Searching for Irving Fisher

Transcription

1 Searching for Irving Fisher Kris James Michener Deparmen of Economics Sana Clara Universiy and NBER Marc D. Weidenmier Deparmen of Economics Claremon McKenna College and NBER December 2008 There is a long-sanding debae abou he wheher here was a Fisher effec during he classical gold sandard period. We break new ground on his quesion by examining he ineres-rae differenial beween Ausrian paper and gold perpeuiy bonds o impue a marke-based measure of general inflaion expecaions during he gold sandard. Our empirical invesigaion of inflaion expecaions during he gold sandard period suggess several conclusions: (1) inflaion expecaions exhibied significan persisence a he weekly, monhly, and annual frequencies; (2) marke paricipans forecased a significan porion of boh he gold deflaion and inflaion episodes during he classical gold sandard period; and (3) inflaion expecaions are highly correlaed wih he nominal ineres raes. Using our unique measures of inflaion expecaions, we find evidence in favor of he operaion of he Fisher effec during he classical gold sandard period. Keywords: Fisher Effec, Inflaion Expecaions, Gold Sandard JEL Classificaion Numbers: E4, G1, N2 Michener: Deparmen of Economics, Sana Clara Universiy, 500 El Camino Real Sana Clara, CA 95050;kmichener@scu.edu; Weidenmier: Deparmen of Economics, Claremon McKenna College, Claremon, CA 91711; marc_weidenmier@cmc.edu;

2 Searching for Irving Fisher Several sudies have found ha ex pos inflaion raes are no correlaed wih he level of nominal ineres raes during he classical gold sandard period even hough he Fisher equaion is an empirical relaionship based on inflaion expecaions. 1 One possible explanaion for he failure of he Fisher equaion in his earlier period is ha inflaion expecaions were nearly zero given he low level of persisence in annual measures of ex pos inflaion raes (Barsky, 1987; Bordo and Kydland, 1995; Fisher, 1930; Friedman and Schwarz, 1982). 2 Oher possible explanaions for he lack of correlaion beween nominal ineres raes and inflaion raes include he hypoheses ha financial markes had money illusion or ha invesors did no undersand he quaniy heory of money (Summers, 1983; Cagan, 1984; Choudry, 1996; Barsky and Delong, 1991). Several scholars have aemped o esimae price and inflaion expecaions during he gold sandard period o es for he presence of a Fisher effec using ime series economeric models (Capie, Mills, and Wood, 1991). An obvious limiaion o his approach is ha we do no have a very good idea of he economic model ha marke paricipans used o form inflaion expecaions (Barsky and DeLong, 1988). Alernaively, some more recen sudies have used daa on agriculural fuures o derive a measure of inflaion expecaions (Siegler and Perez, 2003). Bu, as he auhors imporanly poin ou, rading in fuures markes was quie hin during he gold sandard 1 Many sudies have found ha nominal ineres raes during he gold sandard are correlaed wih he price level raher han he rae of inflaion. Barsky and Summers (1983) argue ha ha he posiive correlaion of nominal ineres raes wih he price level is a direc resul of he fac ha he price level in he inverse of he price of gold. Benjamin and Kochin (1984) argue ha Gibson s Paradox was spurious. 2 For a discussion of inflaion expecaions during he gold sandard period, see Barsky and DeLong (1983). For an analysis of he persisence of inflaion during he gold sandard period and laer, see Burdekin and Siklos (1996). Harley (1977) analyzes prices and ineres raes in he UK during he gold sandard period o es Gibson s Paradox. 1

3 and esimaion someimes requires inerpolaion echniques o consruc a coninuous imes series. In addiion, agriculural fuures only cover individual commodiies, and hence are based on he assumpion ha agriculural prices rack overall inflaion. This approach suffers from a failure o disinguish beween changes in relaive prices versus changes in he overall price level (Bordo and Schwarz, 1981). Summarizing he lieraure, McCallum (1984) poins ou ha sudies of pre-world War I changes in he price level have simply shown ha pas inflaion canno forecas fuure inflaion. Bu hese resuls say very lile abou inflaion expecaions or he failure of he Fisher effec during he gold sandard period. To improve upon he exising lieraure, we have colleced a unique daa se based on high-frequency, asse price daa ha allows us o compue a marke-based measure of inflaion expecaions for he 19 h cenury. In paricular, we use daa on Ausrian paper, gold, and silver governmen bonds o compue measures of inflaion expecaions a high and low frequencies during he classical gold sandard period o es for he presence of a shor and long-run Fisher effec. Ausria was he only major European counry during he classical gold sandard period ha issued gold, paper, and silver perpeuiy bonds ha acively raded on he leading financial exchanges of Europe including London, Paris, Berlin, Amserdam and Vienna. We use he ineres-rae differenial beween Ausrian paper and gold bonds o derive an ex ane marke-based measure of inflaion expecaions. We sugges ha gold bonds were essenially inflaion-indexed securiies, much like modern-day TIPS, whereas paper bonds required an inflaion premium o make he deb obligaions aracive o invesors in inernaional capial markes. Our empirical analysis suggess ha inflaion 2

