999 and found ha here s a long run falure of he uncovered neres rae pary condon for he Swss Franc. Afer correcng for exchange rae changes, mean reurns on Swss asses were sgnfcanly lower han n oher currences, an anomaly no found n any oher major currency. The mean reurn dfferenal over he long run was sable over he perod and ransory srucural breaks were only found n mes of currency urmol. We offered a possble explanaon for he reurn anomaly, namely ha may be due o an nsurance premum agans very rare evens, such as a major war. We found suppor for hs hypohess from wo emprcal fndngs: frs, we showed ha he dfferenal beween exchange rae correced foregn currency neres raes and he Swss franc neres rae - whch s normally posve - s hghly negavely affeced n he shor run by unexpeced evens ncreasng he world-wde polcal uncerany. Second he examnaon of hsorcal daa of neres raes durng he pre-war and ner-war perod showed ha he abnormal low level of Swss neres raes arses afer he Frs World War n a perod of very hgh moneary and polcal uncerany. The las en years were characerzed by whch wo poenally very mporan moneary and polcal changes: frs, he replacemen of 5 somemes raher unsable and hgh nflaon currences by he relavely sable common new currency, he Euro. In such a world a sable currency lke he Swss Franc may become less aracve. Second, he collapse of he Sove block n he early nnees decreased he probably of a large scale war as he wo world wars n he 20 h cenury. Ths mgh also affec he safe haven properes of he Swss Franc, makng less aracve. Las bu no leas, he fnancal urmol of las weeks and monhs have provded a new even sudy of sress and a sharp apprecaon of he Swss Franc. Overall, seems a good me o revs he orgnal puzzle, o esablsh wheher he sylzed facs sll hold and re-evaluae our soluons. The paper s organzed as follows. Secon 2 revss he sylzed facs up o 998 an gves a bref accoun of he explanaons offered for he Swss Ineres Rae Puzzle. Secon 3 dscusses he sylzed facs over roughly he las en years und looks for srucural breaks and secon 4 concludes. 2

5 2. The Puzzle Sylzed Facs and Possble Explanaons In a seres of hree prevous papers (Kugler and Weder, 2002, 2004, 2005) we esablshed he followng sylzed facs for he perod : () () () (v) (v) (v) The long run volaon of uncovered neres rae pary was unque o he Swss Franc, whch was no found n any oher major currency. The puzzle appled equally o all currences under consderaons. The reurn dfferenals were sable and dd no dmnsh over me snce he end of World War I. The puzzle of low real neres raes for he Swss Franc money marke nsrumens and bonds was manly due o a long run deparure from uncovered neres rae pary. Devaons from relave PPP play no mporan role. The Swss Franc was no specal wh respec o he well known shor run volaons of uncovered neres rae pary (forward premum puzzle), bu The reurn anomaly was presen for fxed ncome asses (money marke nsrumens and bonds) bu and no for equy. Wha are he explanaons offered for he puzzlng sylzed facs whch are called he Swss Ineres Rae Puzzle? Bankng secrecy and capal nflows no Swzerland Snce here s a wde spread suspcon ha he naure of he Swss bankng secrecy may be responsble for he neres rae sland s worh resang ha hs was one of he blnd alleys. The popular argumen runs lke hs: ha bankng secrecy aracs huge funds ha are evadng axes no Swzerland and hs drves down Swss neres raes. There are several reasons o dscard hs as a soluon o he puzzle. Frs of all hs proposon suggess ha he reurn rae dfferenal s ha should only apply o asses whch are 3

