Is Quantity Theory Still Alive? - Universidade Nova de Lisboadocentes.fe.unl.pt/~azevedoj/Web...
Transcript of Is Quantity Theory Still Alive? - Universidade Nova de Lisboadocentes.fe.unl.pt/~azevedoj/Web...
Is Quantity Theory Still Alive?∗
Pedro Teles
Banco de Portugal, Universidade Catolica Portuguesa, and CEPR
Harald Uhlig
Dept. of Econ., Univ. of Chicago, NBER and CEPR
Joao Valle e Azevedo
Banco de Portugal and Nova School of Business and Economics
This revision: July 9, 2015
Abstract
This paper investigates whether the quantity theory of money is still alive. We demon-
strate three insights. First, for countries with low inflation, the raw relationship between
average inflation and the growth rate of money is tenuous at best. Second, the fit markedly
improves, when correcting for variation in output growth and the opportunity cost of money,
using elasticities implied by theories of Baumol-Tobin and Miller-Orr. Finally, a subsample
characterized by the adoption of inflation targeting shows considerably less inflation variability,
worsening the fit of a one-for-one relationship between money growth and inflation.
Keywords: quantity theory, money demand, money demand elasticity, inflation targeting
JEL codes: E31, E41, E42, E50
∗The addresses of the authors are Pedro Teles and Joao Valle e Azevedo, Banco de Portugal, DEE, 1150-165,Lisbon, e-mails: [email protected], [email protected], and Harald Uhlig, Univ. of Chicago, Dept. of Economics,1126 East 59th Street, Chicago, IL 60637, U.S.A, [email protected]. Uhlig’s research has been supported by theNSF grant SES-0922550 and by a Wim Duisenberg fellowship at the ECB. Teles gratefully acknowledges the financialsupport of FCT. Uhlig has an ongoing consulting relationship with a Federal Reserve Bank, the Bundesbank andthe ECB. The views here are entirely our own.
1
1 Introduction
One of the most established folk wisdoms in monetary economics is a relationship, which in its
practical version for monetary policy might be stated as follows: long run inflation is related one-
for-one with long-run money growth. This “quantity theory” relationship seems firmly established
at least since Friedman (1956) and Lucas (1980).
This paper takes a cross-section of countries from 1970 to 2005 and re-investigates the relation-
ship between money growth and inflation. We demonstrate three insights. First, for countries with
low inflation, the raw relationship between average inflation and the growth rate of money is tenuous
at best. Second, the fit markedly improves, when correcting for variation in output growth and the
opportunity cost of money, using elasticities implied by money demand theories of Baumol-Tobin
and Miller-Orr. Finally, the sample after the implicit or explicit adoption of inflation targeting (IT)
shows considerably less inflation variability, worsening the fit of a one-for-one relationship between
money growth and inflation but maintaining a low residual variance.
To demonstrate these insights, we provide a series of graphs and tables. We start by showing
that, for countries with moderate inflation, the raw relationship between money growth and inflation
is tenuous or even nonexistent. Quantity theory suggests to take into account the growth rate of
real GDP. Additionally, monetary theory points out the dependence of velocity on yields. The
correction for GDP growth alone turns out not to help. However, the correction for a yield effect has
a remarkable impact. Indeed, one would expect a rise in nominal yields to increase the opportunity
cost of holding money, and thus lead to a reduction in the real quantity of money per real unit
of output. This should translate to either lower nominal money growth or higher inflation. Lucas
(2000) has documented a rather tight fit of the ratio of the real quantity of money to real output
vis-a-vis the yield on government bonds, which furthermore is close to a relationship predicted
by theories on the transaction demand for money, see Baumol (1952), Tobin (1956), Miller and
Orr (1966). Taking into account the relationship suggested by Lucas, we demonstrate that the
fit between money growth and inflation indeed markedly improves. An even better fit is obtained
when using the (lower) elasticity value suggested by Miller and Orr (1966). We finally estimate the
relationship and find just a small improvement over the Miller-Orr specification.
2
The estimation of money demand has been under debate recently. The 90s and 2000s have been
testing decades for it, see in particular the discussion in e.g. Ball (2001), Carlson et al. (2000),
Coenen and Vega (2001), and Teles and Zhou (2005), Ireland (2009), Sargent and Surico (2011),
Lucas and Nicolini (2013). Motivated mostly by Ireland (2009) and Sargent and Surico (2011), we
split the data into two parts, choosing as split point for each country the dates coinciding with the
implicit or explicit adoption of inflation targeting.1 We also consider 1990 as an alternative breaking
point, which could reflect changes in financial arrangements and regulation. Teles and Zhou (2005)
and Lucas and Nicolini (2013), consider 1980 as a data break for the US, focusing on the effects
on monetary aggregates of banking deregulation introduced after 1980.2 Those changes together
with the financial innovation in the 1990s associated with the development of electronic payments
suggest that M1 might not be the most appropriate monetary aggregate to use in the later part of
the sample.
The relationship between money growth and inflation has become hard to pin down during the
second part of the sample. Generalized inflation targeting at low inflation rates, by considerably
reducing the dispersion of inflation across countries, has made it virtually impossible to establish
the one-to-one relationship between average inflation and the growth rate of money implied by the
Quantity Theory. This has also been argued by Sargent and Surico (2011) using US time series
data, in a recount of Whiteman’s (1984) arguments.
Another feature of the second part of the sample is that the interest elasticity of the money
demand is also harder to estimate and is lower than in the earlier sample. One of the reasons
for the difficulty in the estimation is the lower variability in interest rates. Reasons for the lower
elasticity are beyond the scope of this paper. They could involve the shape of the money demand
as in Ireland (2009) or the choice of the monetary aggregate as in Teles and Zhou (2005) or Lucas
and Nicolini (2013). As it turns out, the issues with the estimation of the interest elasticity in
the second part of the sample are not relevant for the purposes of our exercise. Given the lack of
variation in inflation and interest rates, the interest elasticity of the money demand does not make
a significant difference.
1We thank a referee for this suggestion.2The Depository Institutions Deregulation and Monetary Control Act of 1980 and the Garn-St Germain Deposi-
tory Institutions Act of 1982.
3
Investigating international cross-sections of countries to analyze the evidence on the quantity
theory of money has obviously been done before, notably by McCandless and Weber (1995), restated
in Lucas (1996). We build up on that literature. More recent literature such as Assenmacher-
Wesche and Gerlach (2006), using information on the interest rate, and Benati (2009) find a long
run relationship between money growth and inflation for several countries, but do not exploit the
cross section evidence as we do.
The outline of the paper is as follows. Section 2 provides a basic perspective on the cross-
country data. Section 3 describes a simple model that allows for additional “corrective” terms.
Section 4 examines the issue of subsample instability. The data is described in appendix A. An
online appendix and a .zip file provide further graphs and tables, as well as the data used and the
programs for calculating all results.
We conclude that quantity theory is still alive. Whether it should be used as a guide to long-term
monetary policy is more debatable, and it is certainly beyond the scope of this paper. Woodford
(2008) has argued that there is no independent role for tracking the growth rate of money, if
a central bank is already willing and able to stabilize inflation rates at short and medium-term
horizons, without making an explicit use of monetary aggregates. The practice of central banks
seems to be reassuring, that it is possible to keep inflation low using, as it appears to be the case,
some form of interest rate feedback rule. However, theory is more sceptical about that capacity,
pointing out that local determinacy does not imply uniqueness (see e. g. Benhabib, Schmitt-Grohe
and Uribe, 2001). In particular an interest rate feedback rule could be ineffective in raising inflation
from very low inflation at the zero bound. The tracking of money supply could be a means of
avoiding that policy trap (see also Atkeson, Chari and Kehoe, 2010, as well as the analysis in
Fischer et al., 2006).
2 The World
Teachers of intermediate macroeconomics may have consulted Barro (1993 or 2007) in order to
teach students the relationship between money growth rates and inflation. His figure 7.1 in the
1993 edition shows a large sample of countries, and plots this relationship, having calculated the
4
growth rates of money and prices from, typically, the fifties to 1990. The figure is reproduced here
as figure 1: one apparently gets a nice fit to the 45 degree line.
Figure 1
0 20 40 60 800
10
20
30
40
50
60
70
80
Money growth in Percent
Infla
tion
in P
erce
nt p
er y
ear
Money versus Inflation per year
This figure, which just restates figures drawn in Barro (1993, 2007), McCandless and Weber (1995) or Lucas (1996)
shows the relationship between average monetary growth rates and average inflation in a sample of 79 countries.
The data is from Barro (1993). Also drawn is the 45 degree line: it seems that, indeed, long-term monetary growth
is synonymous with long-term inflation.
However, that picture turns out to be misleading and mainly driven by high inflation countries.
Concentrating on the subset of countries we analyze, whose inflation rates were all below 12 percent,
the points no longer assemble nicely around a straight line, but produce a rather randomly looking
scatter plot with all the points below the 45 degree line, see figure 2. The question is thus: is the
relationship between money growth and inflation too loose to be of any relevance for low inflation
countries?
These pictures should be considered disturbing by anybody who believes in a tight relationship
between money growth and inflation and bases monetary policy advice on such a belief. Additional
issues may be of relevance at low rates of inflation, however. In particular, GDP growth, changes
5
Figure 2
AU
AT
CADK
FIN
FR
DE
IE
IT
JP
KR
NL
NZ
PT
ES
CHUK
US
02
46
810
Infla
tion
0 5 10 15 20Money Growth
Not Corrected1970*−2005
This figure is similar to figure 1, but restricts attention to only those countries in our sample whose average
inflation rate was below 12 percent in the period 1970-2005. The inflation measure is Consumer Price Index (CPI)
inflation and the monetary aggregate is M1. Average CPI inflation is plotted vis-a-vis average M1 growth.
Averages of the variables are from 1970 (or later if data is unavailable) until 2005. Instead of a tight relationship
between monetary growth and inflation, one can just see a cloud below the 45 degree line.
in interest rates, technological progress in transaction technologies as well as production may make
a difference. We next discuss how these can be taken into account.
3 Correcting the Quantity Theory relationship
In deriving an equilibrium money demand relationship, we follow the simple analysis in Attanasio,
Guiso and Jappelli (2002).3 A (representative) agent needs transaction services proportional to real
consumption ct, which are produced with time st and real money balances mt = Mt/Pt, according
to
st = l(A−1t ct,mt), (1)
where At measures progress in the transactions technology.
3See also Lucas (2000).
6
Given ct, the agent will want to minimize the cost of transactions, wtst + Rtmt, subject to (1),
where wt is the wage rate and Rt is the opportunity cost of holding money, the nominal interest
rate. The marginal condition
−wtlm(A−1t ct,mt) = Rt (2)
will have to hold.
Let the transactions function l be
l(c,m) = ηcamb (3)
for some η, a and b, with b < 0 and η > 0. When a = 1 and b = −1, the form for the transactions
technology can be justified by assuming, inspired by Baumol (1952) and Tobin (1956), that the
consumer spends cash holdings intended for the purchase of the good at a constant rate ct per unit
of time. ctmt
is the number of times cash balances for transactions of the good are exhausted and
must be restored, the number of trips to the bank. This time cost is a constant η. The Miller-Orr
(1966) specification amounts to l(c,m) = η(
cm
)2, i.e. a = 2 and b = −2.
The marginal condition (2) can be written as −ηbwtA−at catm
b−1t = Rt. In logs, the condition is
logmt = B − 1
1− blogRt +
a
1− blog ct +
1
1− blogwt −
a
1− blogAt, (4)
with B = log(−ηb)1−b
.
To make contact with the data, we wish to examine a panel of countries i = 1, . . . , N and a time
period t = 1, . . . , T . For a country i and a variable xi,t, generally denote the sample growth rate of
that variable between time one and time T with ∆xi =
log xi,T−log xi,1
T.
Equation (4) implies
∆Mi −∆P
i = − 1
1− b∆R
i +a
1− b∆c
i +1
1− b∆w
i − a
1− b∆A
i , (5)
in which ∆Mi is the average growth rate of the stock of nominal money in country i and ∆P
i is the
average inflation in that country. This suggests that the long run relationship between money and
7
prices ought to be corrected by the variation of nominal interest rates and by the growth rates of
consumption, as a measure of transactions, real wages and technical progress in transactions. There
are implications from the transactions technology for the relevant elasticities. For instance, for the
Baumol-Tobin, with a = 1 and b = −1, the interest elasticity would be one half, and so would be
the elasticity to the growth rate of consumption. For the Miller-Orr technology (a = 2 and b = −2),
the two elasticities are one third and two thirds, respectively.
Before taking the theoretical relationship to the data, we take two further steps. The first is to
use balanced growth to impose that the average growth rate of consumption is equal to the growth
rate of real wages, and to the growth rate of output, ∆yi ,
∆ci = ∆w
i = ∆yi . (6)
Notice that with this assumption, the relationship between money and prices is to be corrected by
the growth rate of output with an elasticity of 1+a1−b
. This long run elasticity is equal to one for both
Baumol-Tobin and Miller-Orr technologies.4
The second step is to assume that the cross-country level shift in the transactions technology
can be captured by a random fixed effect,
a
1− b∆A
i = ϵi, (7)
where we assume, and this is a strong assumption, that ϵi is independent of ∆yi and ∆R
i . With this
assumption as well as with (6), we obtain the empirical specification5
∆Mi −∆P
i = γ∆yi − α∆R
i − ϵi, (8)
where
γ =1 + a
1− band α =
1
1− b. (9)
4The unit elasticity of the money demand to output in typical formulations of the money demand, as in Lucas(2000), is a feature of long run growth.
5The specification in (8) is log-log in contrast to a semi-log specification. See the discussion in Bailey (1956) andIreland (2009).
8
One can now either proceed to estimate (8), noting that the two structural parameters a and b
are identified per (9), or one can directly measure the fit of that equation for given specifications of
the transaction technology. In particular, we note that
∆Mi −∆P
i = ∆yi −
1
2∆R
i − ϵi (10)
for the Baumol-Tobin specification and
∆Mi −∆P
i = ∆yi −
1
3∆R
i − ϵi (11)
for the Miller-Orr specification.
3.1 Data and Results
For our investigation, we have used data for all OECD countries, drawing on statistics of the IMF,
the OECD, the European Commission, the ECB and other sources. We excluded countries with
average inflation above 12 percent, transition countries and countries with missing data. We focus
on annual data from 1970 until 2005. The reason not to include data after 2005 is to avoid the zero
bound episode in the aftermath of the financial crisis. At zero interest rates, money and bonds are
perfect substitutes and the demand for money is not uniquely pinned down. Put differently, if we
are to find changes in the relationship between money growth and inflation within our sample, they
will not be due to zero bound considerations. We used CPI inflation, short rates as well as M1 for
all countries.6 We agree that more appropriate aggregates, with instruments more closely related to
transactions motives, such as Money Zero Maturity (MZM) could be used. Unfortunately, they are
not readily available and are reliable only in a few countries. The same applies to divisia indices,
see, e.g., Barnett (1988) and Barnett and Apostolos (2000). More information on the data as well
as explanations for the short codes used to denote countries are in appendix A.
Figure 3 “corrects” the money growth rate by subtracting the GDP growth rate. The points
scatter loosely below the 45-degree line. Figure 4 removes the yield effect with the coefficient of
6In earlier versions of this paper we also experimented with M2 and M3, as well as long rates: the data problemsthere were generally greater, but preliminary results looked rather similar to the results documented here.
9
0.5 on the interest rate change as suggested by the Baumol-Tobin specification (10), as well as
suggested by Lucas (2000) and close to the one estimated for Italy by Alvarez and Lippi (2009).
The correction with the yield considerably improves the fit, shifting the data points upwards, that
now line up nicely along the 45-degree line.
Figure 5 contains the result for the Miller-Orr specification, while figure 6 finally contains the
correction implied by estimating (8) with γ = 1 and including a constant, using as variables the
averages from 1970 until 2005, per ordinary least squares.7 Information about the quality of fit, by
calculating an R2 and the variances of ϵi is in table 1, including results for subsamples, see section 4.
The results from the regression imposing γ = 1 are in table 2, including results for subsamples, see
also section 4. Full results with γ unrestricted can be found in the online appendix. The estimated
interest rate elasticity is below the Baumol-Tobin and Miller-Orr values. For the full sample, the
Baumol-Tobin fits worse compared to the other three specifications that control for yields, all of
which provide essentially the same quality of fit.
We complement this analysis with the estimation of the following panel cointegration model:
log(Mit
Pit
) = α0,i + α1t+ γ log(Yit)− α log(Rit) + uit
with ∆
(log(Yit)
log(Rit)
)= λ0,i + vit
where ∆ is the first-difference operator. wit := (uit, vit)′ is possibly serially correlated but assumed
independent across i = 1, ..., N . (1,−γ, α) is a cointegrating vector and the equilibrium error
(log(Mit
Pit) − γ log(Yit) − α log(Rit)) is allowed, in its most general configuration, to have country-
specific fixed effects and a common time trend. Interestingly, panel tests indicate clearly that
log(Mit
Pit), log(Yit) and log(Rit) are integrated of order one and cointegrated.8 Next, we consider
several variations of the Dynamic OLS (DOLS) and Fully Modified OLS (FMOLS) panel estimators,
see e.g., Phillips and Moon (1999), Pedroni (2000, 2001), Kao and Chiang (2000) and Mark and Sul
7We provide results including a constant, which would not be motivated by the theory above but instead by thedesirability to control (if one thinks in the equation in levels) for movements in velocity not captured by changes inthe interest rates, due, e.g., to financial innovation.
8We employ the tests implemented in E-views due to Levin, Lin and Chu (2002), Breitung (2000), Im, Pesaranand Shin (2003), Maddala and Wu (1999), Choi (2001) and Hadri (2000) to test for unit roots and those due toPedroni (1999), Pedroni (2004), Kao (1999) and Maddala and Wu (1999) for cointegration.
10
(2003). Using DOLS the uit are allowed to be correlated with at most pi leads and lags of vit, which
controls for the possible endogeneity one hardly can get rid of in the simple regression setting. For
all the details and full results of this exercise, see the online appendix. Partial results obtained
with the restriction γ = 1 are in table 3. In this setting, the estimates of α obtained within the
1970-2005 sample are in the range 0.10-0.29, i.e., below the 0.33 of the Miller-Orr specification.
11
Figure 3
AU
AT
CADK
FINFR
DE
IE
IT
JP
KR
NL
NZ
PTES
CH UK
US
05
1015
Infla
tion
0 5 10 15Corrected Money Growth
Corrected for GDP growth1970*−2005
Corrected money growth rate here is average M1 growth minus average real GDP growth. Average CPI inflation is
plotted vis-a-vis corrected money growth. For each country, the average of the variables is from 1970 (or later if
data is unavailable) until 2005. The points scatter loosely around, but mostly below the 45-degree line.
12
Figure 4
AU
AT
CADK
FINFR
DE
IE
IT
JP
KR
NL
NZ
PTES
CH UK
US
05
1015
Infla
tion
0 5 10 15Corrected Money Growth
Corrected for yield − Baumol−Tobin1970*−2005
Baumol-Tobin: Corrected money growth rate here is average M1 growth minus average real GDP growth plus the
average growth of a nominal interest rate, divided by two, following (10). Average CPI inflation is plotted vis-a-vis
corrected money growth. For each country, the average of the variables is from 1970 (or later if data is unavailable)
until 2005. The correction with the yield improves the fit to the 45-degree line.
13
Figure 5
AU
AT
CADK
FINFR
DE
IE
IT
JP
KR
NL
NZ
PTES
CH UK
US
05
1015
Infla
tion
0 5 10 15Corrected Money Growth
Corrected for yield − Miller−Orr1970*−2005
Miller-Orr: Corrected money growth rate here is average M1 growth minus average real GDP growth plus the
average growth of a nominal interest rate, divided by three, capturing the transactions technology model due to
Miller and Orr (1966). Average CPI inflation is plotted vis-a-vis corrected money growth. For each country, the
average of the variables is from 1970 (or later if data is unavailable) until 2005. The fit around the 45 degree line is
better than the Baumol-Tobin specification.
14
Figure 6
AU
AT
CADK
FINFR
DE
IE
IT
JP
KR
NL
NZ
PTES
CH
US
05
1015
Infla
tion
0 5 10 15Corrected Money Growth
Corrected for yield − estimated1970*−2005
Corrected money growth rate here is average M1 growth minus average real GDP growth plus the average growth
of a nominal interest rate, multiplied by an estimated elasticity, minus a regression constant. Average CPI inflation
is plotted vis-a-vis corrected money growth. For each country, the average of the variables is from 1970 (or later if
data is unavailable) until 2005. The fit around the 45 degree line is similar to the Miller-Orr specification.
15
4 Subsamples
4.1 The Quantity Theory with inflation targeting
We now draw attention to the results before and after the implicit or explicit adoption of inflation
targeting (IT). For each country we split the full sample in IT and non-IT periods and consider this
splitting to calculate averages, run regressions with these averages or consider (unbalanced) panel
cointegration regressions. The specific dates and justification for these choices are in appendix A.
Starting with the corrections to average money growth, we note that in all samples the Miller-Orr
specification as well as the estimated specifications, which point to a lower interest rate elasticity,
clearly fit better than the Baumol-Tobin specification. But what is more striking is that while the fit
for all specifications in the non-IT part of the sample is essentially as good as for the whole sample,
the fit, as measured by an R2 is much worse for the IT part of the sample, as table 1 shows. Also,
not including a constant in the regression, i.e., not controlling for movements in velocity (through
a linear trend) not attributable to movements in the short rate deteriorates dramatically the fit in
the more recent sample.
Table 1Measures of fit and error dispersion
Period Benchmark Corrections Regression Based Corrections
Not GDP Baumol- Miller- γ = 1, α est. γ = 1, α est.corrected growth Tobin Orr with Constant No Constant
1970*-2005 R2 -2.79 0.31 0.37 0.62 0.65 0.64s.d.(ϵi) 4.76 2.04 1.94 1.51 1.45 1.46
s.d.(∆ log(M/P )) 2.451970*-IT R2 -1.18 0.52 0.23 0.56 0.69 0.66
s.d.(ϵi) 4.47 2.10 2.65 2.00 1.67 1.77s.d.(∆ log(M/P )) 3.02
IT-2005 R2 -6.45 -0.93 -0.18 0.10 0.43 0.10s.d.(ϵi). 5.59 2.94 2.29 2.01 1.59 2.01
s.d.(∆ log(M/P )) 2.12
R2 is calculated as 1− SSR/SST , where SSR is the sum of squared residuals, where the residuals are the differences between average
inflation and (possibly adjusted) average money growth and SST is the total variation in real money growth). s.d.(ϵi) is the estimated
standard deviation of the residuals. s.d.(∆ log(M/P )) is the standard deviation of real money growth. Averages of the variables are
from 1970 (or whenever data is available) until 2005, denoted 1970*-2005; from 1970 (or whenever data is available) until IT adoption,
denoted 1970*-IT and from IT adoption until 2005, denoted IT-2005. Countries included: AU, AT, CA, DK, FIN, FR, DE, IE, IT, JP ,
KR, NL, NZ, PT, ES, CH, US (17 observations). 1970*-IT does not include DE and CH. IT-2005 does not include JP.
The figures provide an even more revealing story. Figure 7 shows the results for the Baumol-
Tobin specification across samples, figure 8 for the Miller-Orr specification, and figure 9 shows the
16
Table 2Regression results
Period Benchmark No constant Constant includedPreferred spec.
γ = 1 γ = 1α R2 α Const. R2
1970*-2005 0.26 0.64 0.34 -0.44 0.65(0.07) (0.14) (0.77)
1970*-IT 0.18 0.66 0.26 -0.74 0.69(0.08) (0.10) (0.59)
IT-2005 0.33 0.10 -0.01 2.51 0.43(0.08) (0.13) (0.88)
Regression of (averages of) real money growth on real output growth and the nominal interest rate. Standard errors below the
estimates. R2 is calculated as 1 minus (sum-of-squared of residuals divided by total variation of real money growth). Averages of the
variables are from 1970 (or whenever data is available) until 2005, denoted 1970*-2005; from 1970 (or whenever data is available) until
IT adoption, denoted 1970*-IT and from IT adoption until 2005, denoted IT-2005. Countries included: AU, AT, CA, DK, FIN, FR,
DE, IE, IT, JP , KR, NL, NZ, PT, ES, CH, US (17 observations). 1970*-IT does not include DE and CH. IT-2005 does not include JP.
Table 3Panel cointegration regression results
Estimation-DOLS Constant Constant and common trendPreferred specification
Period benchmark γ = 1 γ = 1
Grouped Pooled Grouped Pooledα α α α
1970*-2005 0.24 0.18 0.13 0.10(0.02) (0.01) (0.02) (0.01)
1970*-IT 0.26 0.20 0.25 0.19(0.04) (0.02) (0.04) (0.02)
IT-2005 0.25 0.17 0.06 0.02(0.02) (0.01) (0.02) (0.01)
Estimation-FMOLS Constant Constant and common trend
Period benchmark γ = 1 γ = 1
Grouped Pooled Grouped Pooledα α α α
1970*-2005 0.24 0.29 0.14 0.20(0.02) (0.02) (0.02) (0.02)
1970*-IT 0.25 0.24 0.25 0.19(0.03) (0.03) (0.03) (0.03)
IT-2005 0.22 0.34 0.05 0.13(0.02) (0.05) (0.01) (0.05)
Estimation by DOLS and FMOLS considering country fixed-effects only and country fixed-effects and a common time trend in the
cointegration equation. In each setting we consider Grouped and Pooled (weighted) estimation of the panel. Pooled (weighted)
estimation considers cross-section estimates of the error covariances. With FMOLS we consider heterogeneous first-stage long-run
coefficients. Using DOLS the uit are allowed to be correlated with at most pi leads and lags of vit. We choose pi according to the AIC
criterion. HAC standard errors are below the estimates. For each country, data is from 1970 (or whenever available) until 2005, denoted
1970*-2005; from 1970 (or whenever data is available) until IT adoption, denoted 1970*-IT and from IT adoption until 2005, denoted
IT-2005. Countries included: AU, AT, CA, DK, FIN, FR, DE, IE, IT, JP , KR, NL, NO, NZ, PT, ES, CH, US. 1970*-IT does not
include DE and CH
17
results for the estimated specification.
Results for the simple regression with averages are in table 2 and those for the panel cointegration
regression imposing a unitary output elasticity (γ = 1) are in table 3. More complete versions of
the results, including choosing 1990 as the split point, can be found in the online appendix.9 We
should mention that, in our view, the most appropriate specifications include a constant in the
simple regression and, additionally, a trend in the cointegration equation.10
We want to stress two points here. The first is that the estimated interest elasticity is lower for
the IT part of the sample. This result has been observed in the literature on the stability of the
money demand using time series data, as for example in Ireland (2009). Part of the explanation for
the low elasticity is the increasing role of money substitutes that are not included in M1, as argued
by Teles and Zhou (2005), and recently also by Lucas and Nicolini (2013).
The second result, more important for our purposes, is the apparent poor fit of a money growth-
inflation relationship in the IT part of the sample compared to the non-IT part of the sample or
the whole sample, see table 1. It is still the case, however, and very importantly, that the residual
variance, or variability around the 45 degree line in the inflation-(adjusted) money growth plots, is
comparable to the other samples.
The reason for the apparent poor fit of the quantity theory relationship in the IT sample is
that inflation is nearly the same across countries. Without variation in inflation there can be no
one-to-one relationship between inflation and money growth. Examining figures 7, 8 and 9 makes
this point in a striking way. The explanation is simple. Central banks have increasingly focused
on achieving a low and roughly common target for the inflation rate. Apparently, they have been
successful in achieving this goal. Central banks choose a money growth rate that offsets shocks
to the money-inflation relationship, in order to achieve their common target. There is residual
dispersion in money growth, probably due to differing experiences in deregulation and innovation
in transactions technologies, but not enough to question the tightness of the relation as measured
by the residual variance.
9The main messages do not change if we consider that alternative split point.10This way we control, even if in a crude way, for movements in velocity not attributable to changes in interest
rates or output. Not including a constant in the regressions (or a trend in the cointegration equation) has a bigimpact in the estimated elasticities for the IT-2005 sample due to the fact that these movements are being attributedto the fall in interest rates over this period. This is a typical bias due to the omission of deterministics.
18
The fact that the Quantity Theory one-to-one relationship between money growth and inflation
cannot be found in the data does not mean that it is not a feature of a model that generates
comparable data. In the simple model of section 3, if inflation was equal to 2% for every country
and if interest rates were constant over time (possibly equal to 4%), that would make it impossible
to identify the relationship in the data generated by the model, regardless of the interest elasticity.
Even if the relationship is indeed a feature of the model. In this sense, the model can trivially be
used to explain the subsample instability.
Interestingly, the recent work of Sargent and Surico (2011) makes similar points using the time
series evidence for the US, also used in Lucas (1980). They argue that part of the difficulty in
establishing a relationship between money and inflation in the US in the more recent data is due to
the policy of inflation targeting. This has reduced the intertemporal variability of inflation, making
it hard to find a one-to-one low-frequency relationship between money growth and inflation in the
US time series.
19
Figure 7
AU
AT
CADK
FINFR
DE
IE
IT
JP
KR
NL
NZ
PTES
CH UK
US
05
1015
Infla
tion
0 5 10 15Corrected Money Growth
Corrected for yield − Baumol−Tobin1970*−2005
AU
AT
CADKFIN
FR IE
IT
JP
KR
NL
NZ
PT ES
UK
US
05
1015
Infla
tion
−5 0 5 10 15Corrected Money Growth
Corrected for yield − Baumol−Tobin1970*−IT
AUATCA DK
FINFR
DEIE
ITKRNLNZ
PTES CH
UK
US
05
1015
Infla
tion
−5 0 5 10 15Corrected Money Growth
Corrected for yield − Baumol−TobinIT−2005
Baumol-Tobin: Corrected money growth rate here is average M1 growth minus average real GDP growth plus the
average growth of a nominal interest rate, divided by two, following 10 as well as the suggestion of Lucas (2000).
Average CPI inflation is plotted vis-a-vis corrected money growth. For each country, the average of the variables is
from 1970 (or whenever data available) until 2005, denoted 1970*-2005; from 1970 (or whenever data is available)
until IT adoption, denoted 1970*-IT and from IT adoption until 2005, denoted IT-2005.
20
Figure 8
AU
AT
CADK
FINFR
DE
IE
IT
JP
KR
NL
NZ
PTES
CH UK
US
05
1015
Infla
tion
0 5 10 15Corrected Money Growth
Corrected for yield − Miller−Orr1970*−2005
AU
AT
CA DKFIN
FR IE
IT
JP
KR
NL
NZ
PT ES
UK
US
05
1015
Infla
tion
0 5 10 15Corrected Money Growth
Corrected for yield − Miller−Orr1970*−IT
AUATCA DK
FINFR
DEIE
ITKRNLNZ
PTES CH
UK
US
05
1015
Infla
tion
0 5 10 15Corrected Money Growth
Corrected for yield − Miller−OrrIT−2005
Miller-Orr: Corrected money growth rate here is average M1 growth minus average real GDP growth plus the
average growth of a nominal interest rate, divided by three, capturing the transactions technology model due to
Miller and Orr (1966). Average CPI inflation is plotted vis-a-vis corrected money growth. For each country, the
average of the variables is from 1970 (or whenever data available) until 2005, denoted 1970*-2005; from 1970 (or
whenever data is available) until IT adoption, denoted 1970*-IT and from IT adoption until 2005, denoted IT-2005.
The fit around the 45 degree line is better when compared to the Baumol-Tobin in all the samples.
21
Figure 9
AU
AT
CADK
FINFR
DE
IE
IT
JP
KR
NL
NZ
PTES
CH
US
05
1015
Infla
tion
0 5 10 15Corrected Money Growth
Corrected for yield − estimated1970*−2005
AU
AT
CA DKFIN
FR IE
IT
JP
KR
NL
NZ
PT ES
US
05
1015
Infla
tion
0 5 10 15Corrected Money Growth
Corrected for yield − estimated1970*−IT
AUATCA DK
FINFR
DEIE
ITKRNLNZ
PTESCHUS
05
1015
Infla
tion
−5 0 5 10 15Corrected Money Growth
Corrected for yield − estimatedIT−2005
Corrected money growth rate here is average M1 growth minus average real GDP growth plus the average growth
of a nominal interest rate, multiplied by an estimated elasticity, minus a regression constant. Average CPI inflation
is plotted vis-a-vis corrected money growth. For each country, the average of the variables is from 1970 (or
whenever data available) until 2005, denoted 1970*-2005; from 1970 (or whenever data is available) until IT
adoption, denoted 1970*-IT and from IT adoption until 2005, denoted IT-2005. The fit around the 45 degree line is
similar to the Miller-Orr specification.
22
5 Conclusions
A cross section of long term averages for inflation and money growth plotted one against the other as
in e.g. McCandless and Weber (1995) has those averages line up nicely along a 45 degree line. In his
Nobel lecture, Lucas (1996) claims that there is no sharper evidence in monetary economics.11 But
the evidence is by no means as sharp when the sample excludes countries with very high inflation.
For countries with moderate inflation, the overwhelming evidence is just not there.
We use a cross section of countries with moderate inflation to reestablish the one-to-one close
relationship between long term inflation and money growth. For that we need to take into account
the effect of long term movements in nominal interest rates, according to elasticities that match
the ones suggested by both theory on transactions technologies, as in Baumol (1952), Tobin (1956),
Miller-Orr (1966), and estimation using time series data for the US and other countries. Once we
take into account the effect of movements in interest rates for the whole sample, between 1970 and
2005, what appeared to be a random scatter of points is now a 45 degree line through the origin.12
The data is split into two subsamples with the data break coinciding with the explicit or implicit
adoption of inflation targeting. We find that for the later part of the sample the one-to-one relation
between inflation and money growth seems to deteriorate. Such deterioration is only apparent since
the residual variance is comparable to the one found for the full sample. The reason for all this is
moderate dispersion in money growth coupled with successful inflation targeting at a low common
rate. In the later sample the variability of inflation is indeed considerably reduced. The points
seem to form an horizontal line at low inflation. With very low variability of inflation it is hard
to find any relationship between inflation and money growth. A similar difficulty was described by
Whiteman (1984) and met by Sargent and Surico (2011) in their review of Lucas (1980). Sargent
and Surico also find that inflation targeting around low inflation, reducing its variability, made it
hard to extract from the more recent US data the one-to-one relationship that Lucas found. Our
results here complement their findings, using a cross-country analysis compared to their US time
11”(...)Central bankers and even some monetary economists talk knowledgeably of using high interest rates tocontrol inflation, but I know of no evidence from even one economy linking these variables in a useful way, let aloneevidence as sharp as that displayed in figure 1. The kind of monetary neutrality shown in this figure needs to be acentral feature of any monetary or macroeconomic theory that claims empirical seriousness.(...)”
12For the line to be through the origin, the effect of output growth must also be taken into account.
23
series analysis.
References
[1] Assenmacher-Wesche, Katrin and Stefan Gerlach, 2007, “Money at Low Frequencies,” Journal
of the European Economic Association 5 (2-3), 534-542, 04-05.
[2] Atkeson, Andrew, V.V. Chari, Patrick J. Kehoe, 2010, “Sophisticated Monetary Policies,”
Quarterly Journal of Economics 125, 47-89.
[3] Alvarez, Fernando and Francesco Lippi, 2009, ”Financial Innovation and the Transactions
Demand for Cash,” Econometrica 77, 363-402.
[4] Attanasio, Orazio P., Luigi Guiso and Tullio Jappelli, 2002, “The Demand for Money, Financial
Innovation and the Welfare Cost of Inflation: An Analysis with Household Data,” Journal of
Political Economy 110, 317-351.
[5] Bailey, Martin J., 1956, “The Welfare Costs of Inflationary Finance,” Journal of Political
Economy 64, 93-110.
[6] Ball, Laurence, 2001, “Another Look at Long-run Money Demand,” Journal of Monetary
Economics 47, 31-44.
[7] Barnett, William A., 1980, “Economic Monetary Aggregates: An Application of Aggregation
and Index Number Theory,” Journal of Econometrics 14, 11-48.
[8] Barnett, William A. and Apostolos Serletis, 2000, The Theory of Monetary Aggregation, Con-
tributions to Economic Analysis 245, North-Holland, Amsterdam.
[9] Barro, Robert, 1993, Macroeconomics, 4th edition, John Wiley & Sons, New York.
[10] Barro, Robert, 2007, Macroeconomics: A Modern Approach, South-Western College Pub, 1st
edition.
24
[11] Baumol, William J., 1952, “The Transactions Demand for Cash: An Inventory Theoretic
Approach,” Quarterly Journal of Economics 66, 545-556.
[12] Benhabib, Jess, Stephanie Schmitt-Grohe and Martin Uribe, 2001, “The Perils of Taylor
Rules,” Journal of Economic Theory 96, 40-69.
[13] Benati, Luca, 2009, “Long Run Evidence on Money Growth and Inflation,” Working Paper
Series 1027, European Central Bank.
[14] Breitung, Jorg, 2000, “The Local Power of Some Unit Root Tests for Panel Data,” in B.
Baltagi (ed.), Advances in Econometrics, Vol. 15: Nonstationary Panels, Panel Cointegration,
and Dynamic Panels, JAI Press, Amsterdam, 161–178.
[15] Carlson, John B., Dennis L. Hoffman, Benjamin D. Keen, and Robert H. Rasche, 2000, “Results
of a study of the stability of cointegrating relations comprised of broad monetary aggregates,”
Journal of Monetary Economics 46, 345-383.
[16] Choi, In, 2001. “Unit Root Tests for Panel Data,” Journal of International Money and Finance
20, 249–272.
[17] Coenen Gunter and Juan-Luis Vega, 2001, “The demand for M3 in the euro area,” Journal of
Applied Econometrics 16, 727- 748.
[18] Fischer, Bjorn, Michele, Lenza, Huw Pill and Lucrezia Reichlin, 2006, “Money and Monetary
Policy: The ECB Experience 1999-2006,” mimeo, European Central Bank.
[19] Friedman, Milton, 1956, “The Quantity Theory of Money - A Restatement,” in his Studies in
the Quantity Theory of Money, Chicago.
[20] Ireland, Peter, 2009, “On the Welfare Cost of Inflation and the Recent Behavior of Money
Demand,” American Economic Review 99, 1040-52.
[21] Im, K. S., M. H. Pesaran, and Y. Shin, 2003. “Testing for Unit Roots in Heterogeneous Panels,”
Journal of Econometrics 115, 53–74
25
[22] Hadri, Kaddour, 2000, “Testing for Stationarity in Heterogeneous Panel Data,” Econometric
Journal 3,148–161.
[23] Hammond, Gill, 2012, “State of the art of Inflation Targeting,” Centre for Central Banking
Studies, Handbook No. 29, Bank of England.
[24] Kao, Chinwa D., 1999, “Spurious Regression and Residual-Based Tests for Cointegration in
Panel Data,” Journal of Econometrics 90, 1–44
[25] Kao, Chihwa and Min-Hsien Chiang, 2000, “On the Estimation and Inference of a Cointe-
grated Regression in Panel Data,” in Baltagi, B. H. et al. eds., Nonstationary Panels, Panel
Cointegration and Dynamic Panels, 15, Elsevier, Amsterdam, 179–222.
[26] Levin, Andrew, Chien-Fu Lin, and Chia-Shang Chu, 2002. “Unit Root Tests in Panel Data:
Asymptotic and Finite-Sample Properties,” Journal of Econometrics 108, 1–24.
[27] Lucas, Robert E., Jr., 1980, “Two Illustrations of the Quantity Theory of Money,” American
Economic Review 70, 1005-1014.
[28] Lucas, Robert E., Jr., 1996, “Nobel Lecture: Monetary Neutrality,” Journal of Political Econ-
omy, vol. 104, no. 4, 661-682.
[29] Lucas, Robert E., Jr., 2000, “Inflation and Welfare”, Econometrica 68, 247-274.
[30] Maddala, G. S. and Shaowen Wu, 1999, “A Comparative Study of Unit Root Tests with Panel
Data and A New Simple Test,” Oxford Bulletin of Economics and Statistics 61, 631–52.
[31] Mark, Nelson C. and Donggyu Sul, 2003, “Cointegration Vector Estimation by Panel DOLS
and Long-run Money Demand,” Oxford Bulletin of Economics and Statistics 65, 655–680.
[32] McCandlees, JR., George T. and Warren Weber, 1995, “Some Monetary Facts”, Federal Re-
serve Bank of Minneapolis Quarterly Review 19 (3), 2-11
[33] Miller, Merton H. and Daniel Orr, 1966, “A Model of the Demand for Money by Firms,”
Quarterly Journal of Economics 80, 413-435.
26
[34] Pedroni, Peter, 1999. “Critical Values for Cointegration Tests in Heterogeneous Panels with
Multiple Regressors,” Oxford Bulletin of Economics and Statistics 61, 653–70.
[35] Pedroni, Peter, 2000, “Fully Modified OLS for Heterogeneous Cointegrated Panels,” in Balt-
agi, B. H. ed., Nonstationary Panels, Panel Cointegration and Dynamic Panels, 15, Elsevier,
Amsterdam, 93–130.
[36] Pedroni, Peter, 2001, “Purchasing Power Parity Tests in Cointegrated Panels,” The Review of
Economics and Statistics 83, 727–731.
[37] Pedroni, Peter, 2004. “Panel Cointegration; Asymptotic and Finite Sample Properties of Pooled
Time SeriesTests with an Application to the PPP Hypothesis,” Econometric Theory 20, 597–
625.
[38] Phillips, Peter C. B. and Hyungsik R. Moon,1999, “Linear Regression Limit Theory for Non-
stationary Panel Data,” Econometrica 67, 1057-1111.
[39] Sargent, Thomas and Paolo Surico, 2011, “Two Illustrations of the Quantity Theory of Money,”
American Economic Review 101, 109-128.
[40] Teles, Pedro and Ruilin Zhou, “A Stable Money Demand: Looking for the Right Monetary
Aggregate,” Federal Reserve Bank of Chicago, Economic Perspectives 1Q/2005, 50-63.
[41] Tobin, James, 1956, “The Interest Elasticity of the Transactions Demand for Cash,” Review
of Economics and Statistics 38, 241-247.
[42] Whiteman, Charles H., 1984, “Lucas on the Quantity Theory: Hypothesis Testing without
Theory,” American Economic Review 74, 742–49.
[43] Woodford, Michael, 2008, “How Important Is Money in the Conduct of Monetary Policy?,”
Journal of Money, Credit and Banking 40, 1561-1598.
27
A Data Description
The list of countries included in the regressions starts from the OECD countries excluding Chile,
Iceland, Israel, Mexico and Turkey, due to high inflation during relevant parts of the sample.
We furthermore excluded Luxembourg (as it is a financial hub in small country), the UK in the
regression but not in several plots (as it became a large financial hub during this period, with the
consequence that it is an extreme outlier in all the exercises), Japan for the sample IT-2005 to avoid
zero bound considerations and the transition countries (since there is no useful data for the purpose
of the analysis here from 1970 and 1990). For all other countries, we attempted to obtain as much
reliable data as possible, dropping Belgium, Greece, and Sweden due to missing data. In the end, 19
countries remain in the sample for at least part of the calculations. Table 4 lists the country codes.
An Excel file containing all the data as well as detailed remarks regarding sources and corrections
is available as part of an online appendix to the paper. Likewise, the E-Views and Stata programs
that perform all the calculations and produce the graphs as part of an online appendix to the paper.
An original version of the data used was collected by Jan Auerbach, an undergraduate RA in
Berlin 2006, using EcoWin, a commercial data base, which was available and existent then. Ding
Xuan, an undergraduate RA in Chicago 2012, corrected a few entries, using IMF and World Bank
Data. A number of further issues then were dealt with by the authors. In a nutshell, the latest
version of annual CPI and real GDP data is in most instances from the IMF-IFS. As for the short
nominal interest rate the main source is also the IMF-IFS. Data from the OECD, the ECB and
from St. Louis Fed’s FRED database is also used. As for M1, we rely on FRED, the IMF-IFS and
the ECB. Euro zone countries do not have an independent series for M1 for the whole sample, but
data for their contribution to M1 is provided by the ECB from 1980 onwards. For Germany, the M1
as well as the real GDP series was ”spliced” across unification, per extending the series backwards
using the growth rates of West Germany. Details can be found in table 6. Since the paper at
hand focusses on prices, M1, real GDP and short-term interest rates, the data regarding M2, M3
and long-term rates would need further corrections before full use, but appears to be sufficient to
provide a “first pass” at the results.
28
Code CountryAU AustraliaDK DenmarkDE GermanyFI FinlandFR FranceIE IrelandIS IcelandIT ItalyJP JapanCA CanadaKR S.KoreaNZ New ZealandNL NetherlandsNO NorwayAT AustriaPT PortugalCH SwitzerlandES SpainUK United KingdomUS US
Table 4Country Codes
Country Year of IT adoption Source and/or justificationAU 1993 Gill Hammond (2012)DK 1995 Nominal interest rate reaches average of subsequent periodDE 1970 Very low inflation throughout the sampleFI 1993 Bank of Finland statementFR 1999 Adoption of the Euro; no previous statement or low πIE 1999 Adoption of the Euro; no previous statement or low πIT 1999 Adoption of the Euro; no previous statement or low πJP 2001 Gill Hammond (2012)CA 1991 Gill Hammond (2012)KR 1998 Gill Hammond (2012)NZ 1989 Gill Hammond (2012)NL 1999 Gill Hammond (2012)NO 2001 Gill Hammond (2012)AT 1999 Adoption of the Euro; no previous statement or low πPT 1999 Adoption of the Euro; no previous statement or low πCH 1970 Very low inflation throughout the sampleES 1994 Bank of Spain statementUK 1992 Gill Hammond (2012)US 1984 Onset of great Moderation, common break point
Table 5Beginning of inflation targeting dates
29
Series
Fre
q.
Countries
Sourc
eSam
ple
Desc
ription
and
Note
s
P Consu
merPriceIn
dex,All
Item
s,2010=100
Annual
All
IMF
IFS
1970-2
005
Data
forW
est
Germ
any
before
1991,
Seriesexte
nded
backward
sis
ourcalculation
Y RealGDP,in
dex
2010=100,SA
Annual
All
exc.DE
and
PT
IMF
IFS
1970-2
005
RealGDP
at2010
Mark
etprices
Annual
PT
EC
AM
ECO
1970-2
005
RealGDP
at2010
Mark
etprices
Annual
DE
EC
AM
ECO
1970-2
005
Data
forW
est
Germ
any
before
1991,
Seriesexte
nded
backward
sourcalculation
R Nomin
alIn
tere
stRate
:DiscountRate
Annual
AT
IMF
IFS
and
ECB
1970-2
005
ECB
rate
from
1999
onward
s,Dec.31
valu
eand
ECB
Main
Refinancin
gOpera
tionsra
teDepositRate
Annual
CA
IMF
IFS
1971-2
005
Moneta
ryPolicy
Inte
rest
Rate
Annual
DK
IMF
IFS
1970-2
005
Nomin
alIn
tere
stRate
:DiscountRate
Annual
FIN
IMF
IFS
and
ECB
1970-2
005
ECB
rate
from
1999
onward
s,Dec.31
valu
eand
ECB
Main
Refinancin
gOpera
tionsra
teNomin
alIn
tere
stRate
:3-m
onth
T-b
illra
teAnnual
FR
IMF
IFS
and
ECB
1970-2
005
ECB
rate
from
1999
onward
s,Dec.31
valu
eand
ECB
Main
Refinancin
gOpera
tionsra
teNomin
alIn
tere
stRate
:Short
Rate
,T-b
illra
teAnnual
DE
OECD,IM
FIF
Sand
ECB
1970-2
005
OECD
short
Rate
from
1970-1
974,T
-Billfrom
1975-1
998,
and
ECB
Main
Refinancin
gOpera
tionsra
teECB
rate
from
1999
onward
s,Dec.31
valu
eNomin
alinte
rest
rate
:DiscountRate
Annual
ISIM
FIF
S1970-2
005
and
ECB
Main
Refinancin
gOpera
tionsra
teNomin
alinte
rest
rate
:Discountra
teAnnual
IEIM
FIF
S,OECD
and
ECB
1970-2
005
IMF
Discountra
teuntil1999,1992
filled
by
Dec1992
OECD
Short
rate
,and
ECB
Main
Refinancin
gOpera
tionsra
teECB
rate
from
1999
onward
s,Dec.31
valu
eNomin
alinte
rest
rate
:DiscountRate
Annual
IT,PT,ES
IMF
IFS
and
ECB
1970-2
005
ECB
rate
from
1999
onward
s,Dec.31
valu
eand
ECB
Main
Refinancin
gOpera
tionsra
teNomin
alinte
rest
rate
:DepositRate
Annual
JP
IMF
IFS
1970-2
005
Nomin
alinte
rest
rate
:DepositRate
Annual
KR
IMF
IFS
1970-2
005
Nomin
alinte
rest
rate
:Call
Money
Rate
Annual
NL
Sta
tisticsNL
and
ECB
1970-2
005
Sta
tisticsNeth
erlandsDecembervalu
es
and
ECB
Main
Refinancin
gOpera
tionsra
teECB
rate
from
1999
onward
s,Dec.31
valu
eNomin
alinte
rest
rate
:Short
rate
and
T-b
illra
teAnnual
NZ
IMF
IFS,OECD
1973-2
005
Short
Rate
OECD
1973-1
977,T-b
illfrom
1978
onward
s,hole
for1985
filled
by
Dec.valu
eofOECD
short
rate
Nomin
alIn
tere
stRate
:3
month
T-b
illra
teAnnual
GB
FRED
INTGSTGBM
193N
1977-2
005
Decembervalu
esfrom
month
lyse
ries
Nomin
alIn
tere
stRate
:3
month
T-b
illra
teAnnual
US
FRED
TB3M
S1970-2
005
Decembervalu
esfrom
month
lyse
ries
M Moneta
ryaggr.
M1,NationalCurrency,SA
Annual
US,KR
IMF
IFS
1970-2
005
Moneta
ryaggr.
M1,NationalCurrency,SA
Annual
AU
FRED
MANM
M101AUA189S
1970-2
005
Dec.valu
esfrom
month
lyse
ries
Moneta
ryaggr.
M1,NationalCurrency,SA
Annual
CA
FRED
MANM
M101CAA189S
1970-2
005
Dec.valu
esfrom
month
lyse
ries
Moneta
ryaggr.
M1,NationalCurrency,NSA
Annual
CH
FRED
MANM
M101CHA189N
1970-2
005
Dec.valu
esfrom
month
lyse
ries
Moneta
ryaggr.
M1,NationalCurrency,NSA
Annual
DK
FRED
MANM
M101DKA189N
1970-2
005
Dec.valu
esfrom
month
lyse
ries
Moneta
ryaggr.
M1,NationalCurrency,NSA
Annual
GB
FRED
MANM
M101GBA189N
1986-2
005
Dec.valu
esfrom
month
lyse
ries
Moneta
ryaggr.
M1,NationalCurrency,NSA
Annual
ISFRED
MANM
M101IS
A189N
1970-2
005
Dec.valu
esfrom
month
lyse
ries
Moneta
ryaggr.
M1,NationalCurrency,NSA
Annual
NO
FRED
MANM
M101NOA189N
1970-2
005
Dec.valu
esfrom
month
lyse
ries
Moneta
ryaggr.
M1,NationalCurrency,NSA
Annual
NZ
FRED
MANM
M101NZA189N
1970-2
005
Dec.valu
esfrom
month
lyse
ries
Moneta
ryaggr.
M1,NationalCurrency,SA
Annual
JP
FRED
MANM
M101JPA189S
1970-2
005
Dec.valu
esfrom
month
lyse
ries
Moneta
ryaggr.
M1,NSA
Annual
AT,DE
IMF
IFS
and
ECB
1970-2
005
Data
from
ECB
startin
gin
1980,using
Dec.valu
esfrom
origin
alm
onth
lyse
ries;
exte
nded
backward
susing
gro
wth
rate
sofIM
FIF
SAnnualData
Moneta
ryaggr.
M1,NSA
Annual
ES
IMF
IFS
and
ECB
1970-2
005
Data
from
ECB
startin
gin
1980,using
Dec.valu
esfrom
origin
alm
onth
lyse
ries;
exte
nded
backward
susing
gro
wth
rate
sofIM
FIF
SAnnualData
.Use
May
2005
instead
ofDec.dueto
very
larg
esp
ike
Moneta
ryaggr.
M1,NSA
Annual
FR
IMF
IFS
and
ECB
1977-2
005
Data
from
ECB
startin
gin
1980,using
Dec.valu
esfrom
origin
alm
onth
lyse
ries;
exte
nded
backward
susing
gro
wth
rate
sofIM
FIF
SAnnualData
Moneta
ryaggr.
M1,NSA
Annual
FIN
,IE
,IT
,ECB
1980-2
005
Dec.valu
esfrom
origin
alm
onth
lyse
ries.
IEwith
larg
esp
ikein
Jan
2003.
PT,NL
Wese
tJan
2003=Dec2012
and
exte
nd
serieswith
subse
quentgro
wth
rate
s
Tab
le6
DataDescription
30