4 expecaions were no a whie noise process during he gold sandard period. We find economically meaningful persisence in inflaion expecaions a he weekly, monhly, and annual frequencies ha can be used o forecas realized inflaion. As prediced by he Fisher equaion, we also find ha expeced inflaion was posiively relaed o he level of nominal ineres raes. Inflaion expecaions for Ausria also Granger-cause changes in he gold-silver price raio. Using our unique measures of inflaion expecaions, we find evidence in favor of he operaion of he Fisher effec during he classical gold sandard period. We begin wih a discussion of our daa on Ausrian bonds during he classical gold sandard period. We hen analyze he ime series properies of inflaion expecaions during our sample period in Ausria. We hen es he shor and long-run Fisher effec using ineres rae daa from Ausria and oher counries during he gold sandard. The las secion discusses our findings and heir implicaions for he lieraure on he Fisher effec. II. Ausria and he Gold Sandard Some scholars have characerized nineeenh-cenury Ausria as a financially underdeveloped, agriculurally-oriened economy, suggesing ha i belonged o he periphery raher han o he core of European gold sandard counries like he Unied Kingdom, France, or Germany. Alhough is GDP per capia was lower han he UK, Ausria s sandard of living compares favorably wih France and Germany prior o he oubreak of World War I (see Figure 1). Ausria was also one of he leading European 3

5 miliary powers of he lae nineeenh cenury and is financial markes appear o have been well developed and inegraed (Good, 1977). Afer 1816, i had a cenral bank, modeled afer he Bank of France, which had sole righ of noe issue and a nework of join sock banks wih exensive branching neworks (including he imporan Viennese banks) ha len o businesses hroughou he Ausrian empire. I also borrowed from he German model of universal banking in forming insiuions such as he Crediansal fur Handel und Gewerbe (1855). As for exchange rae policy, Ausria was a member of he silver sandard for much of he nineeenh cenury. The Compromise of 1867 beween Ausria and Hungary gave consiuional foundaions for a moneary union wih he silver florin as he moneary sandard. Afer 1879, he florin was no longer converible ino silver and he exchange-rae sysem ofen resembled a floa more han a peg. Silver florin raded for as much as seven percen away from he min par raio and, as shown in Figure 2, weekly exchange raes exhibied significan flucuaions in he 1880s. In Augus 1892, Ausria joined he gold sandard and esablished a new currency, he crown (kronen). Flandreau and Komlos (2001) argue ha even hough Ausria never formally esablished gold converibiliy prior o World War I, he counry was a de faco member of he gold sandard because of he sabiliy of is exchange rae. The Ausrian crown exchange rae flucuaed only /- 0.4 percen from min par afer Figure 2 confirms ha Ausrian exchange raes were remarkably sable afer he counry joined he gold sandard. Based on he behavior of he exchange rae, Flandreau and Komlos (2001) conclude ha Ausria was a counry ha was neiher a core nor peripheral member of he gold sandard, bu raher somewhere in beween. 4

6 However, addiional hisorical evidence from bond markes suggess ha Ausria may have more closely resembled a core gold sandard counry. In conras o periphery counries, i successfully floaed large amouns of governmen deb hroughou Europe in is own currency. I did no suffer from so-called original sin (Eichengreen and Hausmann, 1999; Bordo, Meissner, and Redish, 2005). Only he UK, France, Germany, Neherlands, and he Unied Saes were also able o sell large bond issues in heir home currency on several European markes during he classical gold sandard period. Ausria apped inernaional capial markes on a significan scale following he passage of he Law of March 16, The legislaion auhorized a 16 million florin bond issue ha was exemp from Ausrian axes and paid ineres half-year in gold in Vienna and oher European exchanges including Amserdam, Berlin, Brussels, Frankfur, and Paris. Morys (2008) esimaed ha foreigners held approximaely 20 percen of he deb issue. Ausria also issued paper and silver bonds on he leading European exchanges. Like he gold bonds, he wo deb issues did no conain a sinking-fund and had a fivepercen coupon. The bonds were perpeuiy obligaions and subjec o a 16 percen income ax. The coupon paymens on he paper bonds were payable half-yearly on February or Augus 1 s or on May 1 s and November 1 s (Sock Exchange Official Inelligence, various issues). The marke value of unredeemed paper bonds exceeded 443fl million in The silver deb, issued in 1868, also had hundreds of million fl ousanding. We have assembled weekly prices of he paper, gold, and silver sovereign deb issues of Ausria over he period We employ hese bond series o derive a marke-based measure of inflaion expecaions and o es for he presence of a Fisher effec during he gold sandard period. 3 We are presenly working on exending he daa o cover he enire classical gold sandard period. 5

7 III. Empirical Analysis A. Model The Fisher equaion saes ha he nominal ineres raes on a given sovereign deb obligaion is equal o he real ineres rae plus he expeced rae of inflaion. The nominal ineres rae for Ausrian paper bonds can be wrien as: (1) P i = r π e, P where r is he real ineres rae and π e, P equaion for Ausrian gold bonds can be wrien as is he expeced rae of inflaion. The Fisher (2) G i = r π e, G However, unlike he paper bond, since his deb obligaion is payable in gold, inflaion expecaions, denoed by π e, G, are assumed o be zero. The paper-gold ineres-rae differenial can be obained by subracing equaion (2) from equaion (1). (3) P G i = π e Equaion (3) saes ha he paper-gold ineres rae spread is equal o he expeced rae of inflaion. To carry ou he empirical analysis below, we make hree assumpions abou he bonds and invesor behavior: (1) invesors are risk-neural; (2) he real ineres rae is he same for boh bonds given ha he Ausrian governmen issued he wo deb obligaions; and (3) paper and gold bonds have idenical defaul risk. The hird assumpion of idenical defaul risk appears reasonable given ha Ausria faihfully repaid is sovereign deb during he enire gold sandard period. Even if here were some differenial defaul 6

8 risk beween paper and gold bonds, he premium is no likely large enough o have a qualiaive impac our analysis. The impued measure of inflaion expecaions, π e, along wih Ausrian paperbond ineres raes are presened in Figure 3. The simple correlaion coefficien beween his nominal ineres rae and inflaion expecaions is more han 90 percen. Inflaion expecaions over he enire sample period averaged 1.38 percen (138) basis poins and accouned for approximaely 20 percen of he nominal ineres rae (inflaion expecaions/nominal ineres rae). Since inflaion expecaions were relaively sable (he sandard deviaion is less han hree percen), he empirical evidence also indicaes, as suggesed by Siegel and Shiller (1977), ha movemens in real ineres raes were probably more imporan han inflaion expecaions in driving flucuaions in Ausrian nominal ineres raes. B. Persisence in Inflaion Expecaions To examine he ime-series properies of inflaion expecaions during he classical gold sandard period, we firs es for a uni roo using he Augmened-Dicker Fuller es. The null hypohesis of a uni roo could no be rejeced a he five- or enpercen level of significance. 4 Figure 3, however, suggess ha he adopion of he gold sandard may have led o a srucural break in he formaion of inflaion expecaions as he counry implemened moneary and fiscal reforms in he lae 1880s, and which in urn reduced he level of nominal ineres raes. 4 The es saisic for he ADF es was

9 8 To es formally for a srucural break, we employ he mehodology developed by Zivo-Andrews (1992). This es allows us o examine wheher adoping he gold sandard lowered he level, rend, or level and rend of inflaion expecaions. The Zivo-Andrews procedure does no specify a breakpoin, a priori, given ha he pre-selecion of a change poin biases he resuls owards finding a srucural break (Chrisiano, 1992). The null hypohesis of he srucural break model, H 0, is ha inflaion expecaions are a nonsaionary process: (4) e e e = π μ π where μ is he mean of a given ime series and e is a whie noise disurbance erm. Three differen alernaive hypoheses, 2,3,4 1 H, can be wrien as follows: k j j j k j j j k j j j e y c y DT DU y e y c y DT y e y c y DU y Δ = Δ Δ = Δ Δ = Δ = = = ) (6) (5) α γ θ β μ α γ β μ α θ β μ Equaion (5) is a crash model ha allows for a one-ime change in he level of inflaion expecaions a a break dae denoed by TB. This specificaion capures he large drop in inflaion expecaions in he lae 1880s early 1890s. Equaion (6) is used o es for saionariy around a broken rend a TB. This specificaion allows for a gradual

10 decline in inflaion expecaions if he good housekeeping effec of he gold sandard is incorporaed ino ineres raes over a long period of ime. Equaion (7) is he mos general specificaion ha allows for a change in he inercep and rend of inflaion expecaions a TB. DU is a dummy variable ha capures he shif in he inercep and akes a value of 1 if >TB. DT is anoher indicaor variable ha represens he shif in he deerminisic rend a ime TB. DT is equal o (-TB) if (>TB) and zero oherwise. To conrol for serial correlaion, we also included lagged differences of he dependen variable as covariaes in he hree models. The number of lagged differences employed in he break ess is seleced on he basis of he Akaike Informaion Crieria (AIC). In each of he e hree alernaive hypoheses, π is assumed o be a saionary process wih one srucural break. The null hypohesis is rejeced if he α coefficien is significanly differen from zero. The empirical resuls for he srucural break es are presened in Table 1. The analysis suggess ha he adopion of he gold sandard led o a srucural change in boh he rend and level of inflaion expecaions a he one-percen level of significance for Ausria on Augus 4, 1888; however, we are unable o rejec he null hypohesis of a srucural break for he crash and rend break models. Based on he resuls of he srucural break analysis, we hen esimae ARIMA models for he periods when Ausria was a member of he silver and gold sandards o measure he persisence of inflaion expecaions. Ausria was a member of he silver sandard from he 1848 Revoluion unil July 1892 and a member of he gold club from Augus 1892 unil he oubreak of World War I. 9

11 In order o avoid conaminaing he empirical resuls wih he large drop in inflaion expecaions ha accompanied he economic reforms, we esimae ARIMA models for he silver sandard over he sample period January 1880 o Augus 1888 he period before he srucural break. In he lae 1880s and early 1890s, Ausria, reduced is budge defici, acquired gold, and reired ousanding governmen paper noes as i prepared o adop he gold sandard (Flandreau and Komlos, 2001). As shown in Table 2, inflaion expecaions in Ausria are bes characerized by an AR(2) process when i was a member of he silver sandard. The sum of he auoregressive coefficiens, which has he value of.92, indicaes a very high level of persisence in inflaion expecaions. The coefficien on he consan erm in he equaion suggess ha financial marke paricipans expeced inflaion o average approximaely 1.76 percen per year. For he gold sandard period, we esimae ARIMA models from Augus 1892 o April 1903, and find ha he bes model for inflaion expecaions is an AR(1) model. Table 2 shows ha he level of inflaion persisence drops from abou 95 percen o abou 92 percen. Alhough we observe a similar level of persisence in inflaion expecaions afer Ausria joined he gold sandard, here is a marked reducion in he average level of inflaion expecaions from 1.73 percen o 1.14 percen. These findings sugges ha here was significan persisence in inflaion expecaions and ha joining he gold reduced he average level of inflaion expecaions by roughly 40 percen. One poenial problem wih he empirical resuls is ha he persisence of inflaion expecaions is simply a resul of using high frequency daa impued from weekly ineres raes. To consider his possibiliy, we re-esimaed he baseline empirical resuls using end-of-monh daa. The ARIMA models of monhly inflaion expecaions are presened 10

12 in Table 3. The resuls are similar o hose employing he weekly daa. Inflaion expecaions followed an AR(2) process when he counry adhered o he silver sandard and inflaion persisence is sill nearly 90 percen and significan a he one-percen level. For he gold sandard period, inflaion expecaions are sill bes modeled as an AR(1) process. Alhough inflaion expecaions are once again no a whie noise process, he coefficien on he auoregressive erm falls by abou 30 percen, from 92 o almos 70 percen. We find similar resuls using annual daa even hough he sample period is quie shor and he uni roo es for he annual daa on inflaion persisence is only marginally saisically significan. 5 Table 4 shows ha inflaion expecaions averaged more han 1.3 percen over he period Again, we find ha here is significan persisence wih he sum of he wo auoregressive erms greaer han Hence, using weekly, monhly, or annual daa, we find subsanial persisence in inflaion expecaions during he gold sandard period. One possible criique of our analysis is ha Ausria s commimen o gold migh have been perceived as less credible, and hence he analysis of inflaion expecaions we derive for i may no be very represenaive of gold club members. Tha is, inflaion expecaions for non-credible members of he gold sandard may be much larger han for counries ha sricly adhered o he moneary rule. While Ausria was a newer member of he gold sandard in comparison o France, Germany, and he UK, i does no appear ha marke paricipans viewed is commimen o gold as subsanially less credible han hese counries. Michener and Weidenmier (2008) provide evidence ha Ausria was one 5 The null hypohesis of a uni roo can only be rejeced a he 15 percen level of significance. We hope o exend he daabase o 1913 in fuure drafs of he manuscrip. 11

13 of he mos credible gold sandard moneary regimes during he period : marke paricipans expeced Ausrian kroner o depreciae approximaely hree percen afer he counry joined he gold sandard. The level of expeced depreciaion is considerably smaller han several oher gold sandard counries including he Unied Saes, Argenina, Brazil, Chile, India, Mexico, and Russia. We herefore inerpre our resuls as providing a lower bound on he size and persisence of inflaion expecaions for he average counry during he classical gold sandard period. C. Forecasibiliy of Inflaion The presence of significan persisence in inflaion expecaions suggess ha he ineres-rae differenial beween paper and gold bonds migh be able o predic fuure inflaion raes during he gold sandard. One problem wih esing his hypohesis during he gold sandard is ha price indices from his period do no always provide useful informaion. Governmens did no regularly collec informaion on he goods and services people purchased on a monhly basis (Perez and Siegler, 2003; Hanes, 1999). 6 This makes i difficul o consruc reliable consumer price indices ha can be compared o a general measure of inflaion expecaions impued from financial markes. A possible soluion for dealing wih he problem of incomplee price figures is o examine he ineres-rae differenial beween Ausrian gold and silver bonds. This yield spread should reflec inflaion expecaions regarding he gold-silver price raio, a relaive price ha was widely followed by invesors and repored on by he major financial newspapers during 6 Hanes (1999) consruced a consisen consumer price index series for he Unied Saes from For he gold sandard period, he consrucs a consumer price index based on 1911 survey from he Unied Saes Bureau of he Census. 12

14 he gold sandard. The monhly gold-silver price raio, shown in Figure 4, sharply declines in he firs par of he sample period and hen flucuaes up and down from abou 1894 unil he oubreak of World War I. The gold-silver ineres-rae differenial, our measure of inflaion expecaions, also appears in Figure 4. The correlaion coefficien beween he wo price series is approximaely 60 percen. To es wheher gold-silver inflaion expecaions can forecas acual movemens in he relaive price, we firs esed he gold-silver price raio for a uni roo. The null hypohesis of a uni roo could easily be rejeced a he one percen level of significance. 7 We hen ran a series of Granger-causaliy ess o es if inflaion expecaions impued from he silver-gold ineres rae differenial can predic movemens in changes in he gold-silver price raio. We esimaed VARs wih lags lenghs of one o six for he periods when Ausria was a member of he gold and silver sandard. The resuls for he silver sandard period sugges ha inflaion expecaions do no Granger-cause changes in he gold-silver price raio (Table 5). However, we do find some evidence ha changes in he gold-silver price raio Granger-causes inflaion expecaions. We hope o provide more perspecive on he relaionship beween he iming of flucuaions in inflaion expecaions and changes in he gold-silver price raio wih higher frequency, weekly daa. For he gold sandard period, we find ha he ineres-rae differenial beween Ausrian silver and gold bonds Granger-causes changes in he gold-silver price raio for lag lenghs of one o hree. In all six VARs, he coefficien on he firs lag of he ineresrae spread in he change in he gold-silver price raio regression is wo percen and saisically significan a he one percen level. This suggess ha financial marke 7 The es saisics for Augmened Dicker-Fuller es was nearly -15 wih a p-value of

15 paricipans used inflaion expecaions o help forecas movemens in he gold-silver price raio. D. Fisher Effec 1. Shor-Run Tess To es for he presence of a shor-run Fisher Effec, we analyze he relaionship beween he Vienna Open marke ineres rae and our measure of inflaion expecaions. 8 The Vienna open marke rae is he mos imporan shor-erm money marke rae used o conduc rade wih oher counries. The Ausrian shor-erm rae along wih he papergold ineres rae spread is shown in Figure 5. Unlike inflaion expecaions, here is no visual evidence of a srucural break in he shor-erm ineres rae series. The lack of a srucural break may be explained by he fac ha he Vienna open marke securiy was denominaed in gold. Hence joining he gold sandard would have lile impac on he money marke ineres rae. To analyze he relaionship beween shor-erm ineres raes and inflaion expecaions in more deail, we esimae a series of Granger-causaliy ess. We firs es he shor-erm ineres rae series for a uni roo using he Augmened Dickey-Fuller es. We can easily rejec he null hypohesis of no uni roo a he one-percen level of significance. 9 As shown in Figure 5, he shor-erm open marke appears o be mean revering for Ausria. Given ha we idenified a srucural break in he ime series of inflaion expecaions, we esimae Granger-causaliy ess for he sample periods when 8 For sudies examining he Fisher effec using modern daa, see Mishkin (1981, 1992) and Fama (1975). 9 The fac ha he null hypohesis of a uni roo can easily be rejeced for he shor-erm ineres rae series shows ha here is no a srucural break in he ime series. 14

16 Ausria was a member of he silver (January 1880-July 1888) and gold sandards (Augus 1892-April 1903). We esimae Granger-Causaliy ess wih lag lenghs from one o six for he wo sample periods. The resuls appear in Tables 6. In all six VARs for he wo periods, we are unable o rejec he null hypohesis ha inflaion expecaions Grangercause shor-erm ineres raes a he five or en percen level of significance. One possible explanaion for he lack of Granger-causaliy is ha he Ausrian sovereign deb and money markes are very efficien and clear quie quickly. If his were rue, we should no expec lagged values of inflaion expecaions o predic shor-erm ineres raes. To consider his possibiliy, we calculae he forecas error variance of shor-erm ineres raes and inflaion expecaions using a sandard Choleski decomposiion o deermine he exen o which flucuaions in inflaion expecaions drove shor-erm ineres raes for Ausria. We give inflaion expecaions he firs ordering in he sysem so ha shocks o inflaion expecaions have a conemporaneous effec on shor-erm ineres raes, bu shor-erm ineres raes do no have a conemporaneous impac on inflaion expecaions. We esimae a VAR using six lags o consruc he variance decomposiion. We find ha inflaion expecaions can accoun for less han four percen of he movemens in shor-erm ineres raes for each week up o a 36-week forecas horizon. The evidence suggess ha here is a weak link beween inflaion expecaions and shor-erm ineres raes for Ausria. We believe ha his simply reflecs he fac ha shor-erm ineres raes during he gold sandard were ofen denominaed in hard currency. As a resul, marke paricipans considered shor-erm ineres raes o be inflaion-indexed securiies. 15

17 2. Long-Run Fisher Effec We es for he presence of a long-run Fisher effec using Ausrian silver bonds. The nominal ineres rae should provide some insigh ino wheher here was a relaionship beween nominal ineres raes and expeced inflaion. Since silver became largely demoneized in he early 1890s, he precious meal funcioned more as commodiy as opposed o a meal used as a currency. 10 Figure 6 shows Ausrian inflaion expecaions along wih he silver currency bonds for he period when he European counry was on he gold sandard. The wo series move ogeher over he enire sample, suggesing a srong posiive link beween nominal ineres raes and expeced inflaion. We examine he relaionship beween Ausrian inflaion expecaions and he ineres rae on silver bonds using Granger-causaliy ess. We firs es he ime series of silver ineres raes for a uni roo using he Augmened Dickey-Fuller es. The null hypohesis of a uni roo canno be rejeced using eiher a consan or consan and rend in he model. As discussed earlier, inflaion expecaions calculaed using he paper-gold differenial are saionary regardless of wheher a rend is specified in he model. Alhough he wo ime series have differen orders of inegraion based on uni roo ess, a long-run equilibrium relaionship may neverheless sill exis. To es for one, we look for coinegraion using he Engle-Granger es. Wih a es saisic of 3.6, we can rejec he null hypohesis of no coinegraion a he five-percen level. We inerpre he resuls as srong evidence of he presence of a Fisher effec a long-run relaionship beween 10 A few counries and colonies coninued o use silver hroughou he gold sandard period including China and Hong Kong. 16

18 inflaion expecaions and nominal ineres raes. In he fuure, we would like o expand our analysis of he Fisher effec o include paper currency bonds issued by leading firms on he Vienna Bourse. An analysis of corporae paper bonds would provide addiional informaion on he link beween nominal ineres raes and inflaion expecaions during he classical gold sandard period. V. Conclusions Economiss have now enered a second cenury of searching for Irving Fisher. We believe we have found him, lurking in he inflaion expecaions of he classical gold sandard period. We discovered he presence of he Fisher effec by providing one of he firs, high-frequency marke-based measures of general inflaion expecaions for he classical gold sandard period so ha we could acually es for he presence of such an effec. Previous sudies have used gold bonds and economeric models o examine he relaionship beween nominal ineres raes and inflaion. We believe ha our measure, he ineres rae differenial beween gold and paper bonds, provides a more direc approach for sudying he behavior of inflaion expecaions during he gold sandard period and is relaion o acual inflaion. Our analysis of inflaion expecaions suggess several conclusions. Firs, he adopion of he gold sandard led o a srucural change in he average level of inflaion expecaions in Ausria. Joining he gold sandard led o a 40 percen drop in inflaion expecaions, from 1.7 percen o 1.2 percen, as measured by decisions made in financial markes. We also find ha here is considerable persisence in inflaion expecaions a 17

19 he weekly, monhly, and annual frequencies. Marke paricipans clearly forecas a significan porion of gold deflaion and gold inflaion as well as he swich from a declining price level o an increasing price level in he mid-1890s as shown by our analysis of he gold-silver price raio. Finally, we find evidence of a long-run Fisher effec: silver ineres raes have a long-run equilibrium relaionship wih inflaion expecaions ha were derived using he paper-gold ineres-rae differenial. Marke paricipans during he classical gold sandard period required an inflaion premium ha was buil ino nominal ineres raes as long as he deb obligaion was denominaed in paper raher han gold. 18

20 References Barsky, Rober. (1987). The Fisher Hypohesis and he Forecasibiliy and Persisence of Inflaion. Journal of Moneary Economics 19, Barsky, Rober and J. Bradford Delong. (1991). Forecasing Pre-World War I Inflaion: The Fisher Effec and he Gold Sandard. Quarerly Journal of Economics 106, Benjamin, Daniel and Levis Kochin. (1984). War, Prices, and Ineres Raes: A Marial Soluion o Gibson s Paradox. in, A Resrospecive on he Classical Gold Sandard, (Michael Bordo and Anna Schwarz, eds.), Bordo, Michael and Finn Kydland. (1995). The Gold Sandard as a Rule: An Essay in Exploraion. Exploraions in Economic Hisory 32(4): Bordo, Michael, Chrisopher Meissner, and Angela Redish. (2005). How Original Sin was Overcome: The Evoluion of Exernal Deb Denominaed in Domesic Currencies in he Unied Saes and he Briish Dominions In Oher People s Money: Deb Denominaion and Financial Insabiliy in Emerging Marke Economies, Barry Eichengreen and Ricardo Hausmann (eds.), Chicago: Universiy of Chicago Press, pp Bordo, Michael and Anna Schwarz. (1981). Money and Prices in he Nineeenh Cenury: Was Thomas Tooke Righ? Exploraions in Economic Hisory 18: Burdekin, Richard C.K. and Pierre Siklos. (1996). Exchange Rae Regimes and Shifs in Inflaion Persisence: Does Nohing Else Maer? Journal of Money, Credi, and Banking 31(2): Cagan, Philip. (1984). Mr. Gibson s Paradox: Was I There? A Resrospecive on he Classical Gold Sandard, (Michael Bordo and Anna Schwarz, eds.), Capie, Forres H., Terrence C. Mills, and Geoffrey E. Wood. (1991). Money, ineres raes, and he Grea Depression: Briain from 1870 o 1913, in New Persepecives on he lae Vicorian Economy (James Foreman-Peck, ed.), Choudry, Tafiq. (1996). The Fisher Effec and he Gold Sandard: Evidence from he USA. Applied Economic Leers 3, Dwyer, Gerald. (1984). The Gibson Paradox: A Cross Counry Analysis. Economica. 51: Economis, various issues. Eichengreen, Barry and Ricardo Hausmann. (1999). Exchange Raes and Financial Fragiliy. NBER Working Paper

21 Fama, Eugene F. (1975). Shor-erm Ineres Raes as Predicors of Inflaion. American Economic Review 65, Fisher, Irving (1930). The Theory of Ineres. New York, NY:Macmillan. Flandreau, Marc and John Komlos (2001). Core or Periphery? The Credibiliy of he Ausro-Hungarian Currency. CEPR Working Paper. Flandreau, Marc and Frederic Zumer. (2004). The Making of Global Finance, OECD: Paris. Friedman, Milon and Anna J. Schwarz. (1982). Moneary Trends in he Unied Saes and he Unied Kingdom, Their Relaion o Income, Prices, and Ineres Raes, Chicago, IL: Universiy of Chicago Press. Good, David F. (1977). Financial Inegraion in Lae Nineeenh Cenury Ausria. Journal of Economic Hisory. 37(4), Hanes, Chrisopher. (1999). Degrees of Processing and Changes in he Cyclical Behavior of Prices in he Unied Saes. Journal of Money, Credi, and Banking. 31, Harley, C. Knick. (1977). The Ineres Rae and Prices in Briain, : A Sudy of he Gibson Paradox. Exploraions in Economic Hisory 14: McCallum, Benne (1984). On Low-Frequency Esimaes of Long-Run Relaionships in Macroeconomics. Journal of Moneary Economics 14(1):3-14. Mishkin, Frederic S. (1981). The Real Ineres Rae: An Empirical Invesigaion. Carnegie-Rocheser Conference Series on Public Policy 15, Mishkin, Frederic S. (1992). Is he Fisher Effec for Real? A Reexaminaion of he Relaionship beween Inflaion and Ineres Raes. Journal of Moneary Economics 30, Michener, K. and Marc D. Weidenmier (2008). Are Hard Pegs Ever Credible in Emerging Markes? Sana Clara Universiy Working Paper. Morys, Mahias. (2008). The Classical Gold Sandard in he European Periphery: A Case Sudy of Ausria-Hungary and Ialy, Unpublished Docoral Disseraion, London School of Economics and Poliical Science. Perez and Siegler, Inflaionary Expecaions and he Fisher Effec prior Pre World War I. Journal of Money, Credi, and Banking 35(6):

22 Shiller, Rober and Jeremy Siegel. (1977). The Gibson Paradox and Hisorical Movemens in Real Ineres Raes. Journal of Poliical Economy 85: Sock Exchange Official Inelligence, Various Issues. Summers, Lawrence H. (1983). The Nonadjusmen of Nominal Ineres Raes. In Macroeconomics, Prices, and Quaniies, edied by James Tobin, pp Washingon, DC: Brookings Insiue. Zivo, Eric and Donald Andrews. (1992). Furher Evidence on he Grea Crash, Oil Price Shock, and he Uni Roo Hypohesis. Journal of Business and Economic Saisics 10:

23 Table 1. Zivo-Andrews Srucural Break Tess, Break Tes Minimum T-Tes Break Dae Inercep /8/4 Trend /7/5 Inercep and Trend *** 1888/8/4 Observaions 1,219 * significan a 10%; ** significan a 5%; *** significan a 1% Table 2. ARIMA Models of Inflaion Expecaions (Weekly Daa) Silver Sandard Gold Sandard Consan *** (10.864) *** (2.280) AR(1) 0.759*** (0.046) 0.930*** (0.016) AR(2) 0.209*** (0.047) Observaions * significan a 10%; ** significan a 5%; *** significan a 1% Table 3. ARIMA Models of Inflaion Expecaions (Monhly Daa) Silver Sandard Gold Sandard Consan *** (20.075) *** (2.311) AR(1) 0.659*** (0.098) 0.729*** (0.061) AR(2) 0.281*** (0.100) Observaions * significan a 10%; ** significan a 5%; *** significan a 1% Table 4. ARIMA Models of Inflaion Expecaions (Annual Daa) Whole Period Consan *** (17.532) AR(1) 1.226*** (0.207) AR(2) ** 22

24 (0.209) Observaions 22 * significan a 10%; ** significan a 5%; *** significan a 1% Table 5 Forecasing he Gold-Silver Price Raio Granger-Causaliy Tess (F-ess) Lags Silver Sandard Gold Sandard *** ** * * significan a 10%; ** significan a 5%; *** significan a 1% Table 6 Shor-Term Fisher Effec Granger-Causaliy Tess (F-ess) Lags Silver Sandard Gold Sandard *significan a 10%; ** significan a 5%; *** significan a 1% 23

25 6,000 5,000 4,000 3,000 2,000 1,000 0 Figure 1 GDP Per Capia, Source: Maddison Yearly Inervals Ausria France Germany UK GDP Per Capia

26 Figure 2 Franco-Ausrian Exchange Rae Source: Flandreau and Zumer (2004) Annual Inervals

27 Figure 3 Ausrian Inflaion Expecaions and Nominal (Paper) Ineres Raes, Jan April Jan-80 3-Jan-81 3-Jan-82 3-Jan-83 3-Jan-84 3-Jan-85 3-Jan-86 3-Jan-87 3-Jan-88 3-Jan-89 3-Jan-90 3-Jan-91 3-Jan-92 3-Jan-93 3-Jan-94 3-Jan-95 3-Jan-96 3-Jan-97 3-Jan-98 3-Jan-99 3-Jan-00 3-Jan-01 3-Jan-02 3-Jan-03 Sources: Economis and The Times Weekly Inervals Inflaion Expecaions Paper Ineres Raes Basis Poins

28 Figure 4 Silver Inflaion Expecaions and Gold-Silver Price Inflaion, 1880-April Jan-80 Jan-81 Jan-82 Jan-83 Jan-84 Jan-85 Jan-86 Jan-87 Jan-88 Jan-89 Jan-90 Jan-91 Jan-92 Jan-93 Jan-94 Jan-95 Jan-96 Jan-97 Jan-98 Jan-99 Jan-00 Jan-01 Jan-02 Jan-03 Weekly Inervals Inflaion Expecaions Gold-Silver Inflaion Inflaion Expecaions (Basis Poins) Inflaion (Percen)

29 Figure 5 Shor-Term Ineres Raes and Inflaion Expecaions, 1880-April Inflaion Expecaions Shor-Term Ineres Rae Inflaion Expecaions 3-Jan-80 3-Jan-81 3-Jan-82 3-Jan-83 3-Jan-84 3-Jan-85 3-Jan-86 3-Jan-87 3-Jan-88 3-Jan-89 3-Jan-90 3-Jan-91 3-Jan-92 3-Jan-93 3-Jan-94 3-Jan-95 3-Jan-96 3-Jan-97 3-Jan-98 3-Jan-99 3-Jan-00 3-Jan-01 3-Jan-02 3-Jan-03 Shor-Term Ineres Raes

30 Figure 6 Inflaion Expecaions and Silver Ineres Raes, Augus 1892-April Aug-92 6-Feb-93 6-Aug-93 6-Feb-94 6-Aug-94 6-Feb-95 6-Aug-95 6-Feb-96 6-Aug-96 6-Feb-97 6-Aug-97 6-Feb-98 6-Aug-98 6-Feb-99 6-Aug-99 6-Feb-00 6-Aug-00 6-Feb-01 6-Aug-01 6-Feb-02 6-Aug-02 6-Feb-03 Weekly Inervals Inflaion Expecaions Silver Ineres Raes Basis Poins