Mərkəzi Bank və İqtisadiyyat – N2, 2014
62
In our estimations, we use foreign currency denominated deposits as a proxy for the
cash dollars circulating in the economy. We divide the period (2001-2010) into pre-
boom (2001-2005) and post-boom (2006-2010) years. We find empirical evidence
on the elasticity of substitution between cash manats and cash dollars, and between
manat and dollar denominated deposits. However, the econometric results
underscore the declining role of dollar for transaction and saving purposes, and
correspondingly increasing share of manats in the portfolio of domestic agents in
the post-boom years. The estimation results show that the economic stabilization
years of post oil boom period established confidence in national currency and in
consequence, manat started to enjoy a new, strong status.
The structure of the paper is as follows: Section II provides the details of theoretical
model, Section III gives data description and empirical findings and Section IV
concludes.
Model and Estimation Methodology
The model employed for econometric estimation is borrowed from Imrohoroglu
(1994). The model assumes that the economy is inhabited by infinitely lived
identical households. The decisions of consumption and saving are taken at the
beginning of each period. The household decides consumption level,
, saving
level in the form of domestic real bonds,
, domestic and foreign real money
balances,
and
∗
∗
respectively. Households produce money services,
, by
combining domestic and real balances using CES function:
= [
m
p
+ (1 − )
m
∗
p
∗
]
⁄
The representative household maximizes lifetime utility function, U
∞
(c ,
m
p
,
m
∗
p
∗
)
subject to budget constraint
+
+
∗
∗
+
≤
−
+
+
∗
∗
+ (1 + r
)b
where is the discount factor and per capita consumption. Holding the real bond
one period yields a real return of
r . At the beginning of each period an
exogenous endowment
is received by each individual who pays lump-sum tax of
.
Mərkəzi Bank və İqtisadiyyat – N2, 2014
63
We assume that the utility function of the household is linear in consumption and
money services, which takes the following form:
=
+
Therefore, the equations from the optimization problem of the household can be
used as moment conditions in the non-linear system GMM estimation:
(1 + r ) − 1 = d
,
h
h
∗
ρ
+ (1 − )
⁄
h
h
∗
ρ
+
− 1 = d
,
1 −
h
h
∗
ρ
− (1 − ) 1 −
= d
,
where
h =
and
h
∗
=
∗
∗
. Purchasing power parity condition is imposed by the
equation
=
∗
.
The instrument set
consists of one period lagged values of the variables
entered the above equations
1
:
= {1,
h
h
∗
,
,
, 1 + r
}
Data and Empirical Findings
We split the sample into pre-boom (01.2001-12.2005) and post-boom (01.2006-
12.2010) years due to the structural change taken place in the economy during the
years 2005-2006. After the inauguration of Baku-Tbilisi-Ceyhan (BTC) pipeline in
the mid of 2005, oil export of the country started to increase dramatically. The pre-
boom period covers the years before BTC pipeline inauguration date and post-boom
period covers the boom and post-boom years
2
.
In each case there are 60 observations on per capita domestic and foreign money
holdings, domestic and US price level (CPI), bilateral AZN/USD exchange rate and
domestic real interest rates. To calculate per capita values of respective variables,
we use data on population, M0 (cash manat in circulation), M2 (cash manat plus
1
We also test for including different lag lengths in the instrument set.
2
Our split point of the sample is intended to characterize the structural change taken place in the country during the
period under the study. However, our evidence on the currency substitution remains robust against small variations of
the subsamples.
Mərkəzi Bank və İqtisadiyyat – N2, 2014
64
manat deposits) and dollar denominated deposits in the banking sector. All the data
are taken from the Central Bank of Azerbaijan database except data on population
which is collected from the State Statistical Committee database.
Since almost all variables incorporated into the model are in growth terms, they
demonstrate stationarity which makes our estimations robust
1
. In addition, because
the empirical model includes overidentifying restrictions we also test for the
validity of those restrictions using J-test.
In the Case 1, we interchangeably use both foreign and domestic interest rates. The
corresponding instrument set can be the p period lagged values of respective
variables. We start with the one lag, but also check more than one period lagged
values of respective values. In cases with more than one period lagged values, the
overidentifying restrictions are rejected at 5% significance level. The empirical
findings employing the national currency cash holdings (M0) and the banking
sector foreign deposits as proxies for domestic and foreign money holdings
respectively, are reported in the Table 1 below
In all cases of the non-linear GMM estimation, the objective function collapses to
almost zero after a few iterations and the results are not sensitive to initial
conditions. It is clear from the table below that all four estimated parameter values
are statistically significant at 5% significance level.
[INSERT TABLE 1 HERE]
The estimated discount factor for the period pre-boom years is approximately
0.998 (or 0.995 on quarterly basis) whereas it is around 0.997 (or 0.99 on quarterly
basis) in the post-boom years. The domestic and foreign currencies enjoy
approximately equal weights in the production of liquidity services in both pre and
post boom years. Though small, the estimate of the share of money services in the
utility function is strongly significant.
The estimate of implies that the elasticity of currency substitution
=
is
around 2.93 and 2.88 for the pre and post boom years respectively. Hence the
estimation suggests that there exists at least empirical evidence on the substitution
between Azeri manat and US dollar. Though the elasticity of substitution is larger
in the first half of the period, the difference in their magnitude is not statistically
significant. The post boom period did not lead individuals to review their decisions
and substantially reduce their holdings of cash manats. Surprisingly, the elasticity of
1
All the variables demonstrate stationarity at 10% significance level using both ADF and KPSS tests. In case of a
conflict between test results, we prefer KPSS test.