You are on page 1of 7

MONEY SUPPLY AND REAL GDP: THE CASE OF THE

CZECH REPUBLIC
RADEK BEDNARIK1

Keywords:

Money supply, vector autoregression, causality, product

Introduction
In general, economic theory posits, that (in the long-run) there shouldn’t be a significant
relationship between money supply and real product. Simple explanation of this statement is
provided by generally known equation (1):

‫ܳ × ܲ = ܸ × ܯ‬, (1)

where M is money supply, V measures velocity of money, P stands for price level and Q
means real product. Now, if V is constant or, at least, stable in the long-run, the quantitative
theory of money states that, in the long-run, the increasing of M leads only to increasing of P,
that is, money supply increment only causes growth of nominal, not real, product2.

This is what theory suggests. But is this statement valid empirically? This is the question we
want to answer in this paper. As the empirically examined economy we picked up the Czech
Republic. For analysis, we use rather conventional methods, which are the graphical analysis
of scattergrams and the vector autoregression (VAR) analysis.

Data
Data were taken from online database ARAD provided by Czech National Bank (CNB). The
timeline spans over period between first quarter of 2002 to third quarter of 2009. Money
supply is measured via M3 monetary aggregate, as it is defined by CNB. Real product is
measured as real GDP and was seasonally adjusted.

Visual relationship between data
In this part of analysis, we took money supply (M3) time series, lagged them by various lags
and plotted them against seasonally adjusted real GDP (GDPRSA). The result is displayed in
Fig. 1.

1
This paper is published to show the results of my ongoing work and as such it is not fully polished. It aims to
elicit debate and further comments.
2
More precisely, change of M must be higher than the change of Q, or: ∆‫ܳ∆ > ܯ‬, other things equal.
Fig. 1 Money supply and real GDP relationship

2,800,000 2,800,000

2,400,000 2,400,000

2
M3 L
2,000,000 3 2,000,000
M

1,600,000 1,600,000

1,200,000 1,200,000
500,000 600,000 700,000 800,000 500,000 600,000 700,000 800,000

GDPRSA GDPRSA

2,600,000 2,600,000

2,400,000 2,400,000

2,200,000 2,200,000

4 2,000,000 6 2,000,000
L L
3 3
M M
1,800,000 1,800,000

1,600,000 1,600,000

1,400,000 1,400,000

1,200,000 1,200,000
500,000 600,000 700,000 800,000 500,000 600,000 700,000 800,000

GDPRSA GDPRSA

2,400,000

2,200,000

2,000,000
8
L 1,800,000
3
M

1,600,000

1,400,000

1,200,000
500,000 600,000 700,000 800,000

GDPRSA

M3L[x] means money supply measured as M3 monetary aggregate (defined by CNB), lagged by 2, 4, 6 and 8
quarters. GDPRSA is real gross domestic product, seasonally adjusted.
Source: CNB (ARAD), own computations.

Our visual analysis given by Fig. 1 suggests clear and positive relationship between money
supply and product, no matter how many lags are applied. Quite interesting piece of
information is given by the upper-right part of the scattergrams. At certain point (or threshold)
the curve turns back, suggesting that at some level the linear relationship between supply and
product ends and the money supply doesn’t have positive influence on real GDP anymore.

However, scattergrams can’t tell us, whether the relationship between variables is one-, or
two-way one. Thus, we should test for mutual, two-way relationship and if possible, try to
quantify it.

Confirming mutual, two-way relationship
First, we have to test for (non)stationarity of variables. The summary of these tests gives us
Table 1.

Tab. 1 Unit root tests results

Group unit root test: Summary
Series: GDPRSA, M3
Sample: 2002Q1 2009Q3

Cross-
Method Statistic Prob.** sections Obs
Null: Unit root (assumes common unit root process)
Levin, Lin & Chu t* -2.11950 0.0170 2 52

Null: Unit root (assumes individual unit root process)
Im, Pesaran and Shin W-stat -2.46706 0.0068 2 52
ADF - Fisher Chi-square 17.8640 0.0013 2 52
PP - Fisher Chi-square 0.88431 0.9268 2 60

** Probabilities for Fisher tests are computed using an asymptotic Chi
-square distribution. All other tests assume asymptotic normality.

Source: CNB (ARAD), own computations.

We can see, that time series are stationary, as most of various tests suggest. This allows us to
use VAR method. However, we may want to know, whether the cointegration (that is the
long-run) relationship between variables is present. If that is so, we could use vector error-
correction (VEC) method, which is able to quantify the long-run relationships between
variables.

Results of cointegration test are provided by Table 2.
Tab. 2 Cointegration test results

Sample: 2002Q1 2009Q3
Included observations: 26
Series: GDPRSA M3
Lags interval: 1 to 4

Selected (0.05 level*) Number of Cointegrating Relations by Model

Data Trend: None None Linear Linear Quadratic
Test Type No Intercept Intercept Intercept Intercept Intercept
No Trend No Trend No Trend Trend Trend
Trace 0 0 0 1 2
Max-Eig 0 0 0 0 0

*Critical values based on MacKinnon-Haug-Michelis (1999)

Source: CNB (ARAD), own computations.

Again, testing indicates, that most probably no cointegration relationship is present. So the
VEC is not an option here, we will use VAR.

It may be also useful to test whether the two-way causality exists between variables. This
could be tested by pair-wise Granger causality test, as provided by Table 3.

Tab. 3 Pair-wise Granger causality test

Pairwise Granger Causality Tests
Sample: 2002Q1 2009Q3
Lags: 4

Null Hypothesis: Obs F-Statistic Prob.

M3 does not Granger Cause GDPRSA 27 5.39493 0.0049
GDPRSA does not Granger Cause M3 5.68226 0.0039

Source: CNB (ARAD), own computations.

As we can see, test rejects the null hypothesis of no Granger causality between variables, thus
we may conclude, that the two-way causal relationship between money supply and real
product seems to be present.

Now we can proceed to VAR analysis. The results are presented in Table 4.

Tab. 4 Vector autoregression estimation results

Vector Autoregression Estimates
Sample (adjusted): 2003Q1 2009Q3
Included observations: 27 after adjustments
Standard errors in ( ) & t-statistics in [ ]

GDPRSA M3
GDPRSA(-1) 0.512340 0.781760
(0.26954) (1.02777)
[ 1.90081] [ 0.76064]

GDPRSA(-2) 0.186955 -0.316976
(0.32515) (1.23983)
[ 0.57498] [-0.25566]

GDPRSA(-3) 0.525667 0.830250
(0.34610) (1.31971)
[ 1.51883] [ 0.62911]

GDPRSA(-4) 0.591525 0.097338
(0.41054) (1.56544)
[ 1.44083] [ 0.06218]

M3(-1) -0.264257 1.088398
(0.07868) (0.30003)
[-3.35843] [ 3.62761]

M3(-2) 0.161281 -0.404562
(0.09719) (0.37060)
[ 1.65941] [-1.09164]

M3(-3) 0.098489 -0.209366
(0.10290) (0.39235)
[ 0.95716] [-0.53361]

M3(-4) -0.154154 0.339482
(0.07532) (0.28718)
[-2.04678] [ 1.18210]

C -225898.3 -518627.0
(88986.5) (339313.)
[-2.53857] [-1.52846]

R-squared 0.992422 0.997629
Adj. R-squared 0.989053 0.996576
Sum sq. resids 7.88E+08 1.15E+10
S.E. equation 6617.882 25234.57
F-statistic 294.6482 946.8523
Log likelihood -270.3709 -306.5088
Akaike AIC 20.69414 23.37102
Schwarz SC 21.12609 23.80296
Mean dependent 685199.2 1996887.
S.D. dependent 63253.00 431231.7

Determinant resid covariance (dof adj.) 1.67E+16
Determinant resid covariance 7.41E+15
Log likelihood -569.9279
Akaike information criterion 43.55022
Schwarz criterion 44.41411
Model was tested for normality, auto- and cross-correlation, heteroskedasticity. All tests indicates, that model is
valid.
Source: CNB (ARAD), own computations.

VAR output shows us, that GDPRSA has really strong and positive influence on itself and on
money supply as well. The influence of M3 on itself and GDPRSA is rather weaker and with
mixed signs.

Maybe better way to show relationship between variables is using graphical output from
Impulse-Response analysis. This will show us, how one variable reacts in response to random
shock (change) of other variable. See Fig. 2.

Fig. 2 Response to Generalized One S.D. Innovations
Response of GDPRSA to GDPRSA Response of GDPRSA to M3
20,000 20,000

10,000 10,000

0 0

-10,000 -10,000

-20,000 -20,000
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

Response of M3 to GDPRSA Response of M3 to M3

40,000 40,000

0 0

-40,000 -40,000

-80,000 -80,000
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

Source: CNB (ARAD), own computations.

Thanks to Fig. 2 we may clearly see the reaction and its development of one variable to
change of the other one.

The reaction of GDPRSA to random shock of itself is positive and tends to rise in time. That
may suggest that the real product growth has some self-incentive feature. Or in other words –
it is autoregressive.

The reaction of GDPRSA to random shock in money supply M3 seems to confirm the validity
of quantitative theory of money, as stated above. As the time goes, the reaction of product to
change in money supply tends to be more and more negative. According to this we could say,
that increasing of the money supply in the economy will not have a positive long-run impact
on the real GDP growth.
Response of M3 to GDPRSA seems logical – as the real product grows so must the money
supply, since the number of money needed for transactions rise3.

Response of M3 to random shock of itself is negative and growing stronger over time. This
may be due to reaction of the central bank by increasing the basic official interest rate and
therefore cutting the amount of new loans provided by commercial banks.

Conclusion
In this paper, we wanted to find out, whether the quantitative theory of money holds in the
Czech Republic, that is, whether there is a strong empirical relationship between money
supply and product.

We found out, that there is indeed strong and mutual relationship between these two variables.
Scattergram analysis shows clear and positive relationship between money supply and
product, no matter how much we lag the money supply variable.

But Impulse-Response analysis shows us that real product reacts in negative way to random
shock in money supply. This difference can be explained by the fact, that scattegram analysis
cannot capture the possible two-way relationships, as VAR do. Thus, the indicated positive
relationship between money supply and product may be caused by product to money supply
causal relationship and not vice versa. This is de facto confirmed by Impulse-Response
analysis, as it is given by Fig. 2.

3
If money supply didn’t react, the velocity of money would have to change. And as we know, it is assumed to
be, at least, stable in the long-run.