The Parallel Market for Foreign Exchange in an Oil Exporting Economy
The Case of Iran, 1978-1990

This paper provides a model for the determination of the parallel market exchange rate premium in a country where oil export earnings accrue directly to the government, and foreign exchange is centrally allocated for the importation of specific goods. Next, it studies the parallel market for foreign exchange In the Islamic Republic of Iran during the period 1978-90. The paper then examines the various time series properties of parallel market exchange rate in Iran, and the evidence of the role of oil and non-oil exports in the determination of the parallel market premium.

Abstract

This paper provides a model for the determination of the parallel market exchange rate premium in a country where oil export earnings accrue directly to the government, and foreign exchange is centrally allocated for the importation of specific goods. Next, it studies the parallel market for foreign exchange In the Islamic Republic of Iran during the period 1978-90. The paper then examines the various time series properties of parallel market exchange rate in Iran, and the evidence of the role of oil and non-oil exports in the determination of the parallel market premium.

I. Introduction

The purpose of this paper is to develop a theoretical model for the determination of the parallel market exchange rate premium in an oil exporting economy where oil revenues accrue directly to the government. Furthermore, it attempts to analyze empirically developments in the parallel market premium for the U.S. dollar in Iran during the period 1978-1990, up to the major reform of the exchange system in January 1991.

Iran has experienced severe balance of payments pressures since 1978 on account of a number of domestic and external shocks, such as the Iranian revolution, capital flight, economic sanctions by the West, the protracted war with Iraq (1980-88), and by an overall decline in oil export earnings since the revolution. The decline in oil and gas revenues, which accounted for about 96 percent of Iran’s total exports during the period 1978-90, was particularly detrimental. Such exports declined from about US$21 billion in 1978 to a low of about US$6 billion in 1986, and rose to US$12 billion in 1990. The Iranian authorities responded to these balance of payments pressures through the imposition of severe capital controls and a multiple exchange rate system. Although Iran had a dual exchange rate regime immediately after the revolution, the system gradually grew in complexity and in certain periods over ten different exchange rates were in effect. The authorities also adopted a highly restrictive policy of import compression, and centralized the allocation of foreign exchange through an annual foreign exchange budget which distributed foreign exchange to different sectors based on political and economic priorities.

As a result of the above developments, there arose a very active parallel market for foreign exchange. This market, which had been virtually dormant since the end of the balance of payments crisis of the 1960s, re-emerged on account of political developments in early 1978 and became one of the most important features of Iran’s postrevolutionary economy. 1/ Chart 1 depicts the movements in the parallel market premium for the U.S. dollar. 2/ The premium (PMP), which exceeded 2,000 percent, is among the highest ever observed internationally. The simple parallel market premium maintained its steep ascent from 1978 to 1990 with only one serious interruption, namely, the temporary decline in 1988 when the war with Iraq ended.

Chart 1:
Chart 1:

The Parallel Market Premium for the U.S. Dollar

(In percent)

Citation: IMF Working Papers 1995, 069; 10.5089/9781451849035.001.A001

The Iranian foreign exchange system was partly reformed in January 1991 when the number of exchange rates was reduced from seven rates to three, including a floating exchange rate. Subsequently, the exchange rate system was temporarily unified in March 1993, and a floating exchange rate system was adopted. However, as the unification of the exchange system was anticipated long in advance, imports of goods that were permitted at the official rate rose sharply prior to the unification. At the same time, a decline in world oil prices diminished Iran’s foreign exchange earnings. As a result of these developments, Iran accumulated sizable arrears on its external obligations, and the unified exchange rate system was abandoned in favor of a multiple exchange rate system. 1/

The remainder of the paper is organized as follows: Section II provides a model for the determination of the parallel market premium. Section III examines the statistical properties of the parallel market exchange rate in Iran, and provides an empirical test of the model. Section IV concludes the paper.

II. The Model

The framework presented here is an adaptation of the perfect foresight model provided in Lizondo (1991) to an oil exporting economy. 2/ It incorporates some of the specific features of these economies such as the receipt of oil revenues by the government in the form of foreign exchange. 3/ In addition, the model presented here includes a foreign exchange rationing board that allocates foreign exchange to various private sector imports by determining the type of goods which may be imported at the official exchange rate.

The hypothetical oil exporting economy operates under a dual exchange rate system that comprises an official rate, e, and a parallel market rate, b. 1/ In order to simplify the analysis, the problem of leakages between the official and the parallel markets is ignored; this problem has been studied in detail in previous works such as Bhandari and Vegh (1990), Gros (1988), and Nowak (1984). 2/ It is assumed that the economy produces three types of goods: oil, h, which is produced and exported by the government, a non-oil exported good, x, and a nontraded good, n. The output of the nontraded good, qn, and of the non-oil exported good, qx, are fixed. Both the dollar price and the export quantity of oil (ph, qh) are determined by a deus ex machina, which can be called OPEC. Hence, total oil revenues (H=phqh) received by the country are determined exogenously. The price of the non-oil exported good Is assumed to be equal to one. 3/

Private sector agents spend a fraction, ω), of their nominal wealth on the consumption of nontraded and imported goods. It is assumed that agents behave as if they had a Cobb-Douglas utility function, and that they uniformly distribute a fraction a of their consumption on a continuum of imported goods, which is indexed by θ (0≤θ≤1), and a fraction (1-α) on the nontraded good. The world prices of the imported goods are also fixed at unity, while the price of the nontraded good, pn, is determined endogenously by the market clearing condition in the nontraded good market. The government sector consumes only imported and nontraded goods; the government consumption of imported and of nontraded goods are represented by gm and gn, respectively.

