Transcript Cursul de echilibru si efectul Balassa

The Academy of Economic Studies,
Doctoral School of Finance and Banking
The equilibrium real exchange
rate and the Balassa-Samuleson
effect in Romania
COORDINATOR,
Professor Moisǎ Altǎr
MSc student,
Dumitrescu Bogdan Andrei
Dissertation paper outline
 The importance of the equilibrium real
exchange rate and the Balassa-Samuleson Effect
 The aims of the paper
 Literature review
 The model
 The data
 Empirical analysis
 Concluding remarks
 References
The importance of the equilibrium
real exchange rate and the BalassaSamuleson Effect
The equilibrium real exchange rate is very important
when judging the external competitiveness of a country.
The equilibrium real exchange rate is also a very
important variable for a country who wishes to join ERM
II and has to be taken into consideration when setting
the central parity.
The Balassa-Samuleson effect has major implications for
interpretation of the inflation and the exchange rate
criterion for membership in the European Monetary
Union
The aims of the paper
To determine the variables that influence the
behavior of the equilibrium real exchange rate
To determine the equilibrium value of the real
exchange rate using a BEER approach
To determine the misalignment of the real
exchange rate from its equilibrium value
To determine the Balassa – Samuleson effect the extent to which differences in productivity
growth between tradable and non-tradable
industries explain the observed differences in
inflation between Romania and the euro area.
Literature review
The first attempt to determine a countries equilibrium exchange rate was
made by Gustav Cassel (1922) who introduced the purchasing power
parity. This theory can be seen as a long –term tendency for the exchange
rate
Balassa and Samuleson (1964) were the first who showed that the PPP
theory is not valid. The existence of non-tradable sector creates important
deviations from the level determined by PPP.
Emprical studies that calculated the Balassa- Samuleson effect in Central
and eastern Europe: Egert (2003,2004,2005), Candelon and Kool (2006),
Oomes (2005), Mihaljek and Klau (2003)
Wiliamson (1994) introduced the concept of fundamental equilibrium
exchange rate (FEER) which is the exchange rate consistent with the
macroeconomic balance, both internally and externally
Clark and MacDonald (1998) introduced the BEER approach (Behavioral
Equilibrium Exchange Rate) which consists in estimating a reduced-form
model, that explains the behavior of the real exchange rate on medium and
long term.
Emprical studies using BEER in estimating equlibrium exchange rate for
CEEC’ Halpern and Wyplosz (2001), De Broeck and Slok (2001), Egert
(2002).
The model
This paper uses a BEER model in order to estimate equilibrium real
exchange rate.
The starting point in this model consists in expressing the real
exchange rate as a function of the expected value of the real exchange
rate at maturity t+k, the real interest rate differential and a time-varying
premium-risk:
qt E t (qt k )  (rt  rt* )   t
I assume that the unobservable expectation of the exchange rate is
determined solely by the long run economic fundamentals
qˆt  Et [1Z1t ]  1Z1t
The total misalignment from the equilibrium real exchange rate
can be expressed:
tmt  qt  1 Z1t
The long-run economic fundamentals used in this

paper are:
q  f ( prod, cons, nfa, open)
t
The Balassa-Samuleson model assumes that real exchange
rate is determine by the nominal exchange rate and the
relative price differential :
qt  (et  ptT  ptT )  [(1   * )(ptNT  ptT )  (1   )(ptNT  ptT )]
*
*
*
We assume Cobb-Douglas production function both in
tradable and non-tradable sector:
Y NT  A NT (LNT )  (K NT )1
Y T  AT ( LT )  ( K T )1
The profit functions can be written :
 NT  P NT Y NT  K NT R  LNT W
 P Y K RL W
T
T
T
T
T
Profit maximisation implies :
 T
0
K T
 T
=0
LT
 NT
0
K NT
 NT
LNT
By log-differentiating the equations above I obtain :

p NT

 p NT  p T  ( )a T  a NT
T

p
The foreign economy (the euro area) is introduced by substituting
the above relation into the first equation of the Balassa-Samuleson
model:


p  p  e  (1   )[( )a  a )]  (1   )[( )a  a
t
*
t
t

T*
t
*
NT
t
*
t
*
T*
t
NT *
t
)
The data
The source of data is Eurostat and The National Bank of Romania
database. The economies and periods covered are Romania
(1998:1-2006:3) and Euro area (1998:1 - 2006:3). The frequency of
observations is quarterly and, in the econometric work, all series are
seasonally adjusted using TramoSeats.
Description of variables:
Quarterly observations of value added from the production side GDP
estimates (decomposed into tradables and non-tradables)
CPI rates of inflation with subcomponents enabling a breakdown
into traded and non-traded goods and services
Nominal exchange rates of domestic currency against the euro
(quarterly averages)
Employment (quarterly averages) in traded and non-traded
industries
Consumption, net foreign assets, openness as a share of GDP.
All variables are in constant prices (1998=100) and in the
econometric work are expressed in logarithms.
Empirical results
In order to estimate the equilibrium real exchange rate using a
BEER approach I have followed more steps.
First I checked if the series used are stationary :
Table 1: Unit root tests for variables included in BEER approach
Series
ADF Test
Level
LRER
LPROD_DIF
LCONS
LOPENESS
LNFA
Phillips-Perron Test
First
difference
Level
Conclusion
First
differnce
-1.8843
-4.2162
-2.0237
-4.1935
(0.3354)
(0.0023)
(0.2758)
(0.0025)
-2.8915
-5.193
-1.7621
-5.2575
(0.0578)
(0.0002)
(0.3922)
(0.0001)
-2.1098
-4.4646
-1.9199
-6.3654
(0.2454)
(0.0014)
(0.3196)
(0.000)
-2.3525
-4.0148
-2.6571
-5.8834
(0.1629)
(0.0041)
(0.092)
(0.000)
-2.5709
-5.1184
-2.4083
-9.892
(0.1087)
(0.0002)
(0.147)
(0.000)
I(1)
I(1)
I(1)
I(1)
I(1)
Second I tried to determine a long run relation between variables using cointegration
test. First I estimated a VAR. I have chosen the lag length by using Likelihood ratio
(LR), Akaike informational criterion (AIC), Schwartz (SIC) şi Hanan-Quinn (HQ)
Table 2 : Choosing the lag length
Lag
0
LR
AIC
NA
SC
-10.82199
1
154.0059
-15.18279
2
33.10951
-15.19693
3
58.40837*
HQ
-10.59297
-10.74607
-13.80866*
-17.28496*
-14.72731
-12.6777
-13.62062
-14.36188
-16.07033*
The residual tests performed on the residuals are shown below :
Table 3 : Residual tests on VAR
Test statistic
P- value
LM autocorrelation test
LM(1)
21.83
(0.6455)
LM(2)
12.97
(0.1635)
LM(3)
16.81
(0.8804)
LM(4)
9.32
(0.4078)
92.27
(0.8078)
463.17
(0.3237)
JB normality test
White’s heteroskedasticity test
Var also satisfies the stability condition:
Table 4 : Checking VAR stability
Roots of Characteristic Polynomial
Endogenous variables: LCURS_REAL L_DIF_PROD LCONSUM LNFA LOPENESS
Exogenous variables: C
Lag specification: 1 3
Date: 07/04/07 Time: 15:43
Root
Modulus
0.905770 - 0.181308i
0.923738
0.905770 + 0.181308i
0.923738
-0.852741
0.852741
0.799589 + 0.244422i
0.836113
0.799589 - 0.244422i
0.836113
0.038655 + 0.822617i
0.823525
0.038655 - 0.822617i
0.823525
0.641456 + 0.478957i
0.800541
0.641456 - 0.478957i
0.800541
0.450079 + 0.650484i
0.791013
0.450079 - 0.650484i
0.791013
-0.544867 + 0.567723i
0.786886
-0.544867 - 0.567723i
0.786886
-0.364578 - 0.157192i
0.397022
-0.364578 + 0.157192i
0.397022
No root lies outside the unit circle.
VAR satisfies the stability condition.
Next I performed a Johansen cointegration test. First I estimated
a vector error correction model :
yt  yt k   yt 1  2 yt 2  ...... k 1yt (k 1)  ut
The Johansen rests on examining the rank of the  matrix
via its eigenvalues.
The test showed the presence of two cointegration vectors at
both 1% and 5% level.
Table 5 : Cointegration test for BEER Model
Sample(adjusted): 1998:4 2006:3
Included observations: 32 after adjusting endpoints
Trend assumption: Linear deterministic trend
Series: LCURS_REAL L_DIF_PROD LCONSUM LNFA LOPENESS
Lags interval (in first differences): 1 to 2
Unrestricted Cointegration Rank Test
Hypothesized
No. of CE(s)
Eigenvalue
Trace
5 Percent
1 Percent
Statistic
Critical Value
Critical Value
None **
0,91654135
139,6948506
68,52
76,07
At most 1 **
0,64863781
60,2259233
47,21
54,46
At most 2
0,34761399
26,75591662
29,68
35,65
At most 3
0,223734333
13,08811335
15,41
20,04
At most 4 *
0,144220971
4,983778548
3,76
6,65
Hypothesized
No. of CE(s)
Eigenvalue
Max-Eigen
5 Percent
1 Percent
Statistic
Critical Value
Critical Value
None **
0,91654135
79,46892732
33,46
38,77
At most 1 **
0,64863781
33,47000667
27,07
32,24
At most 2
0,34761399
13,66780328
20,97
25,52
At most 3
0,223734333
8,104334799
14,07
18,63
At most 4 *
0,144220971
4,983778548
3,76
6,65
*(**) denotes rejection of the hypothesis at the 5%(1%) level
Trace test and Max-eigenvalue test indicates 2 cointegrating equation(s) at both 5% and 1% levels
If we normalize the cointegrating vector with respect to RER,
we can obtain the following expression (standard errors in (),
t-statistics in [] ) :
RER= -1.4336*PROD_DIF - 4.8498*CONS - 0.3151*NFA +1.8390*OPENESS - 0.6816
(0.1545)
(0.8637)
( 0.029)
(0.1273)
[9.2795]
[5.6158]
[10.6710]
[-14.4494]
All coefficients are statistically significant and correctly signed.
An increase in the productivity differential ,in the net foreign assets
and in total consumption, will lead to an appreciation of the real
exchange rate.
An increase in the degree of openness will cause the current
account deficit to widen, which will require real exchange rate
depreciation.
The next step in the BEER approach requires the estimation of
the long run sustainable values for the variables used. In order
to do that, I have used a Hodrick-Prescott filter on the extend
ARIMA series.
Figure 1 : Equilibrium values for
total consumption
- Hodrick- Prescott filter
Figure 2 : Equilibrium value for
productivity differential
- Hodrick- Prescott filter
1.00
.4
0.96
.3
0.92
.2
0.88
.1
.0
0.84
-.1
0.80
-.2
0.76
-.3
0.72
-.4
98
99
00
01
CONS
02
03
04
CONS_ECH_
05
06
98
99
00
01
02
DIF_PROD_ECH
03
04
05
L_DIF_PROD
06
Figure 3: Equlilibrium value for net foreign asstes
- Hodrick-Prescott filter
.6
.5
.4
.3
.2
.1
.0
-.1
-.2
98
99
00
01
02
03
NFA_ECH
04
05
06
NFA
Figure 4: Equlibrium value for openness
- Hodrick-Prescott filter
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
98
99
00
01
OPENESS
02
03
04
05
OPENESS_ECH
06
Figure 5 :The real exchange rate and
the real equilibrium exchange rate
In order to determine the
equilibrium real exchange rate, I
have replaced the values
obtained by filtering the series
in the cointegrating relationship
estimated in step 1. :
1.2
1.1
1.0
0.9
0.8
0.7
0.6
0.5
98
99
00
01
02
RER_ECH
03
04
05
06
RER
The total misalignment of the real exchange rate from its equilibrium value was
obtained by using the next formula :
RER  ERER
Total _ misalignme nt 
x100
The results are shown below :
ERER
Figure 6 : Total misalignment real exchange rate
from its equilibrium value
20
10
0
-10
-20
-30
98
99
00
01
02
03
04
Total_misalgnment
05
06
Misalignment of the real exchange
rates – remarks
During 1998:Q1-2001:Q1 the real exchange rate was undervalued
from its equilibrium value, with a maximum misalignment of 28,96%
in the first quarter of 1999. This misalignment was caused by the
rapid depreciation of the nominal exchange rate induced by the
liberalization of prices and of the foreign currency market. Also, the
inflations expectations were at a high level which determined the
population to keep savings in foreign currency, putting even more
pressure on the nominal exchange rate.
During 2001:Q2-2003Q4 the real exchange rate was fairly valued
while in 2005Q3-2006Q4 it was slightly undervalued.
In the last period, 2005Q3 -2006Q4, the real exchange rate was
overvalued. This situation was caused by the increase in the
productivity differential and the appreciation of the nominal
exchange rate caused by the large speculative funds attracted by
the interest rate differential. The overvaluation can have the effect of
losing external competitiveness inducing a larger current account
deficit.
Estimation of the Balassa-Samuleson
effect
The variables used are : the inflation differential between Romania
and the euro area, the nominal exchange rate and the productivity
differential. The latter is calculated using the next formula :
(1   ) t ln(LPT / LP NT ) t  (1   * ) t ln(LPT * / LP NT * )
All variables are stationary at the 10% level :
Table 6: Unit root tests for variables included in Balassa equation
Series
L_INFL_DIF
L_PROD_DIF
L_NOMINAL_ER
ADF Test
Phillips-Perron Test
Level
Level
-3.2559
-6.1916
(0.0267)
(0.0000)
-2.8915
-1.7621
(0.0578)
(0.3922)
-5.0843
-5.9843
(0.0002)
(0.0002)
Conclusion
I(0)
I(0)
I(0)
The estimated equation is :
INFL_DIF = 0.7022*L_INFL_DIF(-1) + 0.1822*L_PROD_DIF + 0.2110*L_NOMINAL_ER + 0.1096
(0.0519)
(0.0458)
(0.0483)
(0.0089)
All the coefficients are statistically significant and correctly signed.
An increase in the nominal exchange rate (which stands for
depreciation) and in the productivity differential cause the inflation
differential to increase. The linear regression has a high value of R
Squared which proves that the inflation differential is well explained
by this variables.
I have obtained the Balassa-Samuleson effect by multiplying the coefficient
previously determined with the productivity differential. The results are shown
below :
Table 7 : The Balassa-Samuleson effect in Romania
Period
Inflation in the
euro area
Inflation in
Romania
Infl_dif
Prod_dif
B-S
1998
2.17%
40,68%
38.51%
-10.66%
-1.94%
1999
4.64%
54,91%
50.27%
6.31%
1.15%
2000
7.55%
40,71%
33.16%
43.56%
7.93%
2001
6.40%
30,20%
23.80%
-12.10%
-2.20%
2002
7.02%
17,89%
10.87%
-14.58%
-2.66%
2003
6.22%
14,24%
8.02%
-8.22%
-1.50%
2004
7.09%
9,16%
2.07%
-5.11%
-0.93%
2005
7.20%
8,72%
1.52%
6.44%
1.17%
2006q3
4.62%
2,77%
-1.85%
7.20%
1.31%
The Balassa-Samuleson effect explained on average only 0.5% of the inflation
differential during the covered period. However in the last 2 years the effect was
larger because of the increase in the productivity differential. It can be concluded
that factors, other than the inflation differential, were responsible for the inflation
differential between Romania and the euro area.
Concluding remarks
In the covered period the real exchange rate had some important deviations
from its equilibrium value which were determined by the liberalization of the
prices and of the foreign exchange market and by the fluctuations of the
nominal exchange rate
The nominal exchange rate had a high volatility because of its use by the
central bank in order to maintain a low inflation and because of the
speculative funds attracted by the interest rate differential. In the last period
covered, the real exchange rate seems to be fairly valued which will lead to
external competitiveness.
This paper found a negative relation between productivity differential, total
consumption, net foreign assets and the real exchange rate which is
consistent with the literature. Also, an increase of the degree of openness is
likely to cause a depreciation of the real exchange rate because of the
increased demand for tradable goods from abroad.
This paper has found evidence of the Balassa-Samuelson effect in Romania
The productivity differential explained on average only 0.5% of the inflation
differential in the period covered with a higher impact in 2005 and 2006
(1.17% and 1.31%). The conclusion is that factors different from the
productivity differential are responsible for the high inflation differential and
that the Balassa-Samuleson effect is not likely to put at risk the Maastricht
inflation criterion.
Selected References
[1] Balassa, B., (1964), “The purchasing power parity doctrine: A Reappraisal”, Jornal of
Political Economy, 6, 584-596
[2] Bergin,P., R. Glick and A.M. Taylor (2006), “Productivity, tradability and the long-run price
puzzle”, Journal of Monetary Economics, 53, 2041-2066.
[3] De Broek, M. and T. Slok, (2001), ”Interpreting real exchange rate movements in
transition countries”, IMF Working Paper 01/56.
[4] Brooks, C., (2002), “Introductory econometrics for finance”, Cambridge University Press
[5] Candelon,B., K. Raabe, T. van Veen and C. Kool (2006), “Long-run real exchange rate
determinants: Evidence from eight new EU member states, 1999-2003”, Journal of
comparative economics,35, 87-107.
[6] Chudik,A. and J.Mongardini, (2007), “In search of equilibrium: Estimating equilibrium real
exchange rates in subsaharan african countries”, IMF Working Paper, 07/90.
[7] Clark, P.B.and R. MacDonald, (1998), “Exchange rate and economic fundamentals: A
Methodological Comparaison of BEERs and FEERs”, IMF Working Paper 98/67
[8] Coricelli, F.and B Jazbec (2004), “Real exchange rate dynamics in transition economies”,
Structural Change and Economic Dynamics, 15, 83-100.
[9] Coudert, V. and C. Couharde, (2006), “Real equilibrium exchange rate in European
Union New Members and Candidate Countries”, Conference on economic policy issues in
the EU, Berlin.
[10] Egert, B.and L. Halpern (2006), ”Equilibrium exchange rates in Central and Eastern
Europe: A Meta-Analysis”, Journal of Banking and Finance, 30, 1359-1374.
[11] Egert ,B.,(2005), “Equilibrium exchange rates in South EasternEurope, Russia,
Ukraine and Turkey: Healthy or (Dutch) diseased?” , Economic Systems, 29, 205-241.
[12] Egert, B., I. Drine, K. Lommatzsch and C. Rault, (2003), “The Balassa–Samuelson
effect in Central and Eastern Europe: myth or reality?” Journal of Comparative
Economics , 31, 552-572.
[13] Egert,B., (2002), “Estimating the impact of the Balassa–Samuelsoneffect on
inflation and the real exchangerate during the transition”, Economic Systems, 26, 1-16.
[14] Faria, J.R. and M. Leon-Ledesma, (2002), “Testing the Balassa–Samuelson effect :
Implications for growth and the PPP “,Journal of Macroeconomics, 25, 241-253
[15] Froot, K.A.and K. Rogoff (1994), “Perspectives on PPP and long run real exchange
rates”, NBER Working Paper 4952
[16] Klau, M. and D. Mihaljek, (2003), “The Balassa-Samuleson effect in central Europe
: a disaggregated analysis”,BIS Working Papers, 143
[17] Lommatzsch, K.and S. Tober, (2005), “What is behind the real appreciation of the
accession countries’ currencies?An investigation of the PPI-based real exchange rate”,
Economic Systems, 28, 383-403.
[18] MacDonald, R., (1997), “What determines real exchange rates? The long and short
of it”, IMF Working Paper 97/21