6 held a Swss banks locaed n Swzerland and subjec o Swss jursdcon. However, he neres rae puzzle was found n Euro marke or currency deposs,.e. shor erm deposs n Swss Francs, whch are held ousde Swzerland. Second, he foregn demand for Swss asses s quanavely small snce non resdens hold her deposs wh Swss banks mosly n oher currences. Furhermore, non resdens hold Swss Franc deposs mosly n equy, where we dd no fnd a reurn dfferenal. We found lle evdence for a role of bankng secrecy snce non resdens have a very lmed preference for fxed ncome nsrumens ssued by foregn debors, whch should be he prme nsrumens for ax evason snce hey are exemp of whholdng axes ha apply o resden ssues. In Kugler and Weder (2004) we conduced a dynamc facor analyss o examne wheher he porfolo shf of resdens and non resdens can explan he reurns on Swss asses and fnd ha foregn demand had almos no mpac on Swss Franc asse prces. The daa on porfolo holdngs were aken from a survey conduced on a monhly bass by he Swss Naonal Bank (SNB) and whch are avalable snce lae 998. They nclude he porfolo holdngs deposed wh 342 banks locaed n Swzerland and Lechensen and cover abou 95% of he oal value deposed. Neverheless, here are some lmaons of hs daa se. In parcular, here s he problem of daa coverage. For example, f a US nvesor holds a Swss Franc secury and deposs wh a non-swss bank, hs would no be ncluded n our daase. Noe ha we are no clamng ha bankng secrecy s rrelevan for he ncome of he Swss wealh managemen ndusry. Bu canno serve as an explanaon for he neres rae dfferenals. Dversfcaon and Porfolo opmzaon To assess he amoun of leakage, ake he example of Novars AG. Based on he share regser n 2005, 53% of he shares regsered by name are deposed whn Swzerland, and 36% are held by approxmaely 850 holders n he US. Roughly 4% of he shares regsered are held by real or ndvdual nvesors, whle 86% are held by nsuonal nvesors. 4

7 In Kugler and Weder (2004) we showed ha some Swss franc asses (money marke and bonds bu no eques) have a very low or even negave correlaon wh he reurns of oher rsky asses and hey can herefore serve o dversfy. Indeed, usng he porfolo shares repored n he SNB survey for Swss resdens and reversed porfolo opmzaon (Black and Lerman 992) o back ou he mpled expeced reurns we found ha he esmaed paern of hese reurns conformed very well o he observed paern and he Swss neres rae puzzle. From hs research we concluded ha he neres rae dfferenal could be explaned by a dversfcaon move. The problem s ha hs does no amoun o a full explanaon. The fndng alone does no ell us whch are he deep facors drvng hs paern of correlaons and equlbrum reurns. Insurance agans caasrophc rsk and Peso Problem Our preferred canddae for a deep facor was an nsurance premum for caasrophc evens (Kugler and Weder (2005)). Invesors may accep somewha lower reurns n normal mes f hey expec ha he Swss Franc would apprecae n he even of a large scale dsaser. Snce such evens are rare and have no been observed n he pas 20 years s dffcul o es he hypohess for hs perod. Therefore we are confroned wh a peso problem when ryng o esmae he hypohess based on daa from ranqul mes. Neverheless we found suppor for hs hypohess from wo emprcal fndngs: frs, we showed ha he reurn dfferenal s negavely affeced n he shor run by large unexpeced geo-polcal evens, such as he fall of he Berln Wall, or he sudden deah of Sove leaders. Second we presened hsorcal evdence on neres raes dfferenals durng he pre-war and ner-war perod. These show ha he abnormally low level of Swss reurns arose afer he Frs World War only. Our evdence on he Swss reurn puzzle s also conssen wh he vew of Barro (2005), who shows ha allowng for low-probably dsasers explans a number of puzzles relaed o asse reurns and consumpon. Oher frendly leraure ncludes Joron and Goezmann (999), who offer a smlar explanaon for he equy premum puzzle. They showed ha he equy prce puzzle also dsappears for counres ha suffered caasrophc evens, n parcular he nerrupon of he sock exchange durng he wars. Over he las years we connued our search for evdence on he nsurance premum. No dscouraged by he peso problem, we se ou o denfy geopolcal evens, whch would no 5

8 qualfy as caasrophc bu mgh be large enough o brng abou some safe haven effecs.e. an apprecaon of he exchange rae. We repor our resuls for he record, snce he overall concluson was ha we faled o srke gold. Here s wha we red. Frs we looked for geopolcal evens whch were large and unexpeced. A large even would no only make all fron pages of newspapers bu would also be engraved n people s memores. Sepember s an example of a large even. I also qualfes as compleely unexpeced, whch s no he case of oher large evens lke he Iraq War. Our approach herefore was o use he daes of errors aacks and es wheher he Swss Franc responded wh a larger han normal apprecaon. I dd no. Nex we urned around and red a reverse even sudy. We defned large currency apprecaons and hen looked wheher hese paerns could be mached o sgnfcan geopolcal evens. We ook our cue from he currency crss leraure and defned apprecaon crss n several ways, on weekly, daly and monhly bass. Agan, we were unable o uncover a paern. Overall, we do no dscard he peso problem as a soluon o he puzzle. Afer all, s he naure of peso problems ha hey canno be found n he daa. Bu we do repor falure n comng up wh furher evdence n s favor. Carry rade and he puzzle Our expecaon was ha he neres rae puzzle would have ncreased over he pas decade for several reasons. Frs he nomnal neres rae dfferenal beween he Euro and he Swss Franc was que sable a around,5 percen. Second here was a prolonged perod n whch he Swss Franc deprecaed vs-à-vs he Euro. The deprecaon lased for more han 2 years and lead many observers o beleve ha he secular rend apprecaon of he Swss Franc vs-à-vs oher European currences had been broken. (Fgure ) 6

9 Fgure : Nomnal Ineres Rae Dfferenal and Exchange Rae Euro /Swss Franc 3 2,5 2,5 0, ,8,75,7,65,6,55,5,45,4 IRD o EURO SFEURSP Two salen feaures of he las years were a hsorcal low of fnancal marke volaly and broadly rendng exchange raes. In hs envronmen nvesors wh shor-erm horzons could earn aracve rsk-adjused reurns on foregn exchange markes deployng carry rade sraeges and usng he Swss Franc as a fundng currency. And even nvesors wh longer me horzons such as Easern European households ncreasngly engaged n carry rades by fnancng her morgages n Swss Francs. The ncreasng use of he Swss Franc as a fundng currency for carry rades s conssen wh he large growh of FX urnover as repored by he BIS. Fgure 2 shows ha he daly urnover of US Dollar / Swss Franc rose from 57 bllon n 200 o 43 bllon n The urnover Euro /Swss Franc ncreased from 2 o 54 bllon per day over he same perod. In fac, he growh rae n Euro/Swss Franc urnover s hgher han n any oher mayor currency par (see fgure 7). 7

10 Fgure 2: Daly averages n bllons of US dollars Swss Franc vs Dollar Swss Franc vs Euro Fgure 3: Growh rae of daly urnover on foregn exchange markes 400,00% Growh Vs-a-vs USD Growh Vs-a-vs Euro 350,00% 300,00% 250,00% 200,00% 50,00% 00,00% 50,00% 0,00% Euro Yen Pound Swss Franc Source: BIS 2007, Table B. 5 8

11 The evdence on he surge of carry rades n Swss Francs and he raw neres rae dfferenals wh he Euro renforced our expecaon ha he puzzle would have ncreased over he las decade. We now proceed o re-examnng he man sylzed facs for he perod Sylzed facs and srucural breaks In hs secon we consder he sylzed fac wh respec o one-monh (euro currency marke) nvesmen n Swss Franc (SFR), Canadan Dollar (CAD), EURO (DM before 999), Yen (JYN), Brsh Pound (UKP) and US Dollar (USD). We focus on he reurn dfferenal afer exchange rae adjusmen,, where s he home neres rae s he foregn currency neres rae and s s he log exchange rae (home currency per un of foregn currency). In Tables and 2 we repor he mean of he reurn dfferenals of all he combnaons of he sx currences for he perods and , respecvely. The home currency s ndcaed n he column,. e. he las column gves us he reurn dfferenal of he foregn currences wh respec o he Swss franc. The las columns of able and 2 reveal an neresng change beween hese wo perods. Before 999 we have for all fve currences consdered a hgher mean reurn han for he Swss Franc. Ths s no longer rue snce 999: Swss Franc nvesmen resuled n a hgher mean reurn han hose n Yen and US Dollar! The case of he Yen wh a change n mean from around 3% o.6% s n parcular srkng. 9

12 Table : Rae of Reurn Dfferenals (afer exchange rae adjusmen), Percen p.a. Mean 999: :0 and Sandard Errors n Parenhess, EURO JYN UKP USD SFR CAD (3.3) (4.9) (2.98) (2.77) (3.6) EURO 2.46 (3.70) (2.09) JYN (3.69) 0.75 (3.3) -.75 (3.04) UKP.00 (2.67) 0.70 (.3) -.63 (3.37). (2.40) USD 0.05 (3.5) Table 2: Rae of Reurn Dfferenals (afer exchange rae adjusmen), Percen p.a. Mean 980:0-998:2 and Sandard Errors n Parenhess, DEM JYN UKP USD SFR CAD.3 (2.85) -.6 (3.0) -.20 (2.64) (.08).97 (3.02) DEM (2.5) (2.2) -.52 (2.75) 0.84 (0.92) JYN 0.06 (2.86) 0.89 (2.94) 3.3 (2.50) UKP 0.84 (2.67) 3.9 (2.3) USD 2.36 (2.96) The sandard errors a relavely hgh because of he very hgh volaly of exchange rae correced reurns n boh perods. Thus no mean s ndvdually sascally sgnfcanly dfferen from zero. In our 2000 paper we esmaed a sysem of AR() equaons jonly for he reurn dfferenal for he Swss Franc agans 0 currences and could clearly rejec he jon hypohess of 0