The country has a foreign exchange board whose function is to allocate foreign exchange to different types of imports and determine the amount of goods that may be imported at the official exchange rate. Demand for foreign exchange for importing other goods, and asset demand for foreign exchange are left to be met in the parallel market. The board obtains foreign exchange from two sources, namely, the government and the surrender requirements on non-oil exports. In order to simplify the analysis, we assume that the fraction of the oil revenues allocated to the public sector for its consumption of imported goods is a fixed share, β, of total oil receipts. 4/ This implies that the government is obliged to sell the remaining portion (1-β) to the central bank at the official exchange rate. 1/ In addition, the foreign exchange board obtains foreign exchange through the surrender requirements on non-oil exports, which amount to (1-ϕ)qx. The surrender requirement rate, (1-ϕ), is determined exogenously. The total expected real foreign exchange available to the foreign exchange board, S, is as follows: 2/

S=(1β)H+(1φ)qx.(1)

Given perfect foresight, the expected receipt of foreign exchange by the board is equal to the realized amount. The foreign exchange board determines the range of goods that may be imported at the official exchange rate based on the value of S. We define the range of the goods that can be imported at the official exchange rate as (1-θ). As the amount of foreign exchange surrendered to the board from various sources declines, it reduces the number of goods that can be imported at the official exchange rate. 3/ In view of the above institutional arrangements, the parameter θ is determined according to the following rule:

θ=θ(S),θ<0.(2)

Equation (2) states that the range of goods imported by the private sector at the official exchange rate diminishes as the amount of foreign exchange available to it declines. 4/

Given the above structure, the consumer price index is:

P=Pn1αeα(1θ)bαθ=eρ1αdαθ,(3)

where d = (b/e) is the parallel market premium for dollars, 1/ and ρ = (Pn/e) is the price of the nontraded good relative to that of the traded goods, which is a measure of the official real exchange rate.

Let us proceed to a discussion of the private sector’s portfolio choice. Agents hold their wealth, W, in the form of noninterest-bearing domestic and foreign currency, symbolized by M and f, respectively. Assuming that asset transactions are conducted on the parallel market, foreign exchange holdings are valued at the parallel market exchange rate. Hence, the representative agent’s nominal portfolio is:

W=m+bf.(4)

If equation (4) is deflated by the price level given in equation (3), the real private sector wealth is obtained:

w=m+fρα1d1αθ,(5)

where m is the real stock of money, M/P. The fraction of the agent’s wealth that is held in the form of domestic currency, λ, is inversely related to the anticipated rate of depreciation of the parallel market exchange rate. Since agents are assumed to possess perfect foresight, the anticipated rate of depreciation of the parallel market rate is equal to the actual rate. Furthermore, given that the official exchange rate is fixed, the rate of change of the parallel market rate, ḃ/b, is equal to the rate of change of the differential between the official and the black market rate, ḋ/d. It is also assumed that foreigners are enjoined from holding domestic currency. In view of the above assumptions, demand for domestic currency is:

m=λ(b˙b)w=λ(d˙d)w,0<λ<1,λ()<0.(6)

Instantaneous clearing of the asset market provides the following portfolio relationship between the domestic and the foreign currencies:

m=λ(d˙d)1λ(d˙d)fρα1d1αθ.(7)

Movements in the supply of money are determined by changes in domestic credit, D, and by variations in official foreign exchange reserves, r:

M˙=D˙+er˙.(8)

Assuming that tax revenues are equal to zero, and that the government cannot resort to foreign borrowing, changes in domestic credit are determined by the budgetary needs of the government. The budget deficit is equal to the difference between the expenditures on the nontraded good and oil revenues net of government purchases of the imported goods:

D˙=e(gmH)+Pngn=Pngne(1β)H.(9)

The above formulation indicates that while the government will be operating with a surplus in its foreign transactions, its overall budget could, nevertheless, be in deficit. It is also evident from equation (9) that the receipt of petroleum revenues by the government in the form of foreign exchange significantly reduces the need to issue domestic credit and diminishes the chances of a foreign exchange crisis.

Movements in the level of foreign exchange reserves are determined by the developments in the official current account: 1/

r˙=(1φ)qx+(1β)Hαω(1θ)(mp1αdαθ+df).(10)

Using equations (8)-(10), changes in the real stock of money are determined as follows:

m˙=(1φ)qxρα1dαθαω(1θ)(m+fρα1d1αθ)+gnραdαθ(P˙P)m.(11)

A significant feature of equation (11) is that neither oil revenues nor the government’s consumption of imported goods affects the changes in the money stock if it is assumed that θ does not depend on oil revenues, H. 1/ Indeed, a corollary of this result is that the value of the government expenditure policy parameter β does not affect changes in the money stock either, so long as government expenditure on the nontraded good remains unchanged. 2/

Changