Determinants of Demand for Money in Romania Coordinator : Professor Moisa Altar Author: Andreea...

Determinants of Demand for Money in Romania

Coordinator: Professor Moisa Altar

Author: Andreea Paunescu

- Bucharest, June 2002 -

Academy of Economic Studies - Doctoral School of Finance

29 June 2002 Andreea Paunescu - Determinants of Demand for Money in Romania

2

First Step: Why the interest in studying the money demand?

(Some practical aspects)

• appropriate monetary policy actions (stable function, limited number of parameters)

• recently, special interest on developing countries (flexible exchange rate, globalization of capital markets, domestic financial liberalization and innovation)

• predict the effects on ultimate target variables through shocks on monetary aggregates


3

Key - features of Romanian economy after 1990

• inflationary process lasted much longer that in the majority of ex-communist countries

• intermediate objectives of monetary policy in Romania• until the end of 1996 - exchange rate

• since the beginning of 1997 - liquidity sterilization

• 1997 - complete liberalization of the foreign exchange market

• the final objective of NBR’s policies is keeping inflation under control


4

Conceptual approach to determinants of demand for money

• Basically, money serves four major functions:

• 1. Medium of exchange• 2. Store of value• 3. Unit of account• 4. Source of deferred payment


5

Functional Form - The Basic Model

• The basic model resulting from the theory can be summarized as follows:

• (Md / P) = f (y, Ror, Roc , I, e) (1)

• where:

• Md / P = demand for real money balances (either M1 or M2)

• y = scale variable (real net income, Gross Domestic Product, etc)

• Ror = own rate of money (for that monetary aggregate)

• Roc = opportunity cost of holding money (for that monetary aggregate)

• I = inflation rate

• e = fx rate.


6

Applying the general model to M1 and M2, respectively

• Narrow Money (M1)• Md / P - real M1 as a ratio

between nominal M1 and Consumer Price Index

• Ror - own rate of M1 computed as a weighted average between 0 (for cash) and the interest rate on current account

• Roc - interest rate for deposits at Commercial Banks

• Broad Money (M2)

• Md / P - real M2 as a ratio between nominal M2 and Consumer Price Index

• Ror - interest rate for deposits at Commercial Banks

• Roc - average nominal interest rate paid on TBills


7

Assumed / expected signs of the dependencies in the general model

• The long-run real demand function is assumed to be:

• increasing in y

• decreasing in those elements of R representing rates of return on alternative assets

• increasing in own rates of return

• decreasing in inflation

• and also decreasing in exchange rate.


8

Empirical Approach• The aim of the present work and the steps followed:

• 1. Testing for stationarity in the series

• 2. Testing for cointegration between the two monetary aggregates and the specified variables, using a VAR Model and Johansen Cointegration Test

• 3. Estimating the short-run dynamics and the speed of adjustment to the long-run relationship through a VEC Model

• 4. Relating the coefficients obtained with the effective Romanian economic stance


9

Data Description

• Sample: 1997:04 2001:12 • Series used:

» ln_real_m1

» ln_real_m2» ln_rio » int_rate_depos » ln_fx_rate » inflation


10

Plots of ln_real_m1 (level and first difference)

4.4

4.5

4.6

4.7

4.8

4.9

5.0

1998 1999 2000 2001

LN_REAL_M1

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

1998 1999 2000 2001

D(LN_REAL_M1)


11

Plots of ln_real_M2 (level and first difference)

5.95

6.00

6.05

6.10

6.15

6.20

6.25

1998 1999 2000 2001

LN_REAL_M2

-0.10

-0.05

0.00

0.05

0.10

0.15

1998 1999 2000 2001

D(LN_REAL_M2)


12

Plots of ln_rio (level and first difference)

11.3

11.4

11.5

11.6

11.7

11.8

11.9

1998 1999 2000 2001

LN_RIO

-0.2

-0.1

0.0

0.1

0.2

1998 1999 2000 2001

D(LN_RIO)


13

Plots of int_rate_depos (level and first difference)

20

40

60

80

100

1998 1999 2000 2001

Monthly Average Interest Rate for Deposits at Commercial Banks

-30

-20

-10

0

10

1998 1999 2000 2001

D(INT_RATE_DEPOS)


14

Plots of ln_fx_rate (level and first difference)

8.8

9.2

9.6

10.0

10.4

1998 1999 2000 2001

LN_FX_RATE

-0.04

0.00

0.04

0.08

0.12

0.16

1998 1999 2000 2001

D(LN_FX_RATE)


15

Plots of inflation (level and first difference)

0

2

4

6

8

1998 1999 2000 2001

InflationRate in Consumer Prices (%)

-4

-2

0

2

4

1998 1999 2000 2001

D(INFLATION)


16

Unit Root TestsApplying the ADF Test all time-series appear to be integrated

of order 1 I(1), less the inflation which is I(0)

ln(real_M1)

Level

ADF Test Statistic -1.940122 1% Critical Value* -4.1314 5% Critical Value -3.4919 10% Critical Value -3.1744

*MacKinnon critical values for rejection of hypothesis of a unit root.

First difference




17

ln(real_M2)

Level



First difference




18

Industrial output

Level



First differences




19

Int_rate

Level



First differences




20

Ln (FX rate)Level



First difference




21

InflationLevel - stationary in level




22

The Specification of the Long-Run Relationships

• Following the traditional approach the basic model is specified in a log-linear form, with the exception of inflation and interest rate:

• ln(real_M1) = a0 + a1 yt + a2 I t + a3 et + a4 Rt + t (2a)

• ln(real_M2) = a0 + a1 yt + a2 I t + a3 et + a4 Rt + t

(2b)

• After determining the order of integration in the variables of interest, the Johansen procedure is applied to a VAR version of equations (2a) and (2b) to test for cointegration .


23

» owing to the short–length of the available time–series I began with a general four order ( a quarter of the year) VAR.» optimal number of lags to be included in the VAR is 4 ( based on LR and AIC criteria)» this makes sense intuitively taking into account that the data is monthly and the sample is relatively short.

Lag LogL LR FPE AIC SC HQ

0 -183.8987 NA 0.000858 7.128255 7.314131 7.199734

1 87.61926 481.5602 7.86E-08 -2.174312 -1.059052* -1.745437*

2 121.7364 54.07244 5.72E-08 -2.518355 -0.473712 -1.732084

3 149.1735 38.30846 5.57E-08 -2.610322 0.363703 -1.466656

4 183.6294 41.60704* 4.45E-08* -2.967146* 0.936263 -1.466083

Lag Length Structure


24

Cointegration Results for Equation (2a)Hypothesized Max-Eigen 5 Percent 1 PercentNo. of CE(s) Eigenvalue Statistic Critical Value Critical Value

None ** 0.545689 41.81562 33.46 38.77At most 1 0.378232 25.18499 27.07 32.24At most 2 0.238052 14.40949 20.97 25.52At most 3 0.133898 7.618907 14.07 18.63

At most 4 * 0.075839 4.180037 3.76 6.65

*(**) denotes rejection of the hypothesis at the 5%(1%) level Max-eigenvalue test indicates 1 cointegrating equation(s) at both 5% and 1% levels

1 Cointegrating Equation(s): Log likelihood 157.9327

Normalized cointegrating coefficients (std.err. in parentheses)LN_REAL_M1 LN_RIO INFLATION LN_FX_RATE INT_RATE_D

EPOS 1.000000 -0.513859 -0.057317 0.083125 6.37E-05

(0.11653) (0.01133) (0.02397) (0.00225)


25

Cointegration Results for Equation (2b)

Hypothesized Max-Eigen 5 Percent 1 PercentNo. of CE(s) Eigenvalue Statistic Critical Value Critical Value

None 0.400773 27.14206 33.46 38.77At most 1 0.375992 24.99441 27.07 32.24At most 2 0.266496 16.42587 20.97 25.52At most 3 0.115473 6.503244 14.07 18.63

At most 4 * 0.086686 4.805825 3.76 6.65

*(**) denotes rejection of the hypothesis at the 5%(1%) level Max-eigenvalue test indicates no cointegration at both 5% and 1% levels

Hence there is no cointegration between broad money and the variables considered.


26

Testing the Errors in the Long-Run Relationship for Narrow Money M1

• The long-run relationship:

• We will compute the series of residuals from the long-run equilibrium relationship and test the resulting series for stationarity:

• Resid01= ln_real_m1 - 0.5139ln_rio - 0.0573infl + 0.0831ln_fx_rate + 6.3*10–5 int_rate

• ln_real_m1 = 0.5139ln_rio + 0.0573infl – 0.0831ln_fx_rate – 6.3*10–5 int_rate

• (rel. 3a)


27

Unit Root Test on Residuals of Equation (3a)

Unit root test for residuals series:

1 lagged differences










28

Plot of Residuals and Their Significance

-1.2

-1.0

-0.8

-0.6

-0.4

1998 1999 2000 2001

RESID01

As the series of residuals is I(0) we can conclude that the values of M1 estimated by the long-run equilibrium equation (3a) don’t drift apart from the effective values so that the estimation is good.


29

The Error Correction Model

• Our purpose of building an ECM was to capture the dynamics of money demand in the short-run and especially to identify the speed of adjustment as a response to departures from the long-run equilibrium

• the first-difference of logarithm of real narrow money, is regressed on the lagged value of the error correction term (the residuals from the long-run relationship), lagged first-differences of time series from the long-run relationship and random disturbance.


30

Speed of Adjustment in the Short-RunVAR Model - Substituted Coefficients:===============================D(LN_REAL_M1) = - 0.1099725159*( LN_REAL_M1(-1) -0.5138588964*LN_RIO(-1) - 0.05731737609*INFLATION(-1) +0.08312549112*LN_FX_RATE(-1) + 6.374730187e-05*INT_RATE_DEPOS(-1) +0.7239219843 ) - 0.6656549525*D(LN_REAL_M1(-1)) -0.2265403831*D(LN_REAL_M1(-2)) + 0.2003159449*D(LN_REAL_M1(-3)) +0.3566877817*D(LN_RIO(-1)) + 0.5400513137*D(LN_RIO(-2)) +0.5087917017*D(LN_RIO(-3)) - 0.01905925299*D(INFLATION(-1)) -0.007448293855*D(INFLATION(-2)) - 0.002961759247*D(INFLATION(-3)) +0.4001598402*D(LN_FX_RATE(-1)) - 0.8777364931*D(LN_FX_RATE(-2)) +0.5836489162*D(LN_FX_RATE(-3)) - 0.002446832921*D(INT_RATE_DEPOS(-1)) - 0.00109833907*D(INT_RATE_DEPOS(-2)) -0.002284587567*D(INT_RATE_DEPOS(-3)) - 0.01029582013

Speed of adjustment : 0.11

The error correction term coefficient of (-0.11) suggest that convergence

towards equilibrium is rather slow.


31

Testing the Residuals in the Short-Run Relationship for Narrow Money M1

-.15

-.10

-.05

.00

.05

.10

.15

.20

1997 1998 1999 2000 2001

RESID02_SHORT_RUN

Plot of the series of residuals

ADF Test on resid02




32

Conclusions• We chose the sample 1997:4 - 2001:12 in order to leave behind the peaks of

inflation at the beginning of the year 1997 due to foreign exchange market liberalization and to test the relationships under the new market conditions

• The results we got on M1 is the following long-run relationship which we demonstrated to have nonstationary residuals:

• ln_real_m1 = 0.5139ln_rio + 0.0573infl – 0.0831ln_fx_rate – 6.3*10–5 int_rate

• With regard to M2 we didn’t find any cointegration. This conclusion is not very surprising given the way than we constructed the model for M2 which is unconsistent with the theory. The reason is the lack of statistical data for a rate that could be assimilated with an opportunity cost for holding M2. This variable should be in the most appropriate case the rate of return on Tbills.


33

Conclusions

• The coefficient of logarithm of real industrial output, has an expected positive sign, meaning that a increase of 1% in the industrial output generates an increase of 0.5139% in the demand for real narrow money. It is the income (GDP) elasticity of money demand.

• The sign of the coefficient of inflation is also positive, meaning that an increase in the inflation rate generates an increase in the demand for real narrow money. Although not expected, the sign of this coefficient has particular meaning in Romania and can be justified by:– an increase in the stock of money holdings due to anticipated higher expenditure

in the future given the increased level of prices (high level inflation memories of the population in the past years)

– the lack of alternative assets which makes currency in circulation, deposits and real assets almost the only possible means of holding wealth

– low investitional culture of population which is basically consumption-oriented

Interpretation of the coefficients:


34

Conclusions• The sign of the coefficient of interest rate has an expected negative sign, but it

has a low absolute value, denoting that the agents didn’t perceived during the studied period the interest rate at deposits as a guide for whether or not to keep their wealth in cash or deposits. This is also because in many periods the real interest rate was negative so that its movements didn’t influence M1.

• The elasticity of foreign exchange rate is negative, meaning that an increase of 1% in foreign exchange rate generates a decrease of 0.0831% in the demand for real narrow money. This result is expected because the depreciation of foreign exchange rate represents a cost of holding the domestic currency.

• From the short-run relationship, we got the value of the error correction coefficient which is - 0.11, which means that if narrow money was below equilibrium in the previous period then it will be increased in the current period. The value indicates a rather slow convergence towards equilibrium.


35

Conclusions• The impact of inflation and foreign exchange rate is similar (opposite

signs, but almost equal absolute values) which reflects the importance that the agents in Romania give to inflation rates and to fx rates in predicting the future evolution of the prices and in allocating their wealth among alternative assets.

• Among the limitations of the model are the following:– limited number of factors taken into consideration, especially the lack of

different types of interest rates (different maturities, different products). This will be achieved as the financial sector in Romania develops

– the speed of transformations in an economy in tranzition makes it unlikely to reach stable results with a high probability of being confirmed in time

– the short time-series considered, which can give little significance to the results.

Determinants of Demand for Money in Romania Coordinator : Professor Moisa Altar Author: Andreea...

Documents

Transcript of Determinants of Demand for Money in Romania Coordinator : Professor Moisa Altar Author: Andreea...