13 zero nerceps and a zero AR() coeffcen, respecvely. Table 3 repors he esmaon resul for our fve currences and he perod Table 3: Resuls from AR() Models for Nomnal Reurn Dfferenals, Swss Franc vs-à-vs 5 Currences, Monhly Daa 999: :0,, = γ + δ (, 2 2 ) + e, SUR-Esmaes, Sandard Errors n Parenheses Currency γ δ CAD.95 (3.6) EURO 0.72 (.3) JYN -.74 (3.36) UKP.7 (2.39) USD 0.06 (3.3) (0.068) (0.085) (0.079) (0.073) (0.065) The Wald Tes ha all nercep erms and AR()-coeffcens are jonly zero canno be rejeced a all usual sgnfcance levels (ch-square value %. 4.75, margnal sgnfcance level of ). Therefore he puzzle of he reurn dfferenal s no longer presen over hs perod. When we esmae hs sysem for he perod we do fnd he orgnal puzzle, we ge a chsquare value of 4.02 for hs hypohess wh a margnal sgnfcance level of Moreover, we noe ha even he shor run valdy of UIP canno be rejeced as he reurn dfferenals appear o be mean zero whe nose processes. These resuls clearly pon o a break n he seres. In he sequel we wan o check wheher we can locae he break n me. To hs end we apply he Ba and Perron srucural break mehodology o our full daa se. Table 4 presens Ba and Perron s (998) sequenal es of he srucural sably of hese equaons esmaed for he nomnal reurn dfferenals over he full sample We

14 assumed a maxmum number of breaks equal o 5 and a mnmum dsance beween wo breaks equal o 2. However he resuls are robus, wh respec o reasonable varaons of hese parameers. Table 4 ndcaes only wo breaks n he nercep, namely for he Yen n 989:04 (only for he reduced 980 o 990 sample) and he US Dollar n 985:02, respecvely. If we consder breaks n nercep and slope jonly we fnd only one for he US Dollar n 987:2. Therefore, hese ess provde raher weak evdence for srucural breaks. Furhermore, f any, hey are locaed n he eghes and no nnees and do no concern European currences as we would expec from he nroducon of he Euro n 999. Table 4: Resuls of he Sequenal Ba-Perron Tes for Mulple Srucural Changes AR() Models for Nomnal Reurn Dfferenals, Swss Franc vs-à-vs 5 Currences (max. breaks = 5, mn. dsance = 2 monh), Monhly Daa 980:0-2008:0, = γ + δ (, 2 2 ) + e, Currency Break n nercep Break n nercep and slope Break pon F-sasc Break Pon F-sasc CAD 85:02 86:02 87: :02 86:08 82: DEM / EURO 8: 86:02 82: JYN 93:07 92:07 90:04 89:04 82:03 UKP 96:07 92:04 87:0 USD 85:02 87:2 89: : 83:03 82: 89:03 87: 90:04 93:07 92:07 96:07 92:04 86:09 85:02 87:04 95: , and ndcaes sgnfcance a he 0, 5 and % level, respecvely 2

15 In order o ge more nsghs we esmaed he AR() models separaely over he perod and appled he cusum es o he recursve resduals of hese esmaed equaons. The resuls obaned are gven n he Fgures 4 o 8. These ess pon o nsably of he parameer for he Yen, he Pound and he US Dollar. However, he model for he Canadan Dollar and he DM/Euro seres shows no sgn of break. These resuls confrm more or less hose of he Ba-Perron approach n he sense ha no break s found for he DM/Euro case. Fgure 4: CUSUM Tes for he recursve resduals of an AR() model of he reurn dfferenal beween Canadan Dollar and Swss Franc, 980:0-2008:0, = (, 2 2 ) (2.345) (0.055) CUSUM 5% Sgnfcance Fgure 5: CUSUM Tes for he recursve resduals of an AR() model of he reurn dfferenal beween Euro/DM and Swss Franc, 980:0-2008:0 3

16 , = (, 2 2 ) (0.734) (0.0558) CUSUM 5% Sgnfcance Fgure 6: CUSUM Tes for he recursve resduals of an AR() model of he reurn dfferenal beween Yen and Swss Franc, 980:0-2008:0, = (, 2 2 ) (2.0035) (0.0549) CUSUM 5% Sgnfcance 4

