INTRODUCTION
International trade has very important role in economical development of countries
(Todaro, 2003). Most of developing countries gain their
exchange incomes by primary goods and raw materials exports, but the market
of these kinds of goods is not stable, exports dependency on these goods faces
with a danger of instability in one hand and these countries need this exchange
incomes for preparing money for raw materials, machinery and capital, consumption,
intermediate and durable goods imports on the other hand. Because most of developing
countries provide money for imports by exports of raw and primary goods, they
always have deficit in balance of trade. Deficit in balance of trade produces
international money saving reduction. It is clear that the deficit in terms
of trade adjustments is one of the fundamental bases of economical growth and
development limitation. It is vivid that in all models of development models,
international trade fulfils a crucial role for all countries and it is applicable
for Iran too. Iran’s imports completely depend on petroleum incomes and the
affect of the other incomes on imports is not mentionable. During 19892003,
327.9 million ton goods were imported whose value was 276.1 billion dollars.
The maximum and minimum of imports is related to 2003 with 30 billion ton and
1994 with 16 million ton, respectively. With considering to 2004, the value
of import had been 311.3 billion dollars in these periods. In these periods
the value of imports was 5.2 times more than non petroleum export income and
the balance of trade deficit was covered export of petroleum whose value 252.8
billion dollars. With considering 2003, the deficit of Iran’s balance of trade
had been 251.1 billion dollars. In other words, because of each ton of import,
5.7 ton petroleum was exported. Intensive imports returns exchange incomes that
was produced by petroleum export to developed countries which most of Iran’s
trade and imports are happened by them. This reality changes consumption pattern
to foreign goods and prevents domestic produce process. Production reduction
will decrease employment in Iran. Because of all reasons which have been told,
focusing on import demand function is one of the most important subjects which
must be estimated by decision makers in every society. In this decade, a lot
of researches have been done in developing countries on import function and
price and income elasticity of it. Toufighi and Mehrabian
(2002) by using vector autoregressive (VAR) model showed petroleum income
and gross domestic income without petroleum have positive effect and relative
price (imported goods price to the price of domestic goods ratio) have negative
effect on import demand (total, intermediate, capital and consumption goods
imports). Mohseni (2006) investigated the effect of liberalization
on import function with using panel data. The results showed reduction of tariffs
and imposing trade liberalization parameters on demand function have a positive
and significant effect on imports growth of developing countries like Iran.
BahmaniOskooee et al. (2005) have studied the
effect of exchange rate on Canada bilateral inpayments and outpayments by
using Unrestricted Error Correction model (UECM). Results showed although the
value of export in Canada is not sensitive to exchange rate, but is sensitive
to the value of import. Yousefi and Wirjanto (2003) investigated
that 10% reduction in the exchange rate of the US dollar, will decrease export
value of Iran, Saudi Arabia and Venezuela as 3.9, 7.7 and 1.2%, respectively.
Since, the markup of Saudi Arabia is lese than Iran and Venezuela in marketing.
This country can achieve to bigger part of petroleum market with reduction in
the exchange rate of the US dollar. Truett and Truett (2003)
by estimated translog cost function for South Africa showed while capital is
a substitution for domestic labor and imported inputs, import inputs and domestic
labor are complement. Mah (2000) by using UECM showed
that the response of the demand for imported information technological goods
related to variation of price and income is negative and positive, respectively.
Therefore, with decreasing of tariffs, demand for importing of these goods will
pile up. Tang (2003ac) believed
that there is a longrun relationship between GDP, GDP without export and private
sector expenditure and government expenditure with imported goods price index
to domestic price index ratio in Japan. By estimating UECM model showed that
the demand for import regards to mentioned factors is inelastic. Hamori
and Matsubayashi (2001) assessed the longrun relationship between real
import, real GDP and relative price in Japan with using EngelGranger and JohnsonJuseleous
cointegration test and showed there is not longrun relation between mentioned
variables. Moreover GarigoriHanson test didn’t show longrun relationship between
real import, real GDP and relative price in Japan in the condition of structural
changes.
Regarding to earlier studies on the function of Iran’s import demand it is
completely vivid that recognition of structure of import in Iran is necessary
in longrun due to changing the direction of economic to an economy which is
based on production and industry instead of petroleum based economy. In fact,
import of commodities in Iran is a key component for planning in domestic production,
allocation of resources to domestic commodity, investment in industry or other
sectors and use of foreign exchange. In other hand, several factors effect import
of commodity that must determine role of each variable on import. This study
tries to estimate Iran’s import demand for 19602005 and compute elasticity
of significant variables. Therefore, in this study for investigating Iran’s
import demand Pesaran et al. (1996) method is
estimated. After estimation of import demand, elasticity of variable will compute
and interpret ate. Results of this study can help to policy making in trade
field for controlling or adjusting of imports of commodities and key variable
influencing on it.
MATERIALS AND METHODS
Data: This study focused on the application of Pesaran
et al. (1996) method for estimating Iran’s import demand function.
For this aim Iran center bank data of 19602005 have been used. Variables were
the value of import (IMP), Gross Domestic Product (GDP), imported price index
to domestic price index ratio (P^{*}/P), partial productivity of labor
(here gross national product to the number of labor ratio) (PRO) and the official
Exchange Rate (ER). Equations have been estimated by Microfit 4.0 Software.
All variables are changed to constant of 1997.
Model: With regard to sample size (relatively small), this study used
the bound testing approach to cointegration and were explored by Pesaran
et al. (1996) to examine the longrun relationship between import
and the independent variables. Pesaran et al. (1996)
suggested Unrestricted Error Correction Model (UECM) for testing cointegration
between variables and showed this method is suitable for small sample. Pesaran
et al. (1996) suggested their method based on Auto Regressive Distributed
Lag (ARDL) and separated it to two part: (1) Cointegration test and (2) estimate
the longrun coefficients. In first stage the relationship between variables
is tested and in the second stage the coefficients are estimated. So, ARDL model
is changed to error correction model like below:
where, K is the number of variables, Δ is the difference operator,
α_{0} is drift, α_{1} is the time’s coefficient,
Φ and β are longrun multipliers. The cointegration test hypothesis
is:
H_{0}: Ψ_{i} = 0, i = 0,
1, …,k 
If the null hypotheses is rejected, then there is longrun relation between
variables, but it is accepted, there is not any longrun relationship
between variables. The Ftest which has a non standard distribution depends
upon: the nonstationary properties of the data, the number of independent
variables and the sample size.
Two sets of critical values are generated. One set refers to I(1) series and
the other for I(0) series. Here, the critical values for I(1) series are referred
to as the upper bound critical values while the critical values for I(0) series
are referred to as the lower bound critical values. When the calculated Fstatistic
is greater than the upper bound critical values, the null hypotheses of no cointegration
is rejected and when the calculated Fstatistic is lower than the lower bound
critical values, the null hypotheses of no cointegration is accepted (Abrishami
and Mehrara, 2002). The UECM method has several advantages over alternatives
such as EngleGranger and Johansen Juselius methods. These advantages are:
• 
The variables can be I(0) or I(1) 
• 
It is really more suitable than another method for small sample
size 
• 
It can distinguish dependent and independent variables 
For instance, in this study UECM is following:
where, IMP, GDP, P^{*}/P, PRO and ER have mentioned explain.
ε_{t} is error term and b_{i} are model parameters.
Null hypothesis for cointegration is:
H_{0:}b_{6} = b_{7}
= b_{8} = b_{9} = b_{10} = 0 
Stationary test: in testing for a unit root in an autoregressive (AR)
model that allows for a linear time trend, Elliot et al.
(1996) showed that a modified DickeyFuller test, referred to as the DFGLS
test can achieve a substantial gain in power over conventional unitroot tests.
In addition, the DFGLS test displays good power and little size distortion
in finite samples with dependent errors. They explicitly derived the asymptotic
power envelope by analyzing the sequence of NeymanPearson tests of the unitroot
null hypothesis (α = 1) against the alternative of
in the finitesample Gaussian AR (p + 1) model, in which T is the sample size
and
is some fixed constant (DeJong et al., 1992; Harris
and Sollis, 2003; Cheung and Laib, 1994).
Let Z_{t }be the variable under examination. The DFGLS test
is carried out based on the following regression:
where, L is the usual lag operator, v_{t} is error term and Z^{d}
is the detrended variable under the local alternative of
is given by:
With W_{t} = (1, t)’ and
being the regression coefficient of ,
for which:
The DFGLS test statistic is given by the usual tstatistic testing H_{0}:
φ_{0}* = 0 against the alternative of H_{1}: φ_{0}*
<0 in regression. Elliot et al. (1996) recommend
that the parameter ,
which is responsible for defining the local alternative through ,
be set equal to 13.5 when Z_{t }is stationary with a trend and it is
equal to 7 when it is stationary without trend (Ghorbani
et al., 2007).
RESULTS AND DISCUSSION
Stationary test: Ouattara (2004a, b)
believed if the variables in UECM would be I(2), Fstatistic is not valid. For
being sure about that which our variables are I(0) or I(1) after detrending
variable by Elliot et al. (1996) method, stationary
was examined by using DFGLS test. Results show IMP, ER and PRO are stationary
in their level (I(0)), but GDP and P^{*}/ P are stationary in their
first level (I(1)) (Table 1).
Estimated model: Table 2 shows the results of UECM
estimation. Fstatistic of our model is 3.75 (F = 3.75) which is more than Fstatistic
in upper level of table that was suggested by Pesaran et
al. (1996) by considering the trend and 5 variable.
Table 1: 
Results of DFGLS test for stationary 

*Significance at 5% level 
Table 2: 
Results of UECM for Iran’s import demand function 

*Significant at 5% level, **Significant at 1% level 
With considering Table 2, it reveals that while the
coefficients of GDP (1) and PRO (1) are 0.78626 and 1728387, respectively,
the coefficients of P^{*}/P and ER(1) are negative, which show
the affects of gross domestic product and partial productivity on import
function are positive but the affects of imported price index to domestic
price index ratio (P^{*}/P) and official exchange rate on import
are negative. R^{2 }is 0.85 that shows 85% of the variations of
import are explained with significant independent variables in the model.
Fstatistic of Lagrangh coefficient for testing the autocorrelation of
residuals was 1.11 that show the equation’s residuals don’t have any autocorrelation.
Also, the Fstatistic for Ramsey’s reset test was 4.14 which is lower
than critical point at 5% significance level, therefore this model doesn’t
have specification error and model misspecification error. Moreover, test
for heteroscedasticity, Fstatistic was 0.922 thus heteroscedasticity
was rejected. CUSUM and CUSUMSQ tests show that estimated coefficients
are stable at 5% significant level. The results of CUSUM and CUSUMSQ tests
are shown in Fig. 1.
Elasticity estimation: Table 3 show by increasing
1% in GDP, import will increase 1.96%. In other words, increasing income in
Iran will create more tendencies for import. Thus economic growth in Iran will
create deficit in balance of trade. Most of Iran’s incomes are because of selling
petroleum and the incomes of that which is not a base factor for making fundamental
facilities.
Table 3: 
Longrun elasticity of import demand variables 


Fig. 1: 
CUSUM and CUSUMSQ tests for stability of coefficients.
(a) Plot of cumulative sum of rescursive residuals. (b) Plot of cumulative
sum of squares of recusive residuals. The strait lines represent critical
bounds at 5% significance level 
Also, it is a factor for increasing import with a consumption direction
without creating desirable production.
By increasing 1% in P^{*}/P, the amount of import will reduce
0.16%. it is vivid that when foreign prices index is more than domestic
prices index, the amount of import will decrease, thus with considering
the price policies which are related to domestic production, when in domestic
prices are decreased, the level of import will reduce too. Being import
function related to P^{*}/P inelastic show most of Iran’s imported
goods are necessary goods. Regarding the Table 3 indicate
that by increasing 1% in exchange rate, import will pill down 0.46% which
means when the value of money decrease, country has to pay more money
for importing commodities. Thus, decreasing the value of money and increasing
exchange rate is not a good policy in longrun, it will cause reduction
in import. Table 3 show with labour productivity promotion,
the import of goods will increase, thus in this model increasing 1% in
labour productivity, will pile up import of goods by the rate of 0.32%.
The most important reason is with increasing productivity, labours’ wages
will rise and then people will demand more commodities. Thus increasing
labours’ productivity, will make more demand for import.
CONCLUSION
The purpose of this study was the estimation of Iran’s import demand function
by using UECM which has been suggested by Pesaran et al.
(1996). Results show that demand for import is really elastic regarding
to GDP. It means with increasing GDP growth, the amount of import pile up and
it makes deficit in balance of payment. In this condition, the best way for
government in order to encounter this problem is downsizing to decreasing the
expenditure and the amount of imports which are related to the government expenditure.
Moreover, the import increasing because of the growth of economic show Iran
is going to the open economy but this trend should not be a factor for reducing
the possibility and power of domestic firms face to foreign goods imports. Gafer
(1988) determined the price elasticity of import demand in Trinidad and
Tobago which was 0.05 to 0.15 and concluded reforming of exchange rate policies
will improve and decrease instability in balance of payment. Estimated the price
elasticity of import demand in Iran is 0.16 which is close to Gafer
(1988) estimation. Thus exchange rate reformation will be useful for decreasing
deficit in balance of payment, because increasing exchange rate will make expensive
the value of imports and it will improve the deficit in balance of payment.