Professional Documents
Culture Documents
Money Demand Paper - Final
Money Demand Paper - Final
Abstract
The aim of this paper is to investigate the degree of currency substitution between the
currencies of CEE countries and the euro. As a novelty, we develop a model with
substitution and the money demand sensitivity to exchange rate variations. More precisely,
we posit that the currency substitution is related to the money demand sensitivity to the
interest rate spread between the CEE countries and the euro area. In addition, we show that
the existence of a channel throughout the exchange rate has implications on the money
demand, even in the absence of a currency substitution effect. This model can be successfully
applied to countries where an international currency offers liquidity services to the domestic
agent, and where afterwards it is parameterized in order to empirically test the long-run
money demand based on two complementary cointegration equations. The opportunity cost
of holding the money, as well as the scale variable represented by household consumption or
output, explain the long-run money demand in CEE countries. Our results are robust
regarding the use of DOLS or FMOLS estimators, and regarding the employment of the
broad and the narrow money for computing the money demand.
Keywords: money demand, open economy model, currency substitution, cointegration, CEE
countries
1
1. Introduction
The standard model of the money demand function (in an open economy) used in the
literature is a two-country portfolio balance model such as the one described in Leventakis
(1976) critique and to disputes.1 Because of its static nature, the estimated money demand
may appear unstable if the monetary policy strategy is modified, even in a limited way.
The empirical studies of the money demand typically do not provide a micro-founded
theoretical model to justify the specification of their empirical money demand functions.2
Moreover, all these models test the existence of a currency substitution phenomenon through
the exchange rate sensitivity of the money demand. Such a framework presents important
of the other (exchange rate sensitivity). Moreover, these models examine the money demand
sensitivity to international factors, without being able to differentiate between the currency
In this context, we contribute to the existing literature in two ways. First, we advance a
throughout the exchange rate impacts on money demand, even in the absence of a currency
substitution effect. Therefore, it appears that the money demand is subject to exchange rate
fluctuations, even after removing any possibility of currency substitution, because the
exchange rate affects the liquidity service associated with the foreign money holding. Indeed,
our model allows us to measure the currency substitution intensity without explicitly
considering the exchange rate. In addition, the model is compatible with both currency
substitution and currency complementarity hypotheses, which enable the assessment of the
1
According to Hueng (1998), Dreger et al. (2007) and Hsieh and Hsing (2009), the overall effect of the
exchange rate on the domestic money demand is not straightforward. Moreover, it is not clear either if it is the
level, or the (expected) exchange rate variation, that should enter into the money demand equation.
2
Chen (1973), Miles (1978), Bordo and Choudri (1982) and Hueng (1998) are exceptions.
2
currency substitution intensity. Further, in order to capture the recent economic
circumstances, where interest rates recorded negative values for short periods, we consider an
additional opportunity cost of holding the monetary asset (domestic or foreign). Given this
additional cost, the overall opportunity cost remains positive, even if the base cost, associated
Our micro-foundations model, which allows the integration of the liquidity production
close to Miles (1978), who, however, does not consider the consumption choice. This way we
demonstrate that Bordo and Choudri’s (1982) criticism addressed to Miles’s (1978) money
demand equation is not relevant, as the equation is not a result of the omission of the
Second, we parameterize this model in order to test the long-run sensitivity of the
money demand to international factors. In particular, our model fits the Central and Eastern
European (CEE) case, where an international currency offers liquidity services to the
domestic agents.3 The money demand in the CEE countries is investigated inter alia by
Dreger et al. (2007), Hsieh and Hsing (2009) and Fidrmuc (2009). These empirical analyses,
however, do not have a theoretical framework. Therefore, in order to overcome this limit,
based on the proposed money demand model with microeconomic foundations, we test the
existence of a long-run relationship between the money demand and its explanatory variables
based on two cointegrating equations. While the first equation investigates the sensitivity of
the real money demand for foreign currency to the opportunity cost spread of holding the
money, the second equation addresses the long-run relationship between the real money
demand for domestic currency on the one hand, and the opportunity cost of holding the
domestic currency, the opportunity cost spread, and a scale variable, on the other hand.
3
In CEE countries, the agents hold foreign money not only for foreign goods consumption, but also for
domestic goods consumption. Even if the preference for foreign money decreases progressively, a part of real
estate or cars transactions are still performed using foreign currencies, especially the euro.
3
For this purpose we use Hansen’s (1992) instability test for checking the existence of a
long-run relationship, as well as the Dynamic OLS (DOLS) and the Fully-Modified OLS
(FMOLS) methods for estimating the cointegration regressions. We employ monthly data
over the time-span 1999:M1 to 2015:M11, with an application to four CEE countries that
have, in place, floating exchange rate mechanisms, namely the Czech Republic, Hungary,
Poland and Romania.4 Even if our model is compatible with any exchange rate regime,
investigating the effect of currency substitution supposes flexibility, but not necessarily the
existence of a free-floating mechanism. This was also the sample justification of previous
The rest of the paper is structured as follows. Section 2 shortly describes the models of
money demand in an open economy. Section 3 presents our money demand model with
Two different strands of literature characterize the money demand models in an open
economy. The first strand addresses the money demand models with microeconomic
foundations and is represented by the theoretical contributions of Miles (1978), Bordo and
Choudri (1982) and Hueng (1998). The second bulk of the literature, represented by
Leventakis (1993), Dreger et al. (2007), Hsieh and Hsing (2009) and Fidrmuc (2009),
4
Croatia also has in place a managed floating exchange rate regime. However, until 2006, an exchange rate
targeting regime was used. In addition, there are no data available for the monetary aggregate M2 for Croatia.
Therefore, in order to have a consistent comparison of results for the CEE countries, we have decided to exclude
Croatia from our sample.
4
empirically tests the money demand functions, without resorting to a micro-founded
One of the first money demand models in the first category of studies is that of Miles
(1978), who uses the Chetty’s (1969) CES liquidity production function to derive the demand
for domestic money relative to foreign money.5 However, according to Bordo and Choudri
(1982), his model is misspecified due to the omission of income. In effect, in Miles’s (1978)
model, the portfolio choice does not depend on the saving-consumption decision. Therefore
the money demand is not expressed as a function of income. The money demand is derived
consequence, the ratio of domestic to foreign money demand is a function of the opportunity
Bordo and Choudri (1982) derive the demand for money from a money-in-utility-
function model. However, their model is oversimplified: it is a static model, based on the
assumption that income is entirely spent each period and based also on the perfect interest
rate arbitrage condition, thus eliminating the effect of the exchange rates.
Hueng (1998) constructs a cash-in-advance model to motivate the demand for money in
Stockman (1980), Lucas (1982) and Guidotti (1989), hinges on the assumption that domestic
and foreign goods have to be purchased with domestic and foreign currencies, respectively.
Thus the foreign money provides liquidity service only for the foreign good consumption,
and the same thing applies for domestic money in relation to the national good consumption.
This assumption is crucial, and Hueng’s (1998) model is of poor interest to describe the
money demand for agents who hold foreign money not only in relation to their consumption
of foreign good, but also in relation to their consumption of domestic good. In addition, it is
5
Chen (1973) is a special case of Miles (1978), assuming a Cobb-Douglas demand function, which constrains
the elasticity of substitution to equal one.
5
the case that in some countries (i.e., the CEE countries), the agents can partially (or even
totally) substitute an international money for their domestic money in relation to their
The second group of papers empirically examines whether currency substitution plays
Leventakis’s (1993) two-country portfolio balance model, a variation in the expected change
rate affects the demand for domestic money by inducing its substitution with foreign money,
what is referred to as the (direct) currency substitution, and with foreign bonds, what is
termed the capital mobility effect.7 If the exchange rate elasticity is high, i.e., if the money
demand in a country is very sensitive to the variations of the exchange rate of the domestic
currency with a foreign currency, this may indicate that currency substitution plays an
important role in the formation of money demand.8 If agents can switch between foreign and
Starting from this theoretical assumption, Dreger et al. (2007), Hsieh and Hsing (2009)
and Fidrmuc (2009) assess the money demand in the CEE countries. Dreger et al. (2007)
study money demand in the new EU member states over the period 1995 to 2004. A well-
behaved long-run money demand relationship can be identified only if the exchange rate is
included as part of the opportunity cost. In the long-run cointegrating vector, the output
elasticity exceeds unity. Over the entire sample period, the exchange rate vis-à-vis the US
6
In the model we propose thereafter, we make no distinction between foreign and domestic consumption goods.
The representative agent’s utility is a function of his whole consumption, which mixes foreign and domestic
goods, and of domestic and foreign currencies that produce liquidity services. The agent can invest in a portfolio
composed of domestic and foreign currencies and bonds. By maximizing the (inter-temporal) utility function,
we derive the money demand.
7
The capital mobility effect is one of the two parts of the indirect currency substitution defined by McKinnon
(1982). The second part is defined by the substitution of domestic money with domestic bonds (under the
assumption that uncovered interest parity holds, and a variation of the expected change rate induces a variation
of the domestic interest rate).
8
In Leventakis’s (1993) general model, it is impossible to isolate the separate effects of currency substitution
and capital mobility on the money demand. But if foreigners do not hold domestic currency assets, as for
example in Cuddington (1983), it becomes possible to separate their effects.
6
dollar turns out to be significant, and is considered a more appropriate variable in money
Fidrmuc (2009) investigates the money demand (between 1994 and 2003 with monthly
data) in six CEE countries (Czech Republic, Hungary, Poland, Romania, Slovakia and
Slovenia), and finds that the money demand is significantly determined by the euro interest
rate and by the exchange rate against the euro, which indicates possible instability of the
money demand in these countries. However, the exchange rate elasticity is low, which is,
according to Fidrmuc, a good precondition for the eventual adoption of the euro by the CEE
countries.9 The euro area interest rates have significantly shaped money demand in the CEE
countries, indicating that the capital mobility effect plays an important role in this region. The
coefficient estimated for the euro area interest rate is much larger than the coefficient of
domestic rates. Hsieh and Hsing (2009) find that the demand for M2 in Hungary is positively
associated with the nominal effective exchange rate and negatively influenced by the deposit
rate, the euro area interest rate, and the expected inflation rate (over the period 1995-2005).
They find an output elasticity near to unity (Fidrmuc (2009) finds lower output elasticity),
and a coefficient of the euro area interest rate higher than the domestic rate coefficient.
Nevertheless, according to Elbourne and de Haan (2006) and Fidrmuc (2009), a stable
money demand and a transmission mechanism similar to that in the euro area are likely to
create good pre-conditions for the eventual introduction of the euro by new EU member
states. Filosa (1995) and Dreger et al. (2007) also conclude that a stable money demand is a
very important condition for using monetary aggregates in the conduct of monetary policy. In
the view of all these authors, currency substitution indicates instability of money demand. A
low exchange rate elasticity would be thus a good pre-condition for the eventual adoption of
9
He also finds that the parameters of money demand in CEE countries is close to those in developed countries,
which is a good pre-condition for the euro adoption.
7
Nonetheless, this point of view is very critical. In fact, currency substitution is the
(1982), Bordo and Choudri (1982), Leventakis (1993) and Hueng (1998) remind us, if
people’s holdings of money change with foreign monetary developments, such as the foreign
interest rate and the exchange rate, it simply means that the isolation mechanism of the
floating exchange rate system will break down, thus providing the policymaker lower
controls in using stabilization policies. It is true that currency substitution reduces the
stability of the money demand in each country, but it does not mean that the global money
demand is less stable. In fact, addressing the related question of defining meaningful
appropriately defined global money supply rather than national money supplies relevant for
studying global inflation. Thus, when currency substitution occurs, it then becomes more
appropriate to conduct a global monetary policy rather than a national monetary policy. From
this point of view, currency substitution between CEE currencies and the euro is a sign of
monetary integration between the two areas and a good pre-condition for the eventual
Our research relies on both strands of literature described above. First, we propose a
money demand model with microeconomic foundations, which allows for the separation of
the currency substitution effect, and of the money demand sensitivity to exchange rate
variations. Second, we parameterize the model and we empirically test it to investigate the
8
3. The model of money demand in an open economy with microeconomic foundations
The domestic agent living in an outlying country (i.e., CEE country) is supposed to
X M S M*
Vt E t i U t i , t i , t i t i . (1)
i 0 Pt i Pt i Pt i
the price index, M t is his domestic money holding, and M *t is his foreign money holding. If
one unit of money 2 is worth S t units of money 1, St M*t represents the value measured in
terms of domestic money of the domestic agent’s foreign money holding. The operator E t
.
M t 1 1 St M *t 1 1 Bt 1 1 i t St B*t 1 1 i*t Z t X t M t St M *t B t St B*t ,
where B t is the monetary value (in terms of domestic money) of his domestic bond holding,
and B*t is the monetary value (in terms of foreign money) of his foreign bond holding, which
is an imperfect substitute for the domestic bond because of exchange rate risk; Z t is the sum
of the non-financial income and monetary transfers (from the government) perceived by the
agent; and i t 1 and i*t 1 are the nominal domestic and foreign interest rates. As bonds are
The parameter represents the cost the agent faces for holding money. This cost is
modeled as a proportional cost to simplify the analysis. It stands for the charges related to the
9
use of a bank account, the cost of a bank card, the hiring of a bank safe deposit box, and the
cash theft or loss. In standard models of money demand, the proportional cost is neglected
( 0 ) hence the interest rate cannot be negative. Considering a non-zero is the simplest
* *
X t M t 1 Pt 1
1 St M t1 Pt1 1 Bt1 Pt1 1 i t St Bt1 Pt1 1 i*t Z t
Pt Pt 1 Pt Pt 1 Pt Pt 1 Pt Pt 1 Pt Pt
. (2)
M t St M *t B t St B*t
Pt Pt Pt Pt
M t St M*t Bt S B*
The agent maximizes (1) with respect to , , and t t under (2). Let U H denote
Pt Pt Pt Pt
V
0 E t U X t U X t 1 t 1 U M t 0
P
Et t
. (3)
M
t Pt 1 Pt
Pt Pt 1
Pt
V
Et t 0 E U U
St 1 Pt
1 U 0. (4)
St M t
t Xt X t 1 *
* St M t
S P
P Pt Pt 1
t t 1 Pt
t
V
0 E t U X t U X t 1 t 1 i t 1 0 .
P
Et t (5)
Bt Pt Pt 1
Pt 1
Pt
V
Et t S P
0 E t U X t U X t 1 t 1 t 1 i*t 1 0 . (6)
St B t
*
St Pt 1
P
Pt Pt 1
t
Equations (4), (5) and (6) describe the direct and indirect currency substitution.
10
First consider equation (4). This equation results in the optimal choice of the foreign
money holding. It supposes that direct currency substitution is possible. On the contrary,
imagine that the agent cannot substitute foreign and domestic currencies. Then the agent
St M*t St M*t 1
cannot choose his level of foreign money holding which is fixed: , or in a
Pt Pt
St M*t
no currency substitution, the agent cannot decide the level of and he cannot optimize
Pt
his utility function with respect to it. So the equation (4) is not verified if there is no currency
Next consider equation (6). This equation results in the optimal choice of the foreign
bond holding. It supposes capital mobility. If international capital flows are restricted, the
agent cannot choose his level of foreign bond holding which is fixed in the extreme case:
St B*t St B*t 1
S B* S M* eb t
, or in a more general way, t t t t 1
, where eb*t is exogenous. If
Pt Pt Pt Pt
St B*t
the capital mobility condition is not verified, the agent cannot decide the level of and
Pt
Finally, consider equation (5). This equation results in the optimal choice of domestic bond
holding. Indirect currency substitution supposes that the agent can freely choose domestic bond
Bt
holding. If the agent cannot decide the level of , then equation (5) is not valid.
Pt
10
The ownership of foreign money by residents is not a proof of (direct) currency substitution. It is the
responsiveness of foreign money demand to exchange rate or to foreign interest rate which is clear evidence of
currency substitution.
11
Now suppose the absence of currency substitution, direct or indirect, which corresponds
St M*t St B*t B
to the hypothesis of exogeneity of , and t , and equations (4), (5) and (6) are not
Pt Pt Pt
valid. The agent only decides his optimal choice of domestic money, hinging on equation (3),
the only valid equation. Equation (3) shows a relationship between the current and the one-
period ahead marginal utility of consumption, foreign and domestic real cash balances, and
Xt M t S M*
with inflation rate. As the various marginal utilities depend on , and t t , equation
Pt Pt Pt
Mt X S M* X M
(3) indicates that the domestic money demand is a function of t , t t , t 1 , t 1 ,
Pt Pt Pt Pt 1 Pt 1
St 1M *t 1 P
and t . It is a somewhat complex formulation of the money demand, but we
Pt 1 Pt 1
observe that the one-period ahead exchange rate St 1 enters into this money demand function.
So, even after removing any possibility of currency substitution, it appears that the money
demand depends on the exchange rate variation, a result that is in contrast with the whole
The intuition behind this result is simple, and it is surprising that it is ignored. In effect,
the variation of the exchange rate influences the liquidity service provided by the foreign
money holding. Even if the agent will not replace domestic money with foreign currency (or
with domestic or foreign bonds) in response to changes in their relative rate of return, he can
switch between consumption and domestic money holding, in order to respond to liquidity
12
Therefore, the agent responds to the exchange rate variation by a modification of his
domestic money holding. Consequently, in any case, it could be concluded that a non-zero
exchange rate elasticity is necessarily the result of direct or indirect currency substitution.11
equation (3) shows that the domestic money demand depends on the random exchange rate
variation and inflation rate. However, the money demand is independent of domestic and
foreign interest rates. If we make the assumption, which seems realistic, that the agent
controls his domestic bond holdings, then equation (5) is valid too. Multiplying equation (3)
U Xt i t 1 U Mt 1 i t 1 0 . (7)
Pt Pt
rate or exchange rate variation. Equation (7) points out that domestic money demand can be
X t St M*t
written as a function of , and the domestic interest rate i t 1 .
Pt Pt
An important aspect of money demand is that the money demand function can be
written in many ways: equation (3) can be used to express money demand as a complex
function involving exchange rate variation and inflation rate, or a mix of equations (3) and (5)
can be used to express money demand in a way that excludes these two variables. If we add
the hypothesis that the agent controls his foreign bond holdings, which corresponds to the
assumption of capital mobility, equation (6) is also valid. If we multiply equation (6) by
U M t 1 i t 1 E t U X t t 1 t 1 i*t 1 i t 1 0
S P
Pt Pt 1 St Pt 1
11
In fact the consumption – money substitution effect we describe is possibly more important that the currency
substitution effect, depending on the value of liquidity elasticity defined hereafter.
13
This equation leads to a complex, untractable formulation of the domestic money
demand depending on all the variables considered in the model, exchange rate variation,
Finally, we add the assumption of the direct currency substitution, which means that
equation (4) is valid. We than multiply equation (4) by 1 i*t 1 1 and subtract (6) to get:
U Xt i*t 1 U S M* 1 i*t 1 0
t t
. (8)
Pt Pt
Both equations (7) and (8) depend on known (non-random) terms, and can be used to
express the domestic money demand as a function of domestic and foreign interest rates (but
Indeed, a consumption – money substitution effect also influences the money demand, and
thus the sensitivity of money demand to international variables should not be analyzed as a
consequence of the currency substitution only. And even if there is currency substitution, it is
possible to express the money demand as a function independent of exchange rate variations.
depends crucially on this hypothesis. Supposing that equations (3) to (6) are valid, we make
the assumption of indirect and direct currency substitution, which is a core hypothesis of the
model and which cannot be tested. Fortunately, the meaning of this assumption is not as strict
which could be higher or lower, consistent with a high degree of currency substitution or with
between the euro and the currencies of the CEE countries, and to estimate the intensity of the
parameters that explain the sensitivity of money demand to international economic variables.
14
4. A parameterization of the utility function
Following the example of Miles (1978), we parameterize our model by supposing that
the domestic and foreign currency are the inputs of a Constant Elasticity of Substitution
(CES) liquidity production function, and we add the assumption that the produced liquidity
X
1
Mt
St M*t
1
U t 1 , with , (9)
Pt Pt Pt
where 1 /(1 ) is the elasticity of substitution between the domestic money and the
foreign one.
In the case of a zero elasticity of substitution , the CES production function would be
equivalent to the Leontief production function, which would indicate that domestic and
foreign currencies are perfect complements. In the particular case of a unitary elasticity of
When the elasticity increases it is easier to replace one currency with another. In the
extreme case, with perfect substitution, the elasticity tends to infinity. A value of 1 is
monetary integration with the euro area is confirmed, then the currencies of these countries
must be highly substitutable with the euro. Therefore, particular attention must be paid to the
estimation.
1
Mt
St M*t
The term 1 represents the liquidity production function
t
P Pt
whose inputs are domestic and foreign money holdings, and where is the share parameter.
15
The condition > 0.5 ( < 0.5) means that the domestic money is more (less) liquid than the
euro in the eyes of the CEE countries’ representative agent. The CES liquidity production
function and the real consumption are next combined according to a Cobb-Douglas utility
1
M t
1
S M* Xt 1 i t 1 M t
1 t t , (10)
1 Pt Pt Pt 1 i t 1 Pt
and
1
M t X
1
S M* 1 i*t 1 St M*t
1 t t t
. (11)
1 Pt Pt Pt 1 1 i*t 1 Pt
As the right-hand terms of equations (10) and (11) are the same, it comes to:
St M*t 1 i t 1 i*t 1 M t
. (12)
Pt 1 i t 1 1 i*t 1 Pt
i t 1 * i*
Let’s denote oct 1 , oct 1 t 1 * the opportunity costs of holding domestic
1 i t 1 1 i t 1
and foreign monies. If 0 these opportunity costs are equal to the discounted interest
rates.13
12 1
The generalized utility function U1 leads to the same money demand function as the Cobb-Douglas U
1
function, which is independent of the risk aversion parameter . Therefore, we ignore risk aversion, as it has no
impact on the money demand equation.
13
The empirical money demand literature depicts the interest rate as the opportunity cost of money holding
when supposes that 0 , and uses it as a regressor in the money demand equation. But, as the interest rate is
perceived one period later, it has to be discounted, and it is the discounted interest rate that should enter into the
regression equation of the money demand.
16
Mt
ln ln
*
ln oct 1 ln oct 1 .
*
(13)
t t
S M 1
Equation (13) is similar to the relationship between domestic and foreign demand
money derived by Miles (1978), except that the terms 1 i t 1 and 1 i*t 1 in Miles (1978) are
replaced with oc t 1 and oc*t 1 . By integrating the liquidity production function and the
criticism of the Miles’s (1978) money demand equation is not relevant, as the equation is not
a result of omission of the consumption or income level. When the domestic money demand
variable such as the consumption or income level. The share parameter and the elasticity of
substitution are the only parameters needed to explain the shift in domestic money demand
If ln M t St M*t and ln oc t 1 ln oc*t 1 are I(1), then equation (13) describes a
cointegrating relationship.14 In the long run, equation (13) holds exactly, and so it can be seen
as a long-run money demand equation. However in the short run, because adjustment takes
Mt
ln *
0 1 ln oct 1 ln oc*t 1 t , (14)
St M t
where the elements 0 and 1 of the cointegrating vector are related to the structural
parameters by 0 ln 1 and 1 .
The relative money demand function (equation (14)) is not the only result of the model.
Our model also delivers a money demand equation dependent on a scale variable (such as
14
Standard unit root tests and panel unit root tests confirm that the variables involved in equation (17) are I(1)
processes (see Dreger et al. 2007; Fidrmuc, 2009; Hsieh and Hsing, 2009).
17
consumption or income), more consistent with the standard empirical money demand
1
M t 1 i t 1 1 St M*t X
Mt
1 t
, (15)
Pt 1 i t 1 Pt Pt Pt
1
M t 1 oc t 1 X t
1
1
oc t 1 1 * , with . (16)
P1 oc t 1 Pt
Consider the case where the considered country is an outlying country, and its currency
offers no liquidity service to the foreign agent. In this case, we suppose that the money of the
foreign country (i.e., the euro) is an international money which offers liquidity services to the
agent of the outlying but the reverse is not true. The foreign agent has no demand for money
of the outlying country, and the total demand of the money of this country is simply M t Pt .
Mt oct 1
1
X
ln ln ln oct 1 ln 1 * ln t . (17)
Pt 1 oct 1 Pt
We observe that the money demand equation (17) sticks to the standard result of a
18
If ln M t Pt , ln oc t 1 , ln oc t 1 ln oc*t 1 and ln X t Pt are I(1), then equation (18)
describes a second cointegrating relationship. In the long run, this money demand equation
holds exactly, but in the short run, the money depends also on a stationary disequilibrium t :
M X
ln t 0 1 ln oct 1 2 ln oct 1 ln oc*t 1 3 ln t t , (19)
Pt Pt
where the elements of the cointegrating vector are related to the structural parameters by
and 3 1 .
We can notice that money demand equations (14) and (19) are in the line of Meltzer
Baumol (1952) and Tobin (1956)), which consider, as explanatory variables, the log of
opportunity cost, and not the opportunity cost itself, as in Cagan’s (1956) approach.
Cagan’s (1956) semi-log form is currently the most commonly used form in the
empirical analysis of money demand, where the opportunity cost generally resumes to the
interest rates. The existence of very low interest rate levels, even negative, makes it more
difficult or impossible to use a log-log type of money demand when we have 0 . But for a
value of ϕ high enough, or for samples where the interest rates are considerably high, the log-
15
Lucas (2000) compares the two types of money demand and expresses a preference for the log-log form. In
contrast, Ireland (2009) manifests a preference for the semi-log form. The debate on the best choice of the form
for the money demand equation is then far from being settled.
19
5. An application to the CEE countries
The model described in the previous sections can be successfully applied to the CEE
countries, where an international currency (the euro) offers liquidity services to the domestic
agent. In the CEE countries it is well known that the foreign currency denominated deposits
to the total deposits ratio is high, and most of the foreign currency deposits are denominated
in euro.
Therefore, in order to test the two long-run money demand equations (equations (14)
and (19)), we use monthly statistics for the Czech Republic, Hungary, Poland and Romania,
from 1999M1 to 2015M11. The data are extracted from the International Monetary Fund
(IMF) – International Financial Statistics (IFS) database, Eurostat database, the OECD
database, and the statistics provided by the national central banks are used to complete the
series.
represents a proxy of the money demand structure from our model, as the structure of money
in circulation (cash) is unknown. For equation (19), the monetary aggregate M2 is used to
compute the real money demand, as the broad money is commonly used in previous
estimations regarding the money demand in the CEE countries. For robustness purpose, in
equation (19) we also employ the M1 aggregate to compute the real money demand. We use
the money market rate to calculate the discounted interest rate, the consumer price index and
Appendix.
16
Because our model relies on monetary consumption spending, we have retained household consumption
expenditure for the scale variable (equation (19)). This variable is available on a quarterly basis only, and we
have used a cubic spline function in order to get the monthly frequency. However, in line with previous
empirical estimations, we have also used the industrial production index on a monthly basis for the scale
variable. Using the Census X13 approach, both variables were seasonally adjusted before entering into
regression.
20
For the opportunity cost ϕ computation, we do not make the distinction between money
in circulation and deposits, and therefore the proportional cost is supposed to be the same for
all these forms of money substitutes. In addition, the value of ϕ is considered sufficiently
small to be mostly neglected (see previous studies). In this context, we have established for
corresponds to the loss rate of 1% advanced by Lucas and Nicolini (2015) for the narrow
We estimate the cointegration equations (14) and (19) by the DOLS method of
Saikkonen (1991) and Stock and Watson (1993), and by the FMOLS method of Phillips and
Hansen (1990), which both produce asymptotically unbiased estimators even in the absence
Both methods are consistent with the triangular representation of Phillips (1988, 1991)
of cointegrated vector I(1) processes. This representation is valid for any cointegrating rank,
but we assume that this cointegrating rank is 2. Consider a n-vector Yt y1t y 2 t Y3' t ' where
y1t and y 2 t are one-dimensional I(1) processes and Y3 t is a (n-2)-dimensional I(1) process,
Y3 t 3 Z3 t , (21)
17
For cash, it is likely that the value of ϕ increases to 2%. This is the estimate by Alvarez and Lippi (2009) of
the probability of cash theft in Italy for example. However, a value of 1% seems a priori more appropriate to
describe the loss of cash and costs associated with owning a bank account.
21
This triangular representation presents y1t and y 2 t as the “dependent” variables, but in
fact it does not require that they be the only endogenous variables, as the hypothesis of strong
representation supposes that there is only one “dependent” variable in each of the
cointegrating regression and that there is an equation for each “dependent” variable. The
equations (14) and (19) are consistent with this triangular representation with
y1t ln M t St M *t , y 2 t ln M t Pt and Y3t ln oct 1 ln oct 1 ln oc*t 1 ln X t Pt . The
'
DOLS and the FMOLS estimators of the parameters of system (20) are known to be
(Johansen, 1988) based on the vector error-correction model, and they deliver standard
statistics (such as the t-statistics and the Wald statistics) that are asymptotically normally
distributed.
However, before estimating the long-run relationship, we want to be sure that our series
are I(1). Therefore, in the first step we apply the classic ADF and PP unit root tests, allowing
a constant term in the model. The results are presented in Table 1 and show that our variables
are I(1).18 Therefore, we can proceed with the cointegration analysis, for both equations (14)
and (19).
For each equation we test the existence of a long-run relationship based on Hansen’s
(1992) instability test that relies on the L c statistic. The null hypothesis of this cointegration
test is the presence of cointegration. For the DOLS-type estimations, the number of leads and
lags is chosen using the Akaike information criteria, while for the FMOLS (with Bartlett
kernel) we have used a Newey-West automatic bandwidth rule. Different hypotheses on the
18
A small exception is represented by the log of the real industrial production in Romania.
22
Table 1. Unit root tests
Variables Tests Czech Rep. Hungary Poland Romania
Mt ADF -1.788*** -1.460*** -0.808*** -0.954***
ln
* PP -1.674*** -1.593*** -0.769*** -0.960***
St M t
ln oc t 1 ln oc*t 1 ADF -2.669** -1.785*** -1.371*** -2.020***
PP -2.661** -2.350*** -1.573*** -2.058***
M2 t ADF -0.012*** -1.189*** 1.065*** -1.795***
ln PP 0.048*** -1.294*** 0.752*** -0.602***
Pt
M1t ADF -1.370*** -1.365*** -0.408*** -1.650***
ln PP -1.453*** -1.162*** -0.048*** -0.685***
Pt
ln oc t 1 ADF -1.181*** -0.367*** -0.852*** -0.552***
PP -1.269*** -0.302*** -1.016*** -0.191***
XC t ADF -2.505*** -2.202*** -0.078*** -1.070***
ln PP -2.281*** -2.573*** -0.414*** -1.470***
Pt
XIPt ADF -1.604*** -1.396*** -0.463*** -7.620
ln PP -3.114* -2.023*** -1.605*** -8.420
Pt
Notes: (i) the null hypothesis is the presence of unit root and *, **, *** means a p-value for the t-statistic
>1%, >5% and >10% respectively. (ii) ln M1t and ln M2 t are the two forms of ln M t used in
P P P
t t t
equation (19), considering the monetary aggregate M2 and M1, while ln XC t and ln XIPt are the two
P P
t t
forms of ln X t used in equation (19), considering the household consumption expenditure (C) and
P
t
respectively the industrial production index (IP).
The results are presented for each equation (14 and 19) in Tables 2, 3 and 4 below.
First, looking at the results of equation (14), we notice that in almost all the cases the results
of the cointegration test differ between the DOLS and FMOLS estimations, the cointegration
being present for the DOLS estimation only. Hungary represents an exception, as the L c
statistic shows the existence of a long-run relationship for both the DOLS and FMOLS
estimations. Now, if we accept the hypothesis of cointegration, both methods show that the
elasticity of substitution is low, less than 1, and positive (except for Poland). In addition, the
Wald t-statistic shows that 1 1 . We see then that the monetary integration of CEE
countries with the euro area is reduced, as the value of elasticity between the currencies
23
substitution. However, this affirmation is true to a smaller extent for Romania, where a
Mt
ln
* Czech Rep. Hungary Poland Romania
St M t
DOLS FMOLS DOLS FMOLS DOLS FMOLS DOLS FMOLS
κ0 2.20*** 2.17*** 1.56*** 1.43*** 1.56*** 1.69*** 1.21*** 0.93***
κ1 -0.48*** -0.38** -0.27*** -0.15 0.37** 0.24 -0.52*** -0.32***
2
R 0.30 0.16 0.25 0.06 0.29 0.08 0.74 0.48
L c statistic 0.00 1.28 0.00 0.35 0.00 1.83 0.00 0.66
(>0.2) (0.00) (>0.2) (0.10) (>0.2) (0.00) (>0.2) (0.01)
Wald t-statistic 2.83 3.92 6.00 8.72 7.58 7.36 6.18 10.5
1 1 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Notes: (i) ***, **, * means significance at 1%, 5% et 10% significance level; (ii) p-value in brackets; (iii) κ 0
is the intercept of equation (14); κ 1 is the coefficient of ln oc t 1 ln oc*t 1 , with a negative sign.
We present further the results for our second cointegrating relationship (equation (19)).
Two sets of results can be observed in Table 3. A first set considers the real household
consumption as a scale variable, while a second set of results is based on the real industrial
Several conclusions can be drawn from these findings. First, when we look to the
coefficients’ sign and significance level, there is, in general, a strong correspondence between
the DOLS and FMOLS estimations. However, the cointegration relationship is validated in
all cases for the DOLS approach, while for the FMOLS estimation, the findings are more
19
According to the National Bank of Romania statistics (monthly bulletins), in Romania, the foreign currency
deposits to the total deposits ratio over the time-span 1999-2015 is over 50%, decreasing from 70% at the
beginning of the 2000s, to 34% at the present.
24
Table 3. Cointegration test and estimations for equation (19)
M2 t
ln Czech Rep. Hungary Poland Romania
Pt
Consumption DOLS FMOLS DOLS FMOLS DOLS FMOLS DOLS FMOLS
ω0 -7.68*** -2.30 1.92* -0.02 -7.92*** -5.65*** -1.44*** -3.80***
ω1 -0.07* -0.08** -0.23*** -0.19*** 0.21*** 0.04 -0.20*** -0.13***
ω2 0.90*** 0.23*** 0.02 0.10*** 0.01 0.05** 0.03* 0.17***
ω3 2.45*** 1.61*** 0.89*** 1.11*** 2.72*** 2.23*** 1.39*** 1.83***
R2 0.98 0.88 0.98 0.80 0.99 0.98 0.99 0.98
L c statistic 0.00 1.01 0.00 0.68 0.00 2.02 0.00 0.93
(>0.2) (0.02) (>0.2) (0.09) (>0.2) (0.00) (>0.2) (0.03)
Wald t-statistic 27.3 21.5 34.2 29.0 24.0 28.6 46.9 32.3
ω1 1 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Wald t-statistic 5.36 2.66 0.95 0.76 11.2 10.1 5.64 10.4
ω3 1 (0.00) (0.00) (0.34) (0.44) (0.00) (0.00) (0.00) (0.00)
Industrial
DOLS FMOLS DOLS FMOLS DOLS FMOLS DOLS FMOLS
production
ω0 9.38*** 8.65*** 6.60*** 9.60*** 8.75*** 7.56*** 2.98*** 4.48***
ω1 -0.08 -0.21*** -0.79*** -0.36*** 0.02 -0.19*** -0.97*** -0.53***
ω2 5.93*** 0.41*** 0.73*** -0.19 0.26*** 0.24*** -0.35*** 0.10
ω3 2.14*** 1.45*** 3.33*** 0.68 2.27*** 1.68*** -1.14*** -0.46*
R2 0.95 0.78 0.90 0.45 0.99 0.91 0.98 0.83
L c statistic 0.00 0.49 0.00 1.66 0.00 0.50 0.00 1.28
(>0.2) (>0.2) (>0.2) (0.00) (>0.2) (>0.2) (>0.2) (0.00)
Wald t-statistic 15.8 19.8 2.48 8.59 9.58 15.2 0.25 5.06
ω1 1 (0.00) (0.00) (0.01) (0.00) (0.00) (0.00) (0.80) (0.00)
Wald t-statistic 2.05 1.70 3.35 0.60 3.95 3.48 5.87 5.46
ω3 1 (0.04) (0.09) (0.00) (0.58) (0.00) (0.00) (0.00) (0.00)
Notes: (i) ***, **, * means significance at 1%, 5% et 10% significance level; (ii) p-value in brackets; (iii) ω 0
is the intercept of equation (19); (iv) ω1 is the coefficient of ln oc t 1 with a negative sign, ω 2 is the coefficient
of lnoc t 1 lnoc *t 1 and ω 3 is the coefficients of ln XC t or ln XIPt ; (v) M2 aggregate is considered.
P P
t t
Second, the coefficient ω1 has the expected sign and it is significant, except for Poland
when consumption is used as a scale variable. Its low value shows that the CEE domestic
currencies are complements not substitutes, except for Romania, when the industrial
production is used as a scale variable. The spread’s coefficient ω 2 also has the expected and
correct positive sign ( 1 ), in agreement with equation (19). Third, the coefficient ω 3 of
the scale variable is positive as expected, in all the cases for the household consumption, and
25
with one exception (Romania), in the case of the industrial production.20 However, the
consumption and the output elasticity are higher than 1, and we thus reject the hypothesis of a
unitary elasticity.
If we make a comparison between the CEE countries, we notice that the Hungarians
and the Romanians are more sensitive to the opportunity cost computed based on the internal
discounted interest rate, while the Czech agents are more sensitive to the opportunity cost
spread. In addition, the real output has a greater importance for Czech and Polish agents
compared with Hungarian and Romanian agents. All in all, the small elasticity of substitution
For checking the robustness of these findings we use the monetary aggregate M1
instead of M2 for computing the money demand in equation (19). Table 4 presents the
results. As in the previous case, there is a string correspondence between the DOLS and
FMOLS estimations. However, only for the DOLS approach, the Hansen’s (1992) instability
test shows the existence of a cointegrating relationship for all the countries, while in the case
of the FMOLS estimator the null of cointegration is rejected in 5 out of 8 cases. At the same
time ω1 has the expected sign and is very significant in all the cases, except for Romania,
when consumption is used as a scale variable. The spread of opportunity cost ( ω 2 ) positively
impacts the money demand as expected, with two exceptions however, which are Hungary
and Poland for the DOLS estimation when household consumption is used as a scale variable
(this negative result does not stand when industrial production is used in estimations). A
slight difference compared with the results reported for the M2 aggregate appears in the case
of the industrial production estimations, for Poland (FMOLS estimation) and for Romania
20
This result might be influenced by the fact that the real industrial production in the case of Romania has
proved not to be a I(1) process.
26
All in all, our empirical estimations fit the theoretical assumptions synthetized in
equations (14) and (19). The employed tests confirm, in general, the long-run relationship,
explaining the money demand in CEE countries. There is a large agreement between the
DOLS and the FMOLS estimators, although the cointegration relationship is documented in
particular for the DOLS approach. Consequently the money demand in CEE countries is
explained by the opportunity cost of holding the money (computed based on the discounted
money market rate), the spread of the opportunity cost, and the scale variable associated with
Table 4. Cointegration test and estimations for equation (19) – robustness check based on M1
M1t
ln Czech Rep. Hungary Poland Romania
Pt
Consumption DOLS FMOLS DOLS FMOLS DOLS FMOLS DOLS FMOLS
ω0 -12.0*** -11.2*** 2.69*** -0.60 -11.3*** -3.80*** -9.50*** -11.8***
ω1 -0.31*** -0.34*** -0.50*** -0.36*** -0.09** -0.13*** -0.06 0.00
ω2 0.43* 0.15*** -0.08*** 0.04 -0.10*** 0.17*** -0.00 0.12
ω3 2.68*** 2.55*** 0.61*** 1.01 2.92*** 1.83*** 2.82*** 3.26***
R2 0.99 0.96 0.98 0.87 0.99 0.98 0.99 0.96
L c statistic 0.00 0.95 0.00 0.55 0.00 0.93 0.00 0.55
(>0.2) (0.02) (>0.2) (0.17) (>0.2) (0.02) (>0.2) (0.17)
Wald t-statistic 13.3 15.9 26.3 22.7 19.4 32.3 7.36 12.5
ω1 1 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Wald t-statistic 4.75 7.05 4.14 0.07 12.3 10.4 3.40 9.49
ω3 1 (0.00) (0.00) (0.00) (0.93) (0.00) (0.00) (0.00) (0.00)
Industrial
DOLS FMOLS DOLS FMOLS DOLS FMOLS DOLS FMOLS
production
ω0 6.95*** 6.18*** 5.23*** 8.00*** 6.86*** 4.48*** -0.62 2.43**
ω1 -0.40*** -0.55*** -0.96*** -0.53*** -0.22*** -0.53*** -1.62*** -0.77***
ω2 0.77*** 0.42*** 0.53*** 0.17 0.17*** 0.10 -0.79*** -0.01
ω3 3.38*** 2.31*** 3.04*** 0.85* 2.73*** -0.46* -1.66** -0.57
R2 0.99 0.89 0.96 0.70 0.99 0.83 0.97 0.75
L c statistic 0.00 0.58 0.00 1.61 0.00 1.28 0.00 0.96
(>0.2) (0.15) (>0.2) (0.00) (>0.2) (0.00) (>0.2) (0.02)
Wald t-statistic 12.4 10.0 0.72 7.49 11.1 5.06 3.44 1.27
ω1 1 (0.00) (0.00) (0.47) (0.00) (0.00) (0.00) (0.00) (0.20)
Wald t-statistic 5.27 4.36 4.63 0.31 8.20 5.46 3.63 3.06
ω3 1 (0.00) (0.00) (0.00) (0.75) (0.00) (0.00) (0.00) (0.00)
Notes: (i) ***, **, * means significance at 1%, 5% et 10% significance level; (ii) p-value in brackets; (iii) ω 0
is the intercept of equation (19); (iv) ω1 is the coefficient of ln oc t 1 with a negative sign, ω 2 is the coefficient
of lnoc t 1 lnoc *t 1 and ω 3 is the coefficients of ln XC t or ln XIPt ; (v) M1 aggregate is considered.
P P
t t
27
6. Conclusions
We have investigated the money demand in CEE countries starting from a theoretical
model with microeconomic foundations, which allows a separation between the currency
substitution effect and the money demand sensitivity to exchange rate variations effect. A
particular feature of this model is related to the existence of a channel throughout the
exchange rate impacts on the money demand, even in the absence of a currency substitution.
This model can be successfully applied to CEE countries where the euro offers liquidity
services to the domestic representative agent, while the reverse is supposed not to be true.
The model parameterization shows that the money demand in CEE countries can be
result of our model. The empirical findings revealed by Hansen’s (1992) instability test on
the one hand, and the DOLS and FMOLS estimators on the other hand, underline the
existence of the two cointegrating relationships, where the real money demand in CEE
countries is explained by the opportunity cost of holding the money, as well as by the real
consumption or the real output. There is a large consensus between the DOLS and FMOLS
results, and the findings are robust regarding the use of M2 or M1 monetary aggregates for
assessing the money demand in equation (19). In general, the CEE countries’ representative
agents perceive the domestic currency as being more liquid compared to the euro, and a low
level of substitution is obtained between domestic currencies and the euro. If previous
empirical studies on CEE countries’ money demand test the money demand sensitivity to
international factors and report, in general, a high substitution level, then our micro-founded
model shows that the substitution level is reduced, and that the CEE currencies and the euro
are complements rather than substitutes. Therefore, the results of previous work that
advanced an increased degree of CEE monetary integration with the euro area should be
considered with caution, because these studies do not consider the influence of exchange rate
28
variations on money demand in the absence of the currency substitution, nor do they
consider, as possible, the existence of a complementary effect between the CEE currencies
The empirical results we obtain do not represent, however, an incontestable proof of the
existence of two long-run relationships, as the results are mitigated for the FMOLS estimator.
Nevertheless, the empirical findings can be considered as being robust and they clearly show
that the monetary integration of CEE countries with the euro area appears implausible for the
moment. In this context, the policy implications of our study are twofold. Indeed, we show
that the monetary authorities in the CEE countries should consider, in the money demand
estimation, not only the role of opportunity cost of holding the money, the consumption and
the output, but also the impact of the exchange rate, which manifests even in the absence of a
strong currency substitution. At the same time, we posit that these countries are not yet
prepared to join the euro area, and additional efforts must be made to increase their monetary
integration.
References
Alvarez F. and F. Lippi (2009), Financial innovation and the transactions demand for cash,
Baumol W.J. (1952), The transactions demand for cash: An inventory theoretic approach,
Bordo, M.D. and Choudhri, E.U. (1982), Currency Substitution and the Demand for Money:
Some Evidence for Canada, Journal of Money, Credit and Banking, 14, 48-57.
Cagan P. (1956), The monetary dynamics of hyperinflation, in Studies in the Quantity Theory
29
Chen C-N. (1973), Diversified currency holdings and flexible exchange rates, Quarterly
Cuddington J.T. (1983), Currency substitution, capital mobility and money demand, Journal
Dreger C., H-E. Reimers and B. Roffia (2007), Long-run money demand in the new EU
member states with exchange rate effects, Eastern European Economics, 45, 75-94.
Elbourne A. and J. de Haan (2006), Financial structure and monetary policy transmission in
Fidrmuc J. (2009), Money demand and disinflation in selected CEECs during the accession to
Filosa R. (1995), Money demand stability and currency substitution in six European countries
Guidotti P.E. (1989), Exchange rate determination, interest rates, and an integrative approach
to the demand for money, Journal of International Money and Finance, 8, 29-45.
Hansen, B.E. (1992), Tests for Parameter Instability in Regressions with I(1) Processes,
Hsieh W-J. and Y. Hsing (2009), Tests of currency substitution, capital mobility and
959-964.
Hueng C.J. (1998), The demand for money in an open economy: Some evidence for Canada,
Ireland P.N. (2009), On the welfare cost of inflation and the recent behavior of money
30
Johansen S. (1988), Statistical analysis of cointegrated vectors, Journal of Economic
Leventakis J.A. (1993), Modelling money demand in open economies over the modern
Lucas, R. (1976), Econometric Policy Evaluation: A Critique. in The Phillips Curve and
Lucas, R. (1982), Interest Rates and Currency Prices in a Two Country World, Journal of
Lucas R.E. Jr. (2000), Inflation and welfare, Econometrica, 68, 247-274.
Lucas R.E. Jr. and J.P. Nicolini (2015), On the stability of money demand, Journal of
Meltzer A.H. (1963), The demand for money: The evidence from the time series, Journal of
McKinnon R.I. (1982), Currency substitution and instability in the world dollar standard,
31
Stock J.H. and M.W. Watson (1993), A simple estimator of cointegrating vectors in higher
88, 673-698.
Tobin J. (1956), The interest elasticity of the transactions demand for cash, Review of
32