17 Fgure 7: CUSUM Tes for he recursve resduals of an AR() model of he reurn dfferenal beween UK Pound and Swss Franc, 980:0-2008:0 = (, 2 2 ), (.7204) (0.0540) CUSUM 5% Sgnfcance Fgure 8: CUSUM Tes for he recursve resduals of an AR() model of he reurn dfferenal beween US Dollar and Swss Franc, 980:0-2008:0 = (, 2 2 ), (2.28) (0.0537) CUSUM 5% Sgnfcance 5

18 Of course breaks of a lnear model could be caused by non-lneares. For nsance cenral banks could behave dfferenly wh respec o moderae or exreme changes n he exchange rae. These knds of non-lneares can be convenenly esed usng he Tsay (989)-es usng arranged AR() models wh observaons of he curren and lagged value of he reurn dfferenals sored n an ascendng order. If he model s lnear he recursve resdual of hs arranged AR() model s uncorrelaed wh he regressor under lneary. Ths hypohess s esed by regressng he recursve resdual on he regressor and he correspondng coeffcen s esed o be zero by an F-sasc. Table 5 repors he Tsay-sasc for our daa. We deec sgnfcan nonlneares for he Yen and he Pound. Thus n hese wo cases he breaks ndcaed wh he lnear models may be he resuls of asymmerc behavour of small and large devaons from UIP. However for he wo currences whch are cenral for Swzerland, namely Euro and US Dollar, no ndcaon of non-lneary s found and s herefore hghly unlkely ha our fndngs can be explaned by an asymmerc reacon of moneary polcy. Table 5: Tsay es for non-lneary for reurn dfferenals agans he Swss Franc 980:0 2008: 0 Currency F-sasc (p-value) CAD (0.6638) EURO (09275) JPN (0.039) UKP (0.029) USD.6773 (0.962), and ndcaes sgnfcance a he 0, 5 and % level, respecvely Our resuls sugges ha he naure of he puzzle has changed even f s dffcul o locae he me of change and he currences nvolved. The man feaure of he orgnal puzzle has dsappeared n he perod : The mean reurn dfferenal of Swss Franc asses s no longer posve wh respec o all mayor currences. Moreover, jonly he dfferenals wh all major currences are no sgnfcanly dfferen from zero. 6

19 Of course our ess sugges ha even he shor run volaon of UIP, he forward premum puzzle dsappeared. Therefore, we now urn o check hs hypohess n he usual bvarae regresson framework. In our prevous work we had found, as many ohers ha he forward premum puzzle was found n Swss Francs as well as n oher currences. Table 6 and 7 dsplay a remarkable dfference beween he wo perods: In he sample we deec posve and negave slope coeffcen whch are no sascally dfferen from zero whereas n he earler perod he coeffcens are negave and usually sgnfcan. The Wald Tes ha he nercep erm s zero and he coeffcen on he neres rae dfferenal s one canno be rejeced a all usual sgnfcance levels excep for he currency par Canadan Dollar / Swss Franc (ch-square value %. 5.44, margnal sgnfcance level of 0.066). The Wald-Tes ha only he coeffcen on he neres rae dfferenal s one can be rejeced a he 5 % sgnfcance level for Canadan Dollar / Swss Franc, US Dollar / Swss Franc and a he 0 % sgnfcance level for Euro / US Dollar, Canadan Dollar / Euro and he Euro/Pound. Table 6: Tes Resuls for UIP (Unbasedness of he Forward Premum), Monhly Daa 980:0-998:2, 5 Currences Δs = α + β ( + u ) OLS esmaes of α and β, sandard errors n parenheses Currency DEM JYN UKP USD SFR α β α β α β Α β α β CAD -5.4 (3.83) -.40 (0.95) (5.42) (.0) 5.83 (2.86) (.07) (.8) -.65 (0.55) -.63 (4.84) (0.9) DEM (2.94) -.76 (.0) 6.09 (3.68) -.0 (0.76).70 (3.04) (0.83) (.33) -.52 (0.66) JYN (8.9) (.56).88 (4.7) (0.99) 2.92 (2.40) (.00) UKP (3.40) (0.93) (5.) -.96 (0.84) USD (3.77) -.68 (0.77) Table 7: Tes Resuls for UIP (Unbasedness of he Forward Premum), 7

20 Monhly Daa 999: :0, 5 Currences Δs = α + β ( + u ) OLS esmaes of α and β, sandard errors n parenheses Currency EURO JYN UKP USD SFR CAD -.20 (3.44) α β Α β α β α β α β (3.37) (4.84) (4.08) (5.99) (4.26) (2.76) (2.95) (9.73) EURO 6.05 (5.04) 2.8 (4.79) 7.46 (5.6) JYN (26.94) (2.83) 9.80 (5.67) 2.3 (3.20) 5.95 (6.82) UKP (3.77) (.98) -.22 (.76) (2.3) (6.73) 2.83 (6.9) (2.3) USD (5.68) (4.22) 0.20 (3.88) (3.88) (3.59) (2.24) 4. Summary and Conclusons In hree earler papers (Kugler and Weder, 2002, 2004, 2005) we analyzed he long and shor run neres rae pary of 0 mayor OECD currences usng daa before 999 and found ha afer correcng for exchange rae changes, mean reurns on Swss asses were sgnfcanly lower han n oher currences, an anomaly no found n any oher major currency. We offered a possble explanaon for he reurn anomaly, namely ha may be due o an nsurance premum agans very rare evens, such as a major war. The las en years were characerzed by wo poenally very mporan moneary and polcal changes: frs, he replacemen of 5 somemes raher unsable and hgh nflaon currences by he relavely sable common new currency, he Euro. Second, he collapse of he Sove block n he early nnees decreased he probably of a large scale war as he wo world wars n he 20 h cenury. In such a world a sable safe haven currency lke he Swss Franc may become less aracve. Therefore we analyzed he volaons of UIP for he Swss Franc 8

21 agans he Dollar, he Euro, he Yen, he Pound and he Canadan Dollar usng an exended sample up o fall Ths exercse provdes he followng man resuls: The puzzle has dsappeared n s orgnal form,.e. mean reurns on Swss Franc asses are no longer sysemacally and sgnfcanly lower when compared jonly wh oher major currences n he 999 o 2008 perod. We faled o fnd evdence ha geopolcal crses lke errors aacks lead o an apprecaon of he Swss Franc. Neverheless we canno, however, dscard he hypohess of a peso problem snce here were no ruly caasrophc evens n he daa. Even he shor run valdy of UIP canno be rejeced for he Swss Franc agans he fve currences menoned above n he sense ha neres rae dfferenals correced for exchange rae changes appear o be whe nose. Ths resul was confrmed by he usual regresson of UIP for all 5 pars of currences whch provdes only very weak evdence agans UIP. Unforunaely our aemp o locae he me of change and he currences nvolved by he applcaon of ess for srucural breaks wh unknown break pon dd no lead o clear cu concluson for he Euro and he Dollar, he wo mos mporan foregn currences from a Swss perspecve. The same s rue for he applcaon of non-lneary ess for asymmerc reacons o small and large devaons from UIP. Probably we need more han en years of addonal daa n order o ge more precse resuls. 9

22 References: Ba, J. and Perron, P. (998), Esmang and Tesng Lnear Models wh Mulple Srucural Changes, Economerca 66, Black, F. and Lerman R. (992), Global Porfolo Opmzaon, Fnancal Analyss Journal, Sepember-Ocober, Barro, Rober (2005), Rare Evens and he Equy Premum NBER Workng Paper No. 30 Bank of Inernaonal Selemens (2007), Trennal Cenral Bank Survey on Foregn exchange and dervaves marke acvy n 2007, Basel Hodrck, Rober (2000), Inernaonal Fnancal Managemen, Englewood Clffs, Prence Hall (NY). Joron, Phlppe and Wllam N. Goezmann (999), Global Sock Markes n he Tweneh Cenury, The Journal of Fnance, Kugler, Peer and Bearce Weder (2002), The Puzzle of he Swss Ineres Rae Island: Sylzed Facs and a New Inerpreaon, Swss Revew of Inernaonal Economc Relaons 57,, Kugler, Peer and Bearce Weder (2004), Inernaonal Porfolo Holdngs and Swss Franc Asse Reurns 2004, Swss Journal of Economcs and Sascs 3, 2 Kugler, Peer and Bearce Weder (2005) Why Are Reurns on Swss Franc Asses So Low? Rare Evens may Solve he Puzzle ", 2005, Appled Economcs Quarerly 5,3, Tsay, R. S.: Tesng and Modelng Threshold Auoregressve Processes, Journal of he Amercan Sascal Assocaon 82,