Journal of International Money and Finance 25 (2006) 330e347
www.elsevier.com/locate/econbase
Money, inflation and output in Romania,
1992e2000
Nina Budina a, Wojciech Maliszewski b, Georges de Menil c,*,
Geomina Turlea d
a
World Bank, Washington DC, USA
London School of Economics, UK
c
Ecole des Haute Etudes en Sciences Sociales and Stern School, New York University, USA
d
Romanian Centre for Economic Policy and Institute for World Economy, Bucharest, Romania
b
Abstract
Money, inflation and output are tested for stationarity, and found to be integrated of order one. We apply the Johansen procedure for cointegration to test for the rank of the matrix of cointegration relations
(one), to test for the weak exogeneity of output (accepted), inflation (rejected) and money (rejected).
We interpret the unique cointegrating relationship as an extended Cagan money demand function. We
then estimate error correction mechanisms, which explain the short-run movements of real money and inflation. The evidence suggests that in the period considered, including the sub-sample between the liberalization shocks, inflation was largely a monetary phenomenon.
Ó 2005 Elsevier Ltd. All rights reserved.
JEL classification: E52; E41; E31
Keywords: Demand for money; Cointegration; Inflation; Transition
1. Introduction
The end of Central Planning in Europe e whether it came as a result of slow decay or of
rapid collapse e was frequently accompanied by bursts of high inflation. Though the causes
* Corresponding author. Delta, 48, Bd. Jourdan e E.N.S, 75014 Paris, France. Tel.: þ33 1 43 13 63 32; fax: þ33 1 43
13 63 10.
E-mail address: nydemenil@aol.com (G. de Menil).
0261-5606/$ - see front matter Ó 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jimonfin.2005.11.006
N. Budina et al. / Journal of International Money and Finance 25 (2006) 330e347
331
of that inflation have been and remain a subject of debate, statistical tests of alternative hypotheses have not been frequent.1 One of the central questions is whether these early periods of
high inflation were caused by excessive monetary creation or other factors. This paper addresses that question in the case of Romania, a weakly reformed, high inflation, transition
economy.
High inflation cannot persist unless it is caused or validated by excessive money creation.
The objective of this paper is to provide an empirical estimate of the relationship between inflation and money creation in Romania, and to explore statistically the extent to which money
creation is cause or effect.
Romania is an interesting case in its own right, but also because, in that country, liberalization was delayed and drawn out, and, therefore, the dynamics of high inflation can be
observed over a period of several years. We seek first to specify and estimate a demand function for real money in Romania. This is done with monthly data from January 1992, the earliest date for which all of the necessary series are available, to December 2000, when the rate
of inflation first durably declined below 35% per year at an annual rate. Augmented ADF
tests suggest strongly that the principal variables are integrated of order one. In that context,
the search for a real money demand function is a search for a cointegrating, equilibrium relationship; and an important question is whether, and to what degree, the observed high inflation is influenced by divergence between actual money supply and equilibrium money
demand.
The estimation of equilibrium real money demand functions has been one of the central focuses of the empirical literature on applied cointegration analysis. Ever since the path breaking
studies of Johansen and Juselius (1990) and Juselius (1992), cointegrating real money demand
equations have been identified and estimated in many different countries and contexts. Lütkepohl and Wolters (1998a,b) edit the proceedings of a workshop motivated by the run-up to European Monetary Union, in which cointegration estimates of money demand are presented for
the United Kingdom, Finland, Norway, Germany, Spain, Greece and Switzerland.2 Ericsson
et al. (1998) discuss the structural analysis of cointegrating relationships in their introduction
to a series of studies in which this method is applied to money demand in the United Kingdom,
Denmark, Turkey, Brazil and Australia.3
Though the general principles of money demand analysis apply to all non-centrally planned
economies, European transition economies present certain characteristic traits, which call for
a distinctive approach. These are high inflation; large amounts of directed, non-market credits;
periods when the exchange rate is significantly overvalued; and pervasive price controls. We
argue that these call for modelling money demand with an extended version of the Cagan
(1956) specification.
Single equation cointegration methods have been used to estimate long run, Cagan money
demand functions for transition economies by Aarle and Budina (1996). Measures of the difference between money supply and such money demand functions have been used as part of
1
One of the obstacles to statistical inference is the short duration of the initial burst of high inflation in most transition
countries.
2
See Hendry and Mizon (1998), Ripatti (1998), Eitrheim (1998), Scharnagl (1998), Lütkepohl and Wolters (1998a,b),
Vega (1998), Ericsson (1998), Ericsson and Sharma (1998), Peytrignet and Stahel (1998), and Juselius (1998a,b).
3
The studies, in a special section of the Journal of Business and Economic Statistics, are Harbo et al. (1998), Juselius
(1998a,b), Metin (1998), Durevall (1998) and De Brouwer and Ericsson (1998). See also, earlier work by Doornik and
Hendry (1994).
332
N. Budina et al. / Journal of International Money and Finance 25 (2006) 330e347
an error correction mechanism to explain inflation in Russia by Choudhry (1998).4 This paper
can be viewed as an extension to another incompletely reformed economy of the approach of
Choudhry.5
Our finding that we can reject the weak exogeneity of inflation, and that excess money supply affects the course of inflation confirms Choudhry’s result. Our specification and our conclusions differ from those reached in a related analysis in IMF (2001).
Section 2 reviews the context of transition in Romania, presents the data, and discusses the
distinguishing characteristics of money demand and inflation in European transition countries.
Section 3 tests for the stationarity of the variables studied. Section 4 uses the Johansen (1988)
technique to identify and estimate a cointegrating relationship and test for weak exogeneity.
Section 5 estimates the implied error correction models and uses them to compute out-of-sample forecasts. The economic implications of these results are interpreted in Section 6, which is
followed by a conclusion.
2. Romania: a decade of delayed liberalization and high inflation
2.1. Political transformations6
Under Ceausescu, Romania experienced one of the most draconian and repressive regimes
of Communist Europe. After he was overthrown, in December 1989, power remained in the
hands of the successor to the Communist Party, which eventually became the PDSR (Partidul
Democrat Social Roman).
The PDSR introduced partial political and economic reforms, but resisted full liberalization.
On at least two occasions, steps toward liberalization were either thwarted, or offset by increased distortions elsewhere. The unification of exchange rates in November 1991 was accompanied by a large devaluation, but the authorities stopped short of either adopting external
convertibility, or freeing domestic prices. The exchange rate remained controlled, and was
soon again overvalued. It remained overvalued until early 1997.
In June 1993, the share of consumer prices subject to controls was reduced from over 50% to
about 40%, and, in October of the same year, real interest rates on credits from the state-owned
banking system (which had been as low as 230%) were increased to positive levels. By the
spring of the following year, massive directed credits to state-owned enterprises and political
affiliates had resumed. Between 1994 and 1996, the economy grew, but the distortions resulting
from subsidized and uneconomical credits, the overvalued exchange rate, and price controls accumulated. Between 1991 and 1996, the IMF had signed, and then interrupted three Stand-By
Agreements, for lack of compliance with commitments to convertibility and price
liberalization.7
In elections in November and December 1996, opposition forces led by a coalition of historical parties won a majority of the National Assembly, and defeated the incumbent president.
4
See also the study of inflation in Ukraine by Banaian et al. (1998).
The special characteristics of the Cagan money demand function in Romania have been studied by Dobrescu (1994).
Cointegration techniques are used to estimate a complete model of the monetary mechanism in Pelinescu and Scutaru
(2000).
6
The political economy of Romania’s first 10 years of transition are discussed in De Menil (2003).
7
Not withstanding the above, trade was substantially liberalized, notably after the signature of an Association Agreement with the EU in February 1993, and exports and imports were reoriented towards the West. Also, small and medium
sized enterprises were privatized, and 10 foreign banks (six of them branches) were licensed between 1990 and 1996.
5
N. Budina et al. / Journal of International Money and Finance 25 (2006) 330e347
333
The new Government began its term with a comprehensive program of liberalization and stabilization. In February and March 1997, with the support of the IMF, it devalued the exchange
rate, guaranteed its convertibility, and eliminated remaining price controls on all but 10% of
consumer goods.8 Energy and public utility prices were raised to international levels. The burst
of inflation, which immediately resulted, was to be contained by strict budgetary and monetary
policies. Implicit subsidies were eliminated, and the explicit, consolidated, cash budget deficit
was limited to 5%. Subsidized, directed credits were stopped, but political pressures delayed the
closure of major state banks, which had been rendered insolvent by their previous bad loans.
For two years, these banks were bailed out by special credits from the central bank, which eventually found their way into increased money supply. Inflation remained high. Output fell as a result of domestic uncertainty and adverse external shocks.
Four years later, in the elections of November and December 2000, the PDSR e renamed the
PSD (Partidul Social Democrat) e won back the majority of the legislature and the presidency.
Though it temporarily halted privatisation and social security reform, it maintained most of the
other major reforms enacted between 1997 and 2000.
2.2. Economic evolution
Figs. 1 and 2 describe the evolution of inflation, output and money during the period of this
study. Prices are measured by the consumer price index e P; output by the real industrial production index e Y9; and money by domestic M2 e M.10 The data are monthly, seasonally adjusted, and they span January 1992eDecember 2000. We eliminate common nominal trends, by
focusing in the usual manner on the rate of change of prices (Dp), the natural logarithm of real
industrial production ( y), and the natural logarithm of real money balances (m p), where lower
case letters represent the logarithms of upper case letters.
Fig. 1 shows inflation starting at a high level. The first observations of Dp in January and
February 1992 are 18% and 12%. These come after a year of similarly high inflation. The average monthly rate during 1991 (not shown in the figure) was 10%; the more customary January
1991eJanuary 1992 measure was 236%. The average monthly rate remains high in 1992 (9%)
and 1993 (11%). 1993 was marked by an episode of partial liberalization in May 1993, when
consumer prices rose by 27%. The average monthly rate declines in 1994 (4%) and 1995 (2%),
but starts to rise again after July 1996. Liberalization at the beginning of 1997 leads to
8
The IMF supported the new Government’s program with a Stand-By Credit, and the World Bank with Sectoral Adjustment Loans. The IMF loan was interrupted when a major Government reshuffle was accompanied by delays in the
program in 1998.
9
We use real industrial production, because it is the broadest activity variable available on a monthly basis. The velocity of M relative to the nominal value of reported industrial production will also be affected by the evolution of the
share of informal to total output. Since informal output is not directly measurable, we consider that the effect of changes
in its share will be captured in the coefficients of the other factors influencing reported velocity. This approach is also
taken by Dobrescu (1994, 1998).
10
In the studies of money demand in Romania cited in the previous footnote, Dobrescu advocates focusing on total
M2, inclusive of foreign currency deposits, to which he adds estimates of economy-wide net arrears, and estimates of
the value in Lei of foreign currency held outside the banking system. We exclude foreign currency, both because domestic money is a less costly medium of exchange (as opposed to a store of value) in a less than hyperinflationary context, and because no direct measurements are available for foreign currency in circulation. We also prefer to treat arrears
as a form of credit, rather than an alternative medium of exchange. In the estimates reported below, inflation and other
explanatory variables capture the effect on M-velocity of variations in the intensity of the use of both foreign currency
and arrears.
334
N. Budina et al. / Journal of International Money and Finance 25 (2006) 330e347
0.30
-3.00
0.25
-3.50
0.20
-4.00
0.15
0.10
-4.50
0.05
-5.00
inflation rate (left scale)
Jan-00
Jan-99
Jan-98
Jan-97
Jan-96
Jan-95
Jan-94
Jan-93
-0.05
Jan-92
0.00
-5.50
velocity (right scale)
Fig. 1. Romania: trends in velocity of domestic broad money and inflation.
a renewed bout of high inflation, 13% in January 1997, 17% in February, and 27% in March.
The monthly rate then declines slowly. It averages 2% in 1999 and 4% in 2000. At no point in
the period do price increases reach the level of hyperinflation (defined as 50% per month), but
they are, throughout the period, well above the norm in Western Europe or in the US, and remain, at the end of the period, the highest in the region.
In January 1992, Romania was in the middle of its first transition recession, and industrial
production was still declining (Fig. 2). However, output stabilizes in 1993, and grows steadily
in 1994, 1995 and 1996. Between December 1993 and December 1996, the average annual rate
of growth of industrial production is 13%. The OECD has described this as a period of real
growth with weak structural change.11 The second transition recession, beginning in early
1997, leads to a renewed decline of industrial production. Output stabilizes at the end of
1999, and is growing slowly during the last year of our period.
The wide swings which characterize the movement of real money balances between 1992
and 2000 (Fig. 2) are suggestive of a positive relationship with real output and negative one
with inflation. Between January 1992 and January 1994, real domestic M2 declines 69%. Between January 1994 and November 1996, it rises 139%. During the major liberalization episode
of 1997, real M2 falls 41% in four months, and remains low for most of the rest of the period.
Fig. 1 presents the underlying relationship, central to this study, between velocity (m p y)
and inflation.
Fig. 3 represents two additional variables, whose movements are important characteristics of
the inflation process in Romania: the monthly rate of change of the real Leu/dollar exchange
rate (D(e þ p p*)), and the monthly rate of change of administered prices (Dpadm).12
Before 1997, the Leu was not freely convertible, except for brief intervals. Between episodes
of devaluation, the authorities adjusted the official nominal exchange rate in a manner which
resulted in a moderately rising trend of its real value.13 In January and February 1997, the
11
See OECD (1998) and (2002).
D(e p p*) is the monthly rate of change of the official Leu/dollar exchange rate, minus the difference between
the monthly rate of change of the US CPI and the Romanian CPI.
13
Data on the black market premium suggest that the Leu was overvalued throughout this period. See De Menil
(2003).
12
N. Budina et al. / Journal of International Money and Finance 25 (2006) 330e347
1.20
4.80
1.00
4.70
0.80
335
4.60
0.60
4.50
0.40
4.40
0.20
Jan-00
Jan-99
Jan-98
Jan-97
Jan-96
Jan-95
4.20
Jan-94
-0.20
Jan-93
4.30
Jan-92
0.00
real money balances (left scale)
index of industrial output (right scale)
Fig. 2. Romania: trends in real money balances and output. Source: National Bank of Romania and authors’
calculations.
Leu was allowed to float, and it depreciated sharply. At the end of February, the currency became convertible on current account, but not on capital account. The National Bank of Romania
intervened regularly, using reserves to smooth its movements. In the spring of 1999, when
a combination of a peak in foreign debt service and delays in the renegotiations of an IMF
Stand-By Agreement generated the expectation of a high probability of external default, the
bank again let the real value of the Leu slide. Otherwise, the period after February 1997 was
characterized by a moderate appreciation of the real rate of exchange with the US dollar. Statistically the dynamics of real appreciation/depreciation appear similar before and after the
1997 liberalization.
The persistence of price controls over large segments of consumer goods implied that the
economy was subjected to shocks at irregular intervals, when the authorities raised the prices
still subject to administration.14 Fig. 3 depicts the impact effect of these increases. The two
largest shocks coincided with the liberalizations of May 1993 and February and March
1997. In those months, prices still under control rose, respectively, 57%, 24% and 44%. In
the figure, as in our estimations, the increases in administered prices are weighted by the remaining share of controlled prices in the consumer price index. Since this share declined at
each liberalization, the peaks of the weighted impact number in the figure also decline. The
construction of the administered price index, Padm, and its weight, are described in the annex.
Figs. 1e3 do not include any interest rate measures. Until 1997, many of the largest state
banks, which dominated the Romanian banking sector, were heavily engaged in extending
directed credits at subsidized rates. The official measure of average deposit rates implies
negative real rates through all of 1993, most of 1995, and all of 1996. After the liberalization
of 1997, these real interest rates fluctuate around zero. At the same time, the Treasury Bill
market, which was almost non-existent before 1997, remained essentially the narrow preserve of these same banks after 1997. Given the limited and political nature of financial
14
Practitioners in these economies often go so far as to argue that administrative price increases completely determine
the overall course of inflation, while they are working their way through the economy. Our estimates below will support
the view that they alter the short-run dynamics of what is, in the long run, fundamentally a monetary phenomenon.
N. Budina et al. / Journal of International Money and Finance 25 (2006) 330e347
0.10
-0.10
0.05
-0.20
0.00
-0.30
Jan-00
0.00
Jan-99
0.15
Jan-98
0.10
Jan-97
0.20
Jan-96
0.20
Jan-95
0.25
Jan-94
0.30
Jan-93
0.30
Jan-92
336
growth in administrative prices (left scale)
real exchange rate change (right scale)
Fig. 3. Romania: growth in administrative prices and real exchange rate changes. Source: National Bank of Romania
and authors’ calculations.
alternatives to holding money, we have not included any interest rates in our estimates of the
demand for money.15
2.3. Modelling money demand in transition
High rates of inflation and poorly developed financial markets have characterized European
transition countries in their early years. Romania lived with both for longer than most. Under
both conditions, the rate of inflation is a better measure of the opportunity cost of holding
money than available interest rates. This points to the following extended Cagan (1956)
form for the choice of a model of money demand:
m p ¼ b0 þ b1 ðpe p1 Þ þ b2 y
ð1Þ
As before, lower case letters are the natural logs of upper case letters. The second term is the
expected inflation rate. We follow Aarle and Budina (1996), Banaian et al. (1998), Choudhry
(1998) and others in basing our analysis of money demand in transition on this simple form.
Choudhry (1998) adds the rate of depreciation of the nominal exchange rate to Eq. (1) in the
case of Russia. If the rate of change of the Leu/US dollar rate had diverged in an irregular manner from the domestic inflation rate for significant periods of time we could similarly have
added e e1 to Eq. (1) (where e is the logarithm of that exchange rate). However, was we
point out above, in Romania, controls before 1997 and the managed float after 1997 resulted
in relative constancy (with the exceptions mentioned) of the real exchange rate. Therefore,
e e1 does not have sufficiently independent variation to be an additional determinant of
the long-run real demand for money. We test below for the transitory influence of variations
in the real exchange rate.
15
In an earlier version, we experimented with including the real bank lending rate when positive and greater than 5%.
However, this measure was non-zero in such a limited number of months that it acted very much like a dummy variable.
When the sample was extended for the present version, the variable was dropped.
N. Budina et al. / Journal of International Money and Finance 25 (2006) 330e347
337
One of the paradoxes of money demand and the inflation process in less than fully liberalized transition economies is that neither the prevalence of price controls nor their removal appears fundamentally to distort the form of the demand for money. High rates of inflation have
been observed even when countries had pervasive price controls.16 It will be shown below that
Romania is not an exception to this paradox.
The liberalization of February and March 1997 merits particular attention. Did the freeing of
most remaining controlled prices, the passage from an overvalued to a mostly market determined exchange rate, and the termination of years of massive directed credits, constitute
such important structural changes that they led to a regime change in the monetary mechanism
and the inflationary process? In what follows we will test for a structural break in long-run
money demand, and in price dynamics. The answer, reported in Section 5, is that, though
our equations include short-run responses to these shocks, there does not seem to be evidence
of a structural break.
3. Choice of variables and tests for stationarity
We test for the stochastic character of the seasonality in our series using two different methods: the Dickey et al. (1984) method modified by Osborn (1988) and the method developed by
Franses (1991) as an extension of Hylleberg et al. (1990) test for monthly data. Both tests reject
in all cases the null hypothesis of stochastic seasonality. Therefore the use of D12 filter would be
inappropriate. The rejection of non-stationary, stochastic seasonality justifies our maintained,
simplifying assumption that seasonality is deterministic. We use seasonally adjusted data17
for all variables except padm.
Table 1 below presents augmented DickeyeFuller (ADF) unit root tests for all the variables. Following the strategy proposed by Dickey and Pantula (1987), we start the sequence of
tests from the null hypothesis of two unit roots in the series. The lag structure in ADF equations is determined by sequential reduction starting from six lags based on F-statistics,
Schwarz information criterion and tests for remaining autocorrelation in the residuals. The
tests are conducted with constant and trend for the null hypothesis of a single unit root,
and with constant only for testing the hypothesis of two unit roots. The estimation period
runs from January 1992 to December 2000. The tests suggest that the log of real money
(m p), inflation Dp and the log of real industrial production y are all integrated of order
one. The rate of change of the real exchange rate D(e p þ p*) appears to be stationary.
16
In Ukraine in 1993, for instance, extensive controls did not stop price increases from reaching hyperinflationary levels. See De Menil (1997). Some of the controls were administered rates of profit margin, and others e in absolute level
form e were rapidly adjusted, increasingly frequently, by the bureaucracy.
17
In order to remove the deterministic seasonality present in our series we used the X-12-ARIMA program of Census
Bureau of the US Department of Commerce. Ericsson et al. (1994) show that the linear seasonal filter approximating
earlier version of the Census procedure (X-11) may affect inferences about cointegration and alter dynamics and exogeneity status of variables in the model. Siklos et al. (1996) show that the X-11 procedure exhibits strong non-linear
characteristics, and that these characteristics may lead to incorrect inference about the number of cointegrating vectors
in series simulated from linear models. However, empirical series may exhibit certain non-linear features and the impact
of X-11 adjustment on more complicated process has not yet been fully researched. If the periodic component of datagenerating-process exhibit non-linear characteristics, and if seasonality is treated as a noise contaminating the true relationship between the series, application of the X-11 procedure may perform better than using seasonally unadjusted
data. We believe that this is the case for transition data, where seasonal patterns are likely to be more complicated and
volatile than in developed economies.
338
N. Budina et al. / Journal of International Money and Finance 25 (2006) 330e347
Table 1
ADF unit root tests on the SA data
Variable
Two roots
One root
Constant and trend
Constant
Long sample: 1.1992e12.2000
mp
Dp
y
Dpadm
D(e p þ p*)
3.6694**
6.8111**
4.9507**
9.3518**
8.7797**
(4)
(4)
(4)
(4)
(4)
2.6317
2.8751
1.5757
3.5692*
6.7454**
(4)
(4)
(4)
(4)
(4)
2.8203
2.5050
1.5772
2.5974
6.7563**
(4)
(4)
(4)
(4)
(4)
Short sample: 6.1993e11.1996
mp
Dp
y
Dpadm
D(e p þ p*)
3.7527**
5.0742**
3.6099**
12.690**
4.6040**
(1)
(3)
(4)
(4)
(4)
2.5485
1.0994
3.2600
5.7587**
3.9457*
(4)
(4)
(4)
(4)
(4)
1.0021
3.0908*
0.11889
8.1583**
4.2036**
(4)
(4)
(4)
(4)
(4)
**Denotes rejection at 1% critical value; *denotes rejection at 5% critical value; (.) number of lags chosen for testing
unit roots.
The evidence is borderline for the rate of change of administered prices Dpadm, but we take it
also to be stationary.
As a further test of our hypotheses, we choose to estimate the cointegration and error correction relations discussed below twice, once for the full-sample period, and a second time for
a shorter sample period (June 1993eNovember 1996), beginning after the first and ending before the second liberalization shock. The ADF tests for the shorter sample period are reported in
Table 1. They provide additional support for the inferences based on Table 1. In the shorter
sample period, there is strong rejection of a unit root for the rate of change of administered
prices.
4. Johansen’s tests for cointegration and weak exogeneity
We begin our analysis of the dynamics of money, price and output with the formulation of an
unrestricted VAR system with three endogenous variables: (m p), Dp and y. The multivariate
approach is more efficient than single equation modelling if there is a failure of weak exogeneity18
or if variables under investigation form more than one cointegrating relation. Since both of
these conditions may be present in our data, we apply the Johansen (1988) procedure based on
the system estimation to determine the number of cointegration vectors and to test for weak
exogeneity.
The analysis starts from the unrestricted VAR system (UVAR) of the following form:
xt ¼
k
X
Aj xtj þ gDt þ 3t
ð2Þ
j¼1
18
In the cointegration framework the variable is weakly exogenous if it is not influenced by deviations from the longrun relationships. The concept of weak exogeneity is discussed in Engle et al. (1983). Weak exogeneity in the cointegration framework is discussed in Ericsson et al. (1998).
N. Budina et al. / Journal of International Money and Finance 25 (2006) 330e347
339
where xt contains all endogenous variables (m p, Dp, y) and Dt contains exogenous variables.
Finally, 3t is a vector of normally distributed error terms. The system can also be presented in
the equivalent VECM (vector error correction) form:
Dxt ¼
k1
X
Pj Dxtj þ Pxt1 þ gDt þ 3t
ð3Þ
j¼1
P
P
where P ¼ ð ki¼1 Ai IÞ and Pj ¼ ð ki¼jþ1 Ai Þ for j ¼ 1,2,.(k 1). The P matrix can be
decomposed to ab#, where a represents the speed of adjustment to equilibrium and b# the longrun coefficients. The rank of the P matrix, which is equivalent to the number of cointegrating
vectors, is tested by the Johansen (1988) procedure.
The first step in empirical UVAR modelling is to set the appropriate lag structure and select
the set of exogenous variables on which we are conditioning the system.
We experimented with different sets of exogenous variables. Our preferred set includes a constant term, an impulse dummy for January 1997 (more flexible exchange rate regime), an impulse dummy for March 1997 (price liberalization), the current second difference of the log
of the administered price index scaled by the weight of administered prices in CPI, and lags
of the first difference of the real Leu/dollar exchange rate. The two 1997 dummy variables reflect
general characteristics of the liberalization of that winter, not captured by the other two exogenous variables. We also ran the same specification with, in addition, a time trend.19,20
We treat both the rate of change of administered prices and the rate of change of the real
exchange rate as measures of exogenous, supply shocks, expected to have an impact effect
on the overall price level, and therefore on the supply of real money.
We start the analysis with six lags of the endogenous variables, the two dummies mentioned
above, the current, scaled, second difference of the log of administered prices, and six lags of
the first difference of the real exchange rate. After sequential reduction, UVAR with four lags of
each endogenous variable, four lags of the first difference of the real exchange rate, the two
dummies and our measure of administered prices is chosen for further analysis,21 and the Johansen’s (1988) trace (ltrace) and maximum eigenvalue (lmax) tests are used to determine the
rank of P matrix. Asymptotic critical values for these tests are given in Johansen (1988) and
Osterwald-Lenum (1992). Reimers (1992) proposed small sample correction by replacing T
by T-nk in both tests, and results from this test are also reported. It is important to note that
the critical values of these tests are only indicative if we condition on exogenous variables other
than constant and seasonal dummies. Therefore, in order to draw more decisive conclusions on
the number of cointegrating vectors, we also report test results from the specification without
exogenous variables in the system.22
19
The results were similar to those described below, except that the estimated semi-elasticity of inflation in the longrun demand was more than half again higher. With the trend present, acceptance of the weak exogeneity of output, and
rejection of the weak exogeneity of inflation and real money were even clearer than they are below. We prefer to emphasize the results without the trend, because of the more economically reasonable estimated value of the semi-elasticity of
inflation.
20
Two of the variables which were included in our original set of exogenous variables, but which we did not retain
because they were not significant were Dw and D(w p), where w is the average wage rate in industry.
21
The final lag structure is determined by VAR order selection criteria SC (Schwarz’s information criterion) and AIC
(Akaike’s information criterion), together with F-statistics associated with sequential reduction. The results on specification search are available from authors on request.
22
Harbo et al. (1998) and MacKinnon et al. (1999) account fully for the inclusion of exogenous variables. The strength
of our results indicates that they are robust to adjustments in the critical values.
340
N. Budina et al. / Journal of International Money and Finance 25 (2006) 330e347
Table 2
Cointegration statistics
Rank
lmax
lmax using T-nk
95%
ltrace
ltrace using T-nk
95%
Model 1: UVAR with four lags January 1992eDecember 2000, conditional model
0
28.68**
25.49*
22.0
44.33**
39.4*
1
11.47
10.19
15.7
15.65
13.91
2
4.18
3.72
9.2
4.18
3.72
34.9
20.0
9.2
Model 2: UVAR with four lags January 1992eDecember 2000, no exogenous variables
0
30.57**
27.18**
22.0
44.4**
39.46**
1
9.80
8.17
15.7
13.82
12.29
2
4.02
3.58
9.2
4.02
3.58
34.9
20.0
9.2
Model 3: UVAR with four lags June 1993eNovember 1996, conditional model
0
25.95*
18.54
22.0
41.76**
1
9.25
6.61
15.7
15.81
2
6.56
4.69
9.2
6.56
34.9
20.0
9.2
29.83
11.29
4.69
**Denotes rejection at 1% critical value; *denotes rejection at 5% critical value.
The results in Table 2 suggest that there is only one linear combination of the variables,
which is stationary. After imposing this restriction and normalising the parameter of the log
of real money to unity, the cointegrating vector is recognizable as an augmented Cagan
(1956) money demand function.23
In the next step we test for the weak exogeneity of industrial production, inflation and real
money in the conditional system. Since the hypotheses being tested are linear in I(0) variables,24 the test statistic used is the likelihood ratio test with limiting c2 distribution and a number of degrees of freedom equal to the number of independent restrictions to be tested. Tests for
restrictions and estimates of the coefficients with imposed restrictions are reported in Table 3.
Weak exogeneity of industrial production cannot be rejected in any configuration. Economically, this implies that there is no long-run feedback from excess real money to industrial production. We propose an interpretation of this result in Section 6. In contrast, weak exogeneity of
both real money and the rate of inflation are decisively rejected at the 1% level. This rejection
constitutes the underlying rationale for the estimation of error correction mechanisms for the
short-run dynamics of real money and inflation.
The identified equilibrium money demand function (using conditional UVAR with imposed
weak exogeneity of production) has the following form (with standard errors shown under the
estimated coefficients):
m p ¼ 8:865 13:696 Dp þ 2:225 y
ð3:418Þ
ð3:555Þ
ð0:779Þ
ð4Þ
We check the robustness of these results by repeating the tests on a shorter sample, which
excludes the two major price liberalization episodes of 1993 and 1997. In the longer period
tests, it could be argued that the estimated sensitivity of inflation to excess real money is mainly
a reflection of spurts of pent-up inflation released at the time of liberalization. The results for
23
We do not further impose unitary elasticity of real money demand to output, both because there are institutional
reasons to expect velocity to increase, other things being equal, during the period, and because the only available monthly
measure of output is industrial production, and not GDP.
24
xt is I(1) but Dxt and b#xt are I(0) because there is cointegration between elements of xt.
N. Budina et al. / Journal of International Money and Finance 25 (2006) 330e347
341
Table 3
Tests for restrictions in the conditional model
Weak exogeneity tests (cointegrating rank ¼ 1 imposed)
Variable
Jan 1992eDecember 2000
2
mp
Dp
y
c (1) ¼ 17.061 [0.0000]**
c2(1) ¼ 6.6331 [0.0100]**
c2(1) ¼ 1.2261 [0.2682]
June 1993eNovember 1996
c2(2) ¼ 16.662 [0.0000]**
c2(2) ¼ 6.5678 [0.0104]*
c2(2) ¼ 0.39088 [0.5318]
Long-run coefficients b# with weak exogeneity of LIIP imposed (standard errors in parentheses)
mp
Dp
y
Constant
Jan 1992eDecember 2000
1.000
e
13.696
(3.555)
2.225
(0.779)
8.865
(3.418)
June 1993eNovember 1996
1.000
16.069
e
(4.076)
1.587
(0.803)
5.818
(3.746)
Adjustment coefficients a with weak exogeneity of LIIP imposed (standard errors in parentheses)
mp
Dp
y
Jan 1992eDecember 2000
June 1993eNovember 1996
0.05427 (0.0106)
0.01522 (0.0053)
e
0.1274 (0.0279)
0.04797 (0.0194)
e
**,*Denote rejection at 1% and 5% critical values, respectively.
the shorter sample period, reported in Tables 2 and 3, are qualitatively and quantitatively similar
to the full-sample period results.25 The short sample period results thus support the conclusion
that excess supply of real money is a recurrent determinant of inflation, whose influence is not
limited to episodes of liberalization.
5. Error correction model
In the next step of the analysis we map the unrestricted UVAR system to a vector error correction system by restricting the P matrix, as suggested by cointegration analysis, and by reducing the number of the short-run parameters to be estimated. Defining ECMt as a deviation from
the long-run money demand equation, we can map the whole system from I(1) to I(0) as follows:
Dxt ¼
k1
X
Pj Dxtj þ aECMt1 þ gDt þ 3t
ð5Þ
j¼1
The analysis of the vector error correction representation starts with estimation of the system of two equations with three lags of each dependent variable, the ECM term, and exogenous
variables retained from the UVAR representation. We reduce the system by imposing zero
25
The trace and maximum eigenvalue pass the Johansen (1988) test for one cointegrating vector, but not the Reimers
(1992) test, because of the shortness of the sample. Nevertheless, conditional on there being one cointegrating vector,
weak exogeneity of inflation and real money are rejected, and weak exogeneity of output is not. The cointegrating vector is recognizable as a Cagan money demand, with coefficients similar to those obtained for the longer sample.
342
N. Budina et al. / Journal of International Money and Finance 25 (2006) 330e347
restrictions on short-run parameters with significance level lower than 10% as indicated by tstatistics. Since the hypothesis of weak exogeneity of industrial production was not rejected
and parameter reduction reveals that this variable is not Granger-caused by real money and inflation, we conclude that industrial production is strongly exogenous with respect to the remaining two endogenous variables. Consequently, we do not analyse the production
equation. The estimates of the error correction model conditional on this variable are reported
in Table 4. We test and are able to accept the validity of the last stage of reduction, by calculating the likelihood ratio of the model reported in the table versus an error correction model
similarly conditioned on exogenous production, but including the subsequently dropped
variables.26
Diagnostic tests for ninth order serial correlation, ninth order conditional heteroscedasticity
and non-normality did not reveal any problems with the specification of the model.27,28
The constancy of the system is checked through recursive estimation. Recursive residuals29
and recursive Chow tests30 for the complete system show reasonable constancy of all equations
(Fig. 4). They notably do not suggest the presence of a structural break at the beginning of
1997.
6. Economic interpretation
We infer from the tests reported in Table 3 that we cannot reject weak exogeneity of output,
but can reject weak exogeneity of both real money and inflation. In economic terms, in equilibrium, both the price level and real money are determined by the interaction of the supply and
demand for money, whereas output is not. One possible interpretation is that output is supply
determined.31 This interpretation is not inconsistent with the emphasis on restructuring and
reallocation in some analyses of the macroeconomics of transition (see Blanchard, 1997 and
the literature cited there).
The result that the dynamics of inflation depend on the supply and demand for money is one
of the principal results of this study. Month-to-month movements in the rate of inflation are satisfactorily explained by an error correction mechanism (see Table 4) in which the disequilibrium difference between supply and demand of money plays a significant role. Changes in
administered prices and changes in the real exchange rate have transitory effects on the monthly
movements of inflation.
Our results differ in emphasis from those reported in the IMF country report for 2000, IMF
(2001). The authors of that study focus on the levels of p, y, e, m and a measure of unit labor costs.32
26
The test is c2(24) ¼ 18.123[0.797]. We are not able to test the final model versus the original cointegrated UVAR
reported in Table 2, because industrial production is treated as endogenous in Table 2 and exogenous in Table 4. However, that simplification was itself previously validated by the test for the exogeneity of production, reported above.
27
The tests were run in PcFiml econometric package. See Doornik and Hansen (1994) for a description of the tests.
28
Against the charge that, given the shortness of our sample, these estimates constitute data mining, one may cite Hoover and Perez (1999), and subsequent discussion in Campos and Ericsson (1999) and Hendry (2001, Chapter 20).
29
One-step ahead residuals and their confidence intervals, for estimation ending successively in December 1995 (the
first end date chosen by the procedure)eDecember 2000.
30
One-step Chow tests scaled by their critical values at 1% level are shown in the left diagram. N-decreasing Chow
tests for stability between t and December 2000 are shown in the right diagram.
31
Popa (1998) presents evidence against the existence in Romania of a Phillips-like trade-off (working through demand channels) between inflation and output.
32
The period spanned by their data is January 1991eMarch 2000.
N. Budina et al. / Journal of International Money and Finance 25 (2006) 330e347
343
Table 4
Error correction model: FIML estimates
Estimated coefficients (standard errors in parentheses)
Dðm pÞ ¼ 0:0462 ECM1
0:3628Dðm pÞ1 þ0:2893Dðm pÞ3 0:5309DDp1 0:6172DDp2
ð0:0084Þ
ð0:0910Þ
ð0:0764Þ
ð0:1623Þ
ð0:1540Þ
0:3975 uDDpadm þdummies
ð0:0935Þ
DDp ¼ 0:0139 ECM1
þ0:2748Dðm pÞ1
0:1021Dðm pÞ2 0:2361DDp1 þ0:0700Dy3
ð0:0037Þ
ð0:0470Þ
ð0:0367Þ
ð0:0825Þ
ð0:0395Þ
þ0:0855Dðe p þ p Þ1 þ0:0358Dðe p þ p Þ2 þ0:2850 uDDpadm þdummies
ð0:0292Þ
ð0:0134Þ
ð0:0493Þ
Sample: January 1992eDecember 2000
Diagnostic statistics
Individual equations
D (m p):
Normality c2 (2) ¼ 1.4963 [0.4732]
ARCH 9 F(9,36) ¼ 1.2847 [0.2611]
AR 1e9 F(9,45) ¼ 1.7569 [0.0902]
DDp
Normality c2(2) ¼ 2.2736 [0.3208]
ARCH 9 F(9,36) ¼ 0.56493 [0.8210]
AR 1e9 F(9,45) ¼ 2.3453 [0.0213]*
System
Vector AR 1e9 F(36,160) ¼ 0.99663 [0.4831]
Vector normality c2(4) ¼ 2.2683 [0.6866]
They take all five variables to be I(1)33, find evidence for at least one cointegrating vector, and report that in their preferred specification, nominal money does not appear significantly in the longterm, cointegrating relationship. Focussing on the relationship between p, e, and unit labor cost,
they find a stable cointegrating relationship between the levels of those three variables, which they
interpret as a price equation. They conclude that ‘‘unit labor costs, and to a lesser extent the exchange rate.are the leading proximate determinants of inflation,’’ and that ‘‘the role of money
and credit growth, (though) also important, (has been) harder to demonstrate empirically.’’34
Our study finds strong evidence of a cointegrating relationship between real money, inflation
and industrial production, and provides an estimate of a stable inflation equation in which excess
money plays a significant role. In our view, focussing on the levels of nominal values is misleading,
and the evidence strongly suggests that those variables are I(2). The resulting misspecification is
responsible for the implausible rejection of a long-run cointegrating relationship, which can be
interpreted as a money demand equation.35
Our findings agree, on the other hand, with the results for Russia in Choudhry (1998). The
rejection in these studies of the weak exogeneity of inflation in Russia and Romania contrasts
with the acceptance of that weak exogeneity in many cointegration studies of money demand
33
The authors acknowledge that many of the tests reported in Appendix I could be interpreted as supporting a conclusion that the variables are I(2).
34
IMF (2001), pp. 2e3.
35
Table AII.3 of the study reports several estimates of cointegrating relationships in which monetary variables are
highly significant. See models 4a, 5a and 6a.
344
N. Budina et al. / Journal of International Money and Finance 25 (2006) 330e347
.05
ε∆(m-p)
.05
.025
0
0
-.05
-.025
1996
1997
1998
1999
2000
1996
2001
1
ε∆∆p
1997
1998
1999
2000
2001
1999
2000
2001
1
1 crit.
1↑CHOWs
1 crit.
.75
N↓CHOWs
.75
.5
.5
.25
.25
0
1996
1997
1998
1999
2000
2001
1996
1997
1998
Fig. 4. Romania: recursive residuals and Chow statistics for the conditional ECM model.
and inflation in developed, market economies. In the workshop on ‘‘Money Demand in Europe’’ cited in the introduction, weak exogeneity of prices was rejected in only one case,
that of Switzerland.36 In the other cases, the proximate causes of inflation were implicitly or
explicitly seen to be cost-push or excess demand variables. Additional country studies are needed to ascertain whether or not the direct influence of excess money supply on inflation found
here is a characteristic of high inflation, transition economies.
The fact that real money is endogenous implies that nominal money is itself endogenous. In
principle, the endogeneity of real money could reflect exclusively the endogeneity of the price
level. But, in that case, the D(m p) equation would be a simple transformation of the DDp
equation (Table 4). It can be readily shown that this is not the case. Interpretation of the endogeneity of real money is beyond the scope of this paper. If M2 was directly controlled by the
authorities, one could argue that its endogeneity suggests that monetary policy was partially
accommodating. But since we do not model the monetary transmission mechanism, we have
no basis for making that inference.
7. Conclusion
In this paper, we use cointegration techniques to examine the equilibrium relations between
real money, output and inflation in Romania, between 1992 and 2000. The three variables are
found to be linked by a cointegrating relation that can be interpreted as an expanded Cagan
(1956) money demand function. Output is shown to be strongly exogenous. But the dynamics
of inflation and real money are satisfactorily described by error correction mechanisms, which
include significant short-run effects of monetary disequilibria. The coefficients of those dynamic
36
See Peytrignet and Stahel (1998).
N. Budina et al. / Journal of International Money and Finance 25 (2006) 330e347
345
equations are stable, and do not show signs of structural break during the liberalization of 1997.
We interpret our results to imply that inflation was largely a monetary phenomenon, in Romania
between 1992 and 2000.
Acknowledgements
The authors thank Bas van Aarle, Peter Boswijk, Constantin Ciupagea, Lucian Croitoru,
Marek Dabrowski, Emilian Dobrescu, Ion Dragulin, Valentin Lazea, Maryla Maliszewska, Andrea Mihu, Raluca Miron, Elena Pelinescu, Cornelia Scutaru and Constantin Zaman for extensive discussions about data and conceptual issues, and two referees for their helpful
suggestions. This paper was written during our work as experts for the Pro Democratia Foundation International Economic Advisory Group, in Bucharest, Romania. An earlier version of
this study, entitled ‘‘An Application of a Monetary Model for Inflation in Romania’’ circulated
as a working paper of the Pro Democratia International Economic Advisory Group, Bucharest,
May 1998. The work was financed by the Pro Democratia Foundation, the Romanian Centre for
Economic Policy and the Open Society Institute.
Appendix I. Construction of the administered price index
Our administered price index is constructed from disaggregated data on consumer prices and
their weights in the CPI, provided by the Romanian Statistical Office and published in the Price
Statistical Bulletin. The list of administered prices was taken from the succession of laws and
ordinances, which defined what was to be controlled, given to us by Elena Pelinescu (National
Bank of Romania), whose procedure we have followed. Since we do not have information about
the weights used in constructing the 1993 CPI index, the figure for 1993 is based on regulations
in 1993 and product weights in 1994.
According to our estimates, the share of the average consumer basket subject to controls
evolved as follows:
Period
Share of consumer prices controlled
1992.01e1993.04
1993.05e1996.12
1997.01e1997.02
1997.03e1998.08
1998.09e1998.12
1999.01e1999.12
2000.01e2000.12
0.52
0.33
0.37
0.15
0.10
0.12
0.15
Most of the reductions of the share are the result of the freeing of previously controlled
items. Some of its variability reflects changes in the weights in the base consumer basket of
a given list of controlled items.
These shares are higher than those officially reported by the Romanian authorities prior to
1997, but they are close to the estimates reported in IMF (2001). Our worksheets are available
on request.
346
N. Budina et al. / Journal of International Money and Finance 25 (2006) 330e347
References
Aarle, B., Budina, N., 1996. Currency substitution in Eastern Europe. Journal of Policy Reform 1, 279e298.
Banaian, K., Bolgarin, I., de Menil, G., 1998. Inflation and Money in Ukraine. Working paper 98e06. Delta, France.
Blanchard, O., 1997. The Economics of Post-Communist Transition. Clarendon Press, Oxford.
Cagan, P., 1956. The monetary dynamics of hyperinflation. In: Friedman, M. (Ed.), Studies in the Quantity Theory of
Money. University of Chicago Press, Chicago, pp. 25e120.
Campos, J., Ericsson, N.R., 1999. Constructive data mining: modeling consumers’ expenditure in Venezuela. Econometrics Journal 2 (2), 226e240.
Choudhry, T., 1998. Another visit to the Cagan model of money demand: the latest Russian experience. Journal of
International Money and Finance 17, 355e376.
De Menil, G., 1997. The volatile relationship between deficits and inflation in Ukraine, 1992e1996. Economics of Transition 5 (2), 485e497.
De Menil, G., 2003. History, policy and performance in two transition economies: Poland and Romania. In: Rodrik, D.
(Ed.), In Search of Prosperity: Analytic Narratives on Economic Growth. Princeton University Press, Princeton,
Oxford.
De Brouwer, G., Ericsson, N.R., 1998. Modelling inflation in Australia. Journal of Business and Economic Statistics 16
(4), 433e449.
Dickey, D.A., Hasza, D.P., Fuller, W.A., 1984. Testing for unit roots in seasonal time series. Journal of the American
Statistical Association 79, 355e367.
Dickey, D., Pantula, S., 1987. Determining the order of differencing in autoregressive process. Journal of Business and
Economic Statistics 5, 455e461.
Dobrescu, E., 1994. Money velocity in a transition economy. The case of Romania. Romanian Economic Review 38 (2),
110e120.
Dobrescu, E., 1998. Monetary distortion in a transition economy (the case of Romania). In: 25th General Conference of
the International Association for Research in Income and Wealth, Cambridge, UK, 23e29 August.
Doornik, J., Hansen, H., 1994. An Omnibus Test for Univariate and Multivariate Normality. Discussion paper W4&91.
Nuffield College, Oxford, UK.
Doornik, J., Hendry, D., 1994. PcFiml 8: An Interactive Program for Modelling Econometric Systems. London International Thomson Publishing, London.
Durevall, D., 1998. The dynamic of chronic inflation in Brazil, 1968e1985. Journal of Business and Economic Statistics
16 (4), 423e432.
Engle, R., Hendry, D., Richard, F.-D., 1983. Exogeneity. Econometrica 51 (2), 277e304.
Eitrheim, Ø., 1998. The demand for broad money in Norway, 1969e1993. Empirical Economics 23, 339e354.
Ericsson, N.R., Hendry, D.F., Tran, H.-A., 1994. Cointegration, seasonality, encompassing and the demand for money
in the United Kingdom. In: Hargreaves, C. (Ed.), Non-stationary Time-Series Analysis and Cointegration. Oxford
University Press, Oxford, pp. 179e224.
Ericsson, N., 1998. Empirical modelling of money demand. Empirical Economics 23, 295e315.
Ericsson, N., Hendry, D., Mizon, G., 1998. Exogeneity, cointegration and economic policy analysis. Journal of Business
and Economic Statistics 16 (4), 370e387.
Ericsson, N.R., Sharma, S., 1998. Broad money demand and financial liberalization in Greece. Empirical Economics 23,
417e436.
Franses, P.H., 1991. Seasonality, nonstationarity and forecasting of monthly time series. International Journal of Forecasting 7, 199e208.
Harbo, I., Johansen, S., Nielsen, B., Rahbek, A., 1998. Asymptotic inference on cointegrating rank in partial systems.
Journal of Business and Economic Statistics 16 (4), 388e399.
Hendry, D.F., 2001. Econometrics: Alchemy or Science? Essays in Econometric Methodology, second ed. Oxford University Press, Oxford.
Hendry, D.F., Mizon, G.E., 1998. Exogeneity, causality, and co-breaking in economic policy analysis of a small econometric model of money in the UK. Empirical Economics 23, 267e294.
Hoover, K.D., Perez, S.J., 1999. Data mining reconsidered: encompassing and the general-to-specific approach to specification search. Econometrics Journal 2 (2), 167e191.
Hylleberg, S., Engle, R.F., Granger, C.W.J., Yoo, B.S., 1990. Seasonal integration and cointegration. Journal of Econometrics 44, 215e238.
IMF, 2001. Romania: Selected Issues and Statistical Appendix. Country Report No. 01/16.
N. Budina et al. / Journal of International Money and Finance 25 (2006) 330e347
347
Johansen, S., 1988. Statistical analysis of cointegration vectors. Journal of Economic Dynamics and Control 12,
231e254.
Johansen, S., Juselius, K., 1990. Maximum likelihood estimation and inference on cointegration e with application to
the demand for money. Oxford Bulletin of Economics and Statistics 52, 169e210.
Juselius, K., 1992. Domestic and foreign effects on prices in an open economy: the case of Denmark. Journal of Policy
Modelling 14 (4), 401e428.
Juselius, K., 1998a. A structured VAR for Denmark under changing monetary regimes. Journal of Business and Economic Statistics 16 (4), 400e411.
Juselius, K., 1998b. Changing monetary transmission mechanisms within the EU. Empirical Economics 23, 455e481.
Lütkepohl, H., Wolters, J., 1998a. Money demand in Europe: editors’ preface. Empirical Economics 23, 263e266.
Lütkepohl, H., Wolters, J., 1998b. A money demand system for German M3. Empirical Economics 23, 371e386.
MacKinnon, J.G., Haug, A.A., Michelis, L., 1999. Numerical distribution functions of likelihood ratio tests for cointegration. Journal of Applied Econometrics 14, 563e577.
Metin, K., 1998. The relationship between inflation and the budget deficit in Turkey. Journal of Business and Economic
Statistics 16 (4), 412e422.
OECD, 1998. Romania e economic assessment. OECD Economic Surveys 2002 (17).
OECD, 2002. Romania 1997/1998. OECD Economic Surveys 1998 (5).
Osborn, D.R., 1988. Seasonality and habit persistence in a life cycle model of consumption. Journal of Applied Econometrics 3, 255e266.
Osterwald-Lenum, M., 1992. A note with quintiles of the asymptotic distribution of the ML cointegration rank test statistics. Oxford Bulletin of Economics and Statistics 54, 461e472.
Pelinescu, E., Scutaru, C., 2000. A Dynamic Model for the Analysis of Money Demand and Inflation in Romania. Manuscript. National Bank of Romania.
Peytrignet, M., Stahel, C., 1998. Stability of money demand in Switzerland: a comparison of the M2 and M3 cases.
Empirical Economics 23, 437e454.
Popa, C., 1998. Nominal-real tradeoffs and the effects of monetary policy: the Romanian experience. In: 46th International Atlantic Economic Conference, Boston, USA, 8e11 October.
Reimers, H.-E., 1992. Comparisons of tests for multivariate cointegration. Statistical Papers 33, 335e359.
Ripatti, A., 1998. Stability of the demand for M1 and harmonized M3 in Finland. Empirical Economics 23, 317e337.
Scharnagl, M., 1998. The stability of German money demand: not just a myth. Empirical Economics 23, 355e370.
Siklos, P., Ghysels, E., Granger, C.W.J., 1996. Is seasonal adjustment a linear or non-linear filtering process? Journal of
Business and Economic Statistics 14, 374e386.
Vega, J.L., 1998. Money demand stability: evidence from Spain. Empirical Economics 23, 387e400.