9 views

Uploaded by cheema

- New York Metro Data Analysis
- Basic Statistics in Bio Medical
- 1.__DMAIC_-List_of_Tools.ppt
- Statistics
- Effect of Promotion Strategies on the Organizational Commitment of Banking Sector Employees in Kenya
- Predstavitev Rezultatov V5-0451 (a)
- Determinants of Financial Development
- FRM2009Syllabus
- School Principals Leadership Style.pdf
- UT Dallas Syllabus for gisc6379.501.08s taught by Michael Tiefelsdorf (mrt052000)
- Kine Study
- The First Fifty Years of Modern Econometrics
- BSST
- Financial Econometrics Outline Copy
- Hypothesis Testing
- Ba Yes Freq Book
- ECN225sol4.pdf
- Final Project Memo_QNT550
- statment of purpose
- perron1989.pdf

You are on page 1of 10

EMPIRICAL

ECONOMICS

( Springer-Verlag 2001

and interest rates: A reexamination of the Fisher e¨ect*

Markku Lanne

Research Unit on Economic Structures and Growth, Department of Economics, P.O. Box 54

(Unioninkatu 37), FIN-00014 University of Helsinki, Finland

(e-mail: markku.lanne@helsinki.®)

Abstract. Tests of the Fisher e¨ect are plagued by high persistence in interest

rates. Instead of standard regression analysis and asymptotic results, methods

relying on local-to-unity asymptotics are employed in testing for the Fisher

e¨ect with monthly U.S. data covering the period 1953:1±1990:12. These

procedures are extensions of a recently presented method (Cavanagh, Elliott

and Stock (1995)) based on simultaneous con®dence intervals, and they have

the advantage of being asymptotically valid whether interest rates are integ-

rated of order one or zero, or near unit root processes. Taking appropriately

account of the near unit root problem the ®ndings in most of the previous

literature are recon®rmed. There is support for the Fisher e¨ect in the interest

rate targeting period (1953:1±1979:10) of the Federal Reserve but not in the

1979:11±1990:12 period.

1. Introduction

in real interest rates or in expected in¯ation is one of the central problems

in monetary economics. Proponents of the commonly held Fisher e¨ect view

claim that changes in short-term interest rates primarily re¯ect changes in ex-

* I am indebted to Graham Elliott, Seppo Honkapohja, Johan Knif, Erkki Koskela, Pentti

Saikkonen, Matti VireÂn, Anders Warne and the three anonymous referees for useful comments.

The usual disclaimer applies. Financial support from the YrjoÈ Jahnsson Foundation and SaÈaÈstoÈ-

pankkien tutkimussaÈaÈtioÈ is gratefully acknowledged.

358 M. Lanne

view, the results seem to depend considerably on the time period and country

being studied (for previous results see Mishkin (1992) and references therein).

Recently Mishkin (1992, 1995) has suggested that an explanation of this

lack of robustness of the Fisher e¨ect might stem from the use of inappro-

priate econometric methods. Typically empirical studies have involved re-

gressing the in¯ation rate over m periods on the m-period nominal interest rate

and testing the signi®cance of the slope coe½cient. However, there is a wealth

of results from unit root tests suggesting that both interest rates and in¯ation

are integrated (of order one) variables, and therefore standard statistical infer-

ence may be invalid in these regression models. Using critical values obtained

by Monte Carlo simulation under the assumption that both variables (one

and three month in¯ation and interest rates, respectively) are I 1 and not

cointegrated Mishkin (1992) found no support for the Fisher e¨ect in U.S.

monthly data spanning the period 1953±1990. In his 1995 paper Mishkin

generated the critical values under the assumptions of I 1 and cointegrated,

and stationary in¯ation and interest rates, respectively, but in both cases the

result was the same as in the his 1992 paper, except for some evidence for

positive correlation between real and nominal interest rates after 1979.

The main motivation for making the I 1 assumption for in¯ation and

interest rates lies in the idea that it is a better approximation than station-

arity. Moreover, it is well known that unit root test have low power against

close alternatives (see e.g. Elliott et al. (1996)), so that the null hypothesis

of a unit root can seldom be rejected for such highly autocorrelated vari-

ables. Recent results in the econometric literature (Elliott (1994), and Elliott

and Stock (1994), inter alia) suggest, however, that if the regressor is poten-

tially I 1, but its order of integration is unknown, both standard inference

and the procedure of using a pretest for a unit root and then applying the

consequent distribution theory, can result in sizeable overrejection. As re-

cently pointed out by Stock (1997), standard parametric bootstrap is not

asymptotically valid in this case, either. These theoretical results thus cast

doubts on the use of the above mentioned inferential procedures based on

Monte Carlo simulation. Therefore, a reexamination of the Fisher hypothesis

that explicitly takes account of the potential near unit root problem, is re-

quired.

Applying an asymptotically valid inferential procedure to Mishkin's data,

we ®nd support for the Fisher e¨ect in the interest rate targeting period of the

Fed (1953:1±1979:10), whereas the e¨ect is absent in the post 1979 data. This

outcome is in con¯ict with Mishkin (1992, 1995), while it agrees with much of

the earlier literature.

The plan of the paper is as follows. In Section 2 we introduce the in¯a-

tion forecasting equation that has been used to test the Fisher e¨ect, and dis-

cuss the potential econometric problems caused by the high persistence in the

interest rates. Mishkin's (1992, 1995) simulation-based procedures are evalu-

ated from this point of view and an alternative inferential procedure is sug-

gested. Section 3 contains the econometric results using Mishkin's (1992,

1995) data1 along with some Monte Carlo simulation results to evaluate the

In¯ation and interest rates 359

small sample performance of the method. The empirical results are obtained

using extensions of the asymptotically valid method of Cavanagh, Elliott and

Stock (1995). The derivation of these extensions is deferred to the Appendix.

Section 4 concludes.

2. Econometric Methodology

models where the regressor is not exogenous and has a large, potentially unit

root. These characteristics describe well the following in¯ation forecasting

equation that has been used to test the Fisher e¨ect,

(when monthly observations of the price level P are used) and itm is the m-

month nominal interest rate. The test for the Fisher e¨ect involves testing the

signi®cance of bm ; if b m equals zero, then there is no Fisher e¨ect. Under

rational expectations a signi®cance test on bm in (1) can also be interpreted

as testing for the correlation between nominal interest rates and expected

in¯ation.

If itm is I 1, standard inference (based on the normal distribution) may

not be valid as was noted by Mishkin (1992), who instead based inference

on simulated critical values under the I 1 assumption. In general, however,

the order of integration of the regressor in (1) cannot be assumed known. Al-

though a unit root test does not reject (see Table 1 below), it is plausible

that the interest rate variables have a root near but not necessarily equal to

unity. It has recently been shown by Elliott (1994), and Elliott and Stock

(1994) that substantial size distortions can prevail in t tests when the explan-

atory variable in a regression model has a large, possibly unit root.

Table 1. 95% con®dence intervals of the largest autoregressive roots of interest rates

1953:1±1990:12

i1 8 1.984 0.971, 1.007

i3 8 1.804 0.975, 1.008

1953:1±1979:10

i1 6 0.714 0.991, 1.015

i3 6 0.710 0.990, 1.015

1979:11±1982:10

i1 0 1.956 0.638, 1.094

i3 1 2.472 0.471, 1.065

1982:11±1990:12

i1 0 2.262 0.832, 1.029

i3 0 1.312 0.929, 1.044

The lag length k in the ADF tests was determined by step-down testing (Ng and Perron (1995)),

using a 5% level for each lag. The con®dence intervals were obtained by linear interpolation based

on the ®gures in Stock (1991, Table A.1).

360 M. Lanne

Elliott and Stock (1994) studied the properties of the t test statistic using

local-to-unity asymptotics, i.e. they parametrized the largest autoregressive

root of the regressor, r, as 1 c=T, where c is a constant and T is the sample

size. They reached two main conclusions. First, the asymptotic null distribu-

tion of the t statistic is nonstandard and depends on two nuisance parameters,

the local-to-unity parameter c and the long-run correlation between the error

terms of the regression model (1) and the autoregressive representation of the

regressor, i.e. on the simultaneity in the system. Of these the latter is con-

sistently estimable whereas the former is not. The overrejection in standard

inference is the worse the higher is the simultaneity. Second, the often applied

procedure of pretesting for a unit root in the spread and subsequently employ-

ing the consequent distribution theory also leads to overrejection. This is so

because asymptotically the rejection probability of the null hypothesis r 1 is

less than one, meaning that the pretest is not consistent in the local-to-unity

case. Hence the nonstandard critical values corresponding to the I
1 case are

too often selected and overrejection prevails.

Mishkin (1992) considered inference in the forecasting regression (1) as-

suming exactly I
1 in¯ation and interest rates and no cointegration. He ob-

tained the critical values of the t statistic under these assumptions, and found

no evidence for the Fisher e¨ect. The simulated distributions are, of course,

crucially dependent on the restrictive assumptions. If the largest autoregres-

sive roots of the series deviate from unity or are cointegrated contrary to these

assumptions, the estimated time series models used as DGPs are misspeci®ed.

In a related paper Mishkin (1995) considered di¨erent DGPs estimated with

the same data set. First, while still making the exact unit root assumption, he

assumed that the in¯ation and interest rates are cointegrated. Second, he esti-

mated the DGP without imposing the unit roots. In both cases the results

obtained using the consequent simulated critical values are similar to those

in his 1992 paper, with the exception of some evidence for positive

correlation between the real and nominal interest rates after 1979.2 The ®rst

DGP is, however, misspeci®ed, if the variables are not unit root processes,

and, as recently pointed out by Stock (1997), also the latter simulation proce-

dure (parametric bootstrap without imposing the exact unit root) is asymp-

totically invalid when the order of integration of the regressor is not known

(for a proof of this in the closely related case of the cointegration regression,

see Stock and Watson (1996)). Hence these procedures cannot be considered

satisfactory for testing the Fisher e¨ect in the framework of regression (1).

The list of asymptotically valid tests in models where the order of integr-

ation of the regressor is unknown, is still short (for a survey, see Stock (1997)),

and little is known about the relative merits of the di¨erent tests.3 A recently

presented contribution that looks promising, is due to Cavanagh, Elliott and

Stock (1995), who suggested basing inference on simultaneous con®dence in-

2 In Mishkin (1995) the regression model being studied instead of (1) has the ex post real rate as

the dependent variable.

3 It is worth noting that the lag augementation scheme of Toda and Yamamoto (1995) provides a

technically uncomplicated way of achieving asymptotically valid inference. The idea is to add to

the model one additional lag that is not restricted under the null hypothesis. However, as pointed

out by Stock (1997) the local asymptotic power of this procedure is lower than that of the alter-

native tests. Finite sample power is also likely to be low, especially in our model which involves

only one lag (see the simulation results of Dolado and LuÈtkepohl (1996)).

In¯ation and interest rates 361

Sche¨e type con®dence interval consisting of all the values of bm0 for which

there is at least one value of the local-to-unity parameter c0 such that the joint

null hypothesis b m ; c bm0 ; c0 is not rejected. These conservative con®-

dence intervals can be shown to be asymptotically valid when the regressor

has a root local to unity. Cavanagh et al. derived a Wald test that can be

inverted to compute the simultaneous con®dence interval. The critical values

of this test depend on c0 so that the con®dence interval must be solved nu-

merically for its di¨erent values.5 Cavanagh et al. considered rather a stylized

model that corresponds to model (1) with m 1 and no serial correlation in

the error term. In the Appendix it is shown, how their test can be modi®ed

when the error term is serially correlated and m 3.6 In all cases the proce-

dure is rather conservative with the asymptotic rejection rates varying between

2 and 7% when the nominal size is 10%. According to the calculations of

Cavanagh et al. local asymptotic power in the m 1 case is good. Finite

sample simulation results are presented at the end of Section 3.

3. Empirical results

In this section we apply the Sche¨e type procedure of Cavanagh et al. (1995)

to test the Fisher e¨ect. The data set consists of monthly observations of U.S.

one and three month Treasury bill rates and in¯ation rates computed from the

consumer price index from the period 1953:1±1990:12 (456 observations). The

same data were also studied by Mishkin (1992, 1995). Following Mishkin,

three subsample periods are considered in addition to the entire sample. There

is some empirical evidence suggesting that there were structural changes in the

relationship between in¯ation and interest rates when the Federal Reserve

System ceased targeting interest rates in 1979 and again when the targeting

was partially resumed in 1982 (see e.g. Huizinga and Mishkin (1986)). To be

able to compare our empirical results with those of Mishkin we have chosen

to take his subsample division as given. Whether such structural changes

really occurred and which the exact dates may have been, is of course, an in-

teresting question in its own right but it will not be studied here. The three

subsample periods are 1953:1±1979:10 (322 observations), 1979:11±1982:10

(36 observations) and 1982:11±1990:12 (98 observations).

To assess the severity of the near unit root problem let us ®rst take a look

at the estimates of the largest autoregressive roots of the interest rate process

and the simultaneity in the system. The 95% con®dence intervals of the largest

intervals. For sup-bound and Bonferroni intervals they also considered size-adjustments, and the

size-adjusted tests seemed to have somewhat better power properties than the Sche¨e type proce-

dure. However, the Sche¨e type procedure allows for a rather straightforward extension to the

case where the error terms are autocorrelated or there is a longer lag of the explanatory variable

while it is hard to see how the other procedures could be extended in those directions. Therefore,

we have chosen to employ that procedure in this paper.

5 For the empirical results in Section 3 a grid consisting of 160 values of c0 ranging from 40:0

to 9.5 was considered. I am grateful to Graham Elliott for providing the critical values needed in

the computations.

6 The case of a general m > 1 is straightforward, but only the case m 3 is needed in the em-

pirical part of the paper.

362 M. Lanne

level the null of a unit root cannot be rejected for any of the interest rate

variables. The very wide intervals for the period of the `new operating proce-

dures' of the Federal Reserve System 1979:11±1982:10 re¯ect most likely the

small number of observations. The con®dence intervals include in all cases

values that are liable to cause size distortions in t tests on bm in (1) as dis-

cussed in Section 2. In particular, unity is included in all the intervals, and the

lower limits vary between 0.832 and 0.991 in the entire sample and the ®rst

and last subsample periods. In the 1979:11±1982:10 period the intervals are

considerably wider which is probably due to the small number of observations

in this subsample. The values of the local-to-unity parameter c corresponding

to the lower limits of these con®dence intervals range from about 3 to 19.

Whether overrejection really is a substantial problem depends on the magni-

tude of simultaneity between the residuals of the in¯ation forecasting equa-

tion (1) and the univariate autoregressive representation of the interest rate

variable in question. The estimates of the long-run correlation between the

residuals, are presented in Table 2. With the exception of the interest rate

targeting period 1953:1±1979:10, the estimates are high implying that size

distortions are in general to be expected if standard inference or pretesting is

used.

The 95% con®dence intervals for b m in Table 3 accord with our expecta-

tions based on the estimates of the long-run correlation. The null of b m 0 is

rejected in the entire sample period, and the absence of the Fisher e¨ect in the

Table 2. Estimates of the long run correlation between the residuals of the ADF regression for the

m month interest rate and the in¯ation forecasting equation (1)

m 1 3

Sample period

1953:1±1979:10 0.082 0.181

1979:11±1982:10 0.677 0.824

1982:11±1990:12 0.347 0.342

The estimates are obtained by autoregressive approximation. AIC with a maximum of 10 lags was

used to select the lag length.

Table 3. 95% Sche¨e con®dence intervals of the slope coe½ents b m in the in¯ation forecasting

equation (1)

m 1 3

1953:1±1979:10 0.993, 1.496 1 1.136, 1.514 3

1979:11±1982:10 2.135, 3.441 1 0.817, 0.591 2

1982:11±1990:12 0.893, 0.753 1 1.044, 0.914 1

The con®dence intervals were solved on a grid with the values of c ranging from 40.0 to 9.5. The

lag length k was selected by AIC.

In¯ation and interest rates 363

(1992, 1995) but in accordance with several previous results relying on stan-

dard inference there is support for the Fisher e¨ect in the interest rate target-

ing period. As implied by the low estimated simultaneity for this period, in-

ference using critical values from the standard normal distribution or applying

pretests is not likely to lead to size distortions, i.e. they are asymptotically

valid procedures.

The conclusion from the entire sample period is that nominal interest rates

have information about future in¯ation. Closer inspection, however, reveals

that changes in monetary policy seem to have had an e¨ect on this relation-

ship. Recently SoÈderlind (1999) has argued that stronger in¯ation targeting

by the central bank leads to large ¯uctuations of the real interest rate and, if

successful, decreases the Fisher e¨ect by stabilizing in¯ation expectations such

that nominal interest rates primarily re¯ect the real rate. The period of the

new operating procedures was indeed characterized by aggressive disin¯ation

policy while the most recent subsample can be described as a period of estab-

lishing and maintaining central bank credibility with emphasis on the in¯ation

target (see e.g. Goodfriend (1995)). Also, recently Fuhrer (1996) concluded

that in the post 1979 period the relative weight of the in¯ation target in the

Fed's reaction function was higher and the in¯ation target itself lower than

in the 1966±1979 period. Combined with these observations, our results thus

lend support to the conjecture of SoÈderlind.

The results of Table 3 also give information concerning the (near) station-

arity of the real interest rate. Note that under rational expectations the sta-

tionarity of p m i m implies the stationarity of the ex ante real rate. In terms

of model (1) this hypothesis can be expressed as b m 1, and its interpretation

is, of course, that in¯ation and interest rates move one-for-one in the long run.

Mishkin (1992) called this the full Fisher e¨ect. In the entire sample unity is

included in the 95% Sche¨e type con®dence interval in the m 1 case but not

in the m 3 case. In the interest rate targeting period the situation is similar

with the lower bounds of the interval in the m 3 case just above unity. In

the 1982:11±1990:12 period the hypothesis that nominal interest rates and

in¯ation move one-for-one in the long run, is clearly rejected at the 5% level.

The outcome that the hypothesis of common serial correlation properties

between nominal interest rates and in¯ation is clearly rejected only in the

period where the in¯ation target was emphasized, is of course, also consistent

with the results of SoÈderlind (1999).

To evaluate how well our procedures control size in ®nite samples some

Monte Carlo experiments were conducted with models estimated from the

data.7 The rejection rates for the at most 5% nominal level test are presented

in Table 4. In these cases the procedures are not as conservative as the

asymptotic results may lead one to believe. They control size relatively well in

all samples except the one corresponding to the period of the new operating

procedures. In that sample large overrejections prevail presumably due to the

small sample size. Also in the last subsample period especially the test in the

m 3 case tends to overreject.

7 Speci®cally, the in¯ation forecasting equation (1) and an autoregression for the interest rate

with a VAR model for the residuals were estimated for each of the sample periods. The error

^ distribution, where S^ is the estimated covariance matrix of the

terms were generated from N
0; S

®nal residuals.

364 M. Lanne

Table 4. Rejection rates of tests of the Sche¨e type test with nominal size U5%

m1 m3

1953:1±1979:10 322 0.055 0.054

1979:11±1982:10 36 0.381 0.882

1982:11±1990:12 98 0.064 0.125

The entries are based on 2,000 Monte Carlo replications of models for interest rates and in¯ation

(for details see footnote 7) estimated with data on the m-period in¯ation and interest rates from

the indicated sample period. T is the sample size.

4. Conclusion

In this paper we test the Fisher e¨ect according to which changes in short-

term interest rates primarily re¯ect changes in expected in¯ation. It has been

suggested that both in¯ation and nominal interest rates have unit roots, and

therefore the results obtained in much of the earlier empirical work relying

on standard inference are questionable. Since there is no theoretical reason

to assume these variables to be I
1, we apply the Sche¨e type approach of

Cavanagh et al. (1995) which is asymptotically valid under uncertainty about

the order of integration. The data set consists of U.S. short-term in¯ation and

interest rates covering the period from 1953 to 1990. The general conclusion is

that the Fisher e¨ect is absent in the post 1979 subsample periods but there is

support for it in the interest rate targeting period 1953:1±1979:10. A potential

explanation for the presence of the Fisher e¨ect in the former and absence in

the latter period is o¨ered by a change in the emphasis on in¯ation targeting

(see SoÈderlind (1999)).

There is some support for the one-for-one movement of in¯ation and

nominal interest rates in the entire sample and potentially in the interest rate

targeting period. In contrast, this hypothesis is clearly rejected in the post 1982

period. Also, the recent results of Evans and Lewis (1995) and Crowder and

Ho¨man (1996) lend support to the (full) Fisher e¨ect in the U.S. Using data

from 1947 to 1987 Evans and Lewis argue that taking account of changes

in the process of in¯ation it cannot be rejected that in the long run nominal

interest rates re¯ect expected in¯ation one-for-one. Crowder and Ho¨man, on

the other hand, reconcile data from 1952 to 1991 with the Fisher hypothesis

by explicitly adjusting the interest rate for time varying taxes. Neither paper

has subsample results.

Econometric appendix

Cavanagh, Elliott and Stock (1995) needed in the empirical part of the paper.

Throughout, generic notation is used, i.e. the regressor is denoted by x, the

regressand by y etc.

In¯ation and interest rates 365

xt rxt 1 v1t

where r 1 c=T, vt
v1t ; v2t 0 F
L 1 et , F
L is a kth order lag poly-

nomial matrix having all roots outside the unit circle, with typical element

Fij
L (Fij 0 I2 ) and et
e1t ; e2t 0 is a martingale di¨erence sequence with

constant conditional covariance matrix S, and supt Eeit4 < y, i 1; 2 and

2

Ex10 < y. Hence serial correlation in the error terms is assumed to be well

described by an autoregressive process. As far as the empirical application

of the paper is concerned, the latter equation corresponds to the in¯ation

forecasting equation (1), and the ®rst equation can be seen as the generating

mechanism of the interest rate.

Model (2) can be written in the following error-correction form, where the

error-correction term ut g0 xt yt is known under the null hypothesis

g; c
g0 ; c0 :

X

k X

k 1

Dxt c10 c11 xt 1 c12 ut 1 a1i Dxt i b1i Dyt i e1t

i1 i1

X

k X

k 1

Dyt c20 c21 xt 1 c22 ut 1 a2i Dxt i b2i Dyt i e2t :
3

i1 i1

The null hypothesis implies two restrictions on the coe½cients of model (3):

Tc11 c0 F11
1 and Tc21 c0 F21
1. The coe½cients of the lag polyno-

mials are, of course, unknown, but in the test statistics F11
1 and F21
1 can

be replaced by their consistent estimators under the null hypothesis, F ^
1

Pk Pk 1 ^ ^ Pk P11k 1 ^

1 a^ g b g c and ^

F
1 g ^

a g

i1 1i 0 i1 1i 0 12 21 0 i1 2i 0 i1 b2i

^

g0 c22 . We obtain the statistic

!

1 0 ^ 1 2X T

m2

W 1 rT S T xt 1 rT ;
4

2 t2

P

where rT
T c^11 c0 F ^11
1; T c^21 c0 F

^21
1 0 , S^ 1=
T k T ^t e^t0 ,

tk1 e

0 m

and e^t
^ e1t ; e^2t are the OLS residuals from (3), xt 1 xt 1 1=

PT

T 1 t2 xt 1 . It is easy to see that under the null hypothesis
g; c

g0 ; c0 , W1 converges to the same limiting distribution as Cavanagh et al.

obtained in the simpler case with no error autocorrelation since the only dif-

ference between the models is in the dynamics and only the coe½cients of the

lagged level are involved in the test.

If xt 3 is the regressor in (2) instead of xt 1 the error-correction form of

the model corresponding to (3) has lags of Dxt up to k 2 while the longest

lag of Dyt is still k 1. In this case the joint null hypothesis g g0 and c

c0 implies additional restrictions on the coe½cients of the subsequent error-

correction model. The test statistic, W3 , can be written as the sum of W1 and

366 M. Lanne

lagged di¨erences. It can be shown that the asymptotic null distribution of

W3 is the same as that of W1 except for an additional w42 =2 part stemming

from the additional restrictions.

References

Cavanagh CL, Elliott G, Stock JH (1995) Inference in models with nearly integrated regressors.

Econometric Theory 11:1131±1147

Crowder WJ, Ho¨man DL (1996) The long-run relationship between nominal interest rates and

in¯ation: The Fisher equation revisited. Journal of Money, Credit, and Banking 28:102±118

Dolado JJ, LuÈtkepohl H (1996) Making Wald tests work for cointegrated VAR systems. Econo-

metric Reviews 15:369±386.

Elliott G (1994) Application of local to unity asymptotic theory to time series regression. Un-

published Ph.D. dissertation (Harvard University)

Elliott G, Rothenberg TJ, Stock JH (1996) E½cient tests for an autoregressive unit root. Econo-

metrica 64:813±836

Elliott G, Stock JH (1994) Inference in time series regression when the order of integration of a

regressor is unknown. Econometric Theory 10:672±700

Evans MDD, Lewis KK (1995) Do expected shifts in in¯ation a¨ect estimates of the long-run

Fisher relation? Journal of Finance 50:225±253

Fuhrer JC (1996) Monetary policy shifts and long-term interest rates. Quarterly Journal of

Economics 111:1183±1209

Goodfriend M (1995) Acquiring and maintaining credibility for low in¯ation: The US experience.

In: Leiderman L, Svensson LEO (eds.), In¯ation Targets. CEPR, London, pp. 122±141

Huizinga J, Mishkin FS (1986) Monetary policy regime shifts and the unusual behavior of real

interest rates. Carnegie-Rochester Conference Series on Public Policy 24:231±274

Mishkin FS (1992) Is the Fisher e¨ect for real? A reexamination of the relationship between

in¯ation and interest rates. Journal of Monetary Economics 30:195±215

Mishkin FS (1995) Nonstationarity of regressors and tests on real-interest-rate behavior. Journal

of Business & Economic Statistics 13:47±51.

Ng S, Perron P (1995) Unit root tests in ARMA models with data-dependent methods for the

selection of the truncation lag. Journal of the American Statistical Association 90:268±281

Stock JH (1991), Con®dence intervals for the largest autoregressive root in US macroeconomic

time series. Journal of Monetary Economics 28:435±459

Stock JH (1997) Cointegration, long-run comovements, and long-horizon forecasting. In: Kreps

DM, Wallis KF (eds.) Advances in economics and econometrics: theory and applications,

Seventh World Congress, volume III. Cambridge University Press, Cambridge, U.K, pp. 34±

60

Stock JH, Watson MW (1996) Con®dence sets in regressions with highly serially correlated

regressors. Manuscript (Harvard University)

SoÈderlind P (1999) Monetary policy and the Fisher e¨ect. Journal of Policy Modeling (forth-

coming)

Toda HY, Yamamoto T (1995) Statistical inference in vector autoregressions with possibly

integrated processes. Journal of Econometrics 66:225±250

- New York Metro Data AnalysisUploaded bykevin11h
- Basic Statistics in Bio MedicalUploaded byifai.gs7487
- 1.__DMAIC_-List_of_Tools.pptUploaded byjitendrasutar1975
- StatisticsUploaded byMujeeb Abdullah
- Effect of Promotion Strategies on the Organizational Commitment of Banking Sector Employees in KenyaUploaded byIOSRjournal
- Predstavitev Rezultatov V5-0451 (a)Uploaded byTilen Mandelj
- Determinants of Financial DevelopmentUploaded byDaniil Ovchinnikov
- FRM2009SyllabusUploaded bygreathill56
- School Principals Leadership Style.pdfUploaded byRalph Fael Lucas
- UT Dallas Syllabus for gisc6379.501.08s taught by Michael Tiefelsdorf (mrt052000)Uploaded byUT Dallas Provost's Technology Group
- Kine StudyUploaded byErica L. Robinson
- The First Fifty Years of Modern EconometricsUploaded byAsel Ch
- BSSTUploaded byPallav Kumar
- Financial Econometrics Outline CopyUploaded byJunaid Cheema
- Hypothesis TestingUploaded byMalik Babar
- Ba Yes Freq BookUploaded byseamusol
- ECN225sol4.pdfUploaded byJamie1231
- Final Project Memo_QNT550Uploaded bySrinath Meduri
- statment of purposeUploaded byapi-356029006
- perron1989.pdfUploaded bySergio
- ContentsUploaded byRanndolf Javier
- Correlations and Inferential Statistics_Workshop1Uploaded bypemea2008
- Syllabus - Statistics2 FallUploaded byNhím Biển
- 17616752 Advertising Sales ManagementUploaded bykhushi_sharma58101959
- Article Keynes and WgnerUploaded bym_nafees
- ch02Uploaded byarunprasadvr
- BA7_Multiple Regression 7.05Uploaded byandreea143
- A Study on Brand Impact of Apparels on Consumer Buying Behaviour in Kukatpally AreaUploaded byEditor IJTSRD
- organicvsnormalstrawberriesUploaded byapi-358599175
- Session 07 - Data Processing and UnivariateUploaded byyul_nguyen

- زبر زیر پیش عربیUploaded bycheema
- naturallyy coolingg ur homee.pdfUploaded byNur Amira Mardiana Zulkifli
- DW-391HZ.pdfUploaded bycheema
- VocabularyUploaded byAsm Burraq
- PG Cert HRH Expression of InterestUploaded bycheema
- ہائیبرنیشن آن آفUploaded bycheema
- New Text DocumentUploaded bycheema
- New Text Document (2)Uploaded bycheema
- New Text DocumentUploaded bycheema
- Monotype® Keyboard (View Only)Uploaded bycheema

- ch12qaUploaded byHana Lee
- Quiz _ iCPAUploaded byGizelle Taguas
- LN08_Eiteman_85652_14_LN08Uploaded bynotnull991
- FIN1 Midterm Exam1Uploaded byBebelan A. Madera
- EC08Uploaded byMuhammad Awais
- Review Ch_02 AnnuityUploaded byNelson Cabingas
- Hedging With OptionsUploaded byMichael Yuk
- Corporate Finance - Session 1Uploaded bySatish Kun Dalai
- Swap Chapter 1Uploaded bysudhakarhere
- Lecture 3Uploaded byBalu Bhs
- 04 Milestone CCILUploaded bygopalusha
- 20090603FRM_Exam2009AIMSUploaded byjatipatel5719
- Credit One 2Uploaded byBryan
- Example ProblemsUploaded byKhaled Abdelbaki
- econ12Uploaded bysanju das
- FARM-Africa WEDP End of Project Evaluation 2008Uploaded byFARMAfrica
- 5. FINANCIAL CONCEPTS.pdfUploaded byVictor Miguel Vergara Lovera
- A Century of News: 100 Years of the News-BulletinUploaded byVCNews-Bulletin
- Solution Manual Introduction to Corporate Finance 5th Edition by Alex Frino SLP1163Uploaded byThar Adelei
- Financial Econometric ModellingUploaded byMarta Lopes
- ValuationOfCliquetOptionsUploaded byLawrence Chau
- Financial Management: Valuation of Bonds and Shares QuizUploaded bySeri Diyana
- Chapter-7Uploaded bySharmin Rumi
- Interest rate Futures in India- final oneUploaded byVishal Padole
- Auto Loan CalculatorUploaded byPramod Athiyarathu
- Googlespreadsheet Function ListUploaded byAneek Kumar
- Booth Cleary 2nd Edition Chapter 6 - Bond Valuation and Interest RatesUploaded byQurat.ul.ain Mumtaz
- b Do Troubled Debt Restructuring 2013Uploaded bymssidhu88
- Time value of moneyUploaded byHersh Oberoi
- 2009 Hsc Exam General MathematicsUploaded byHussein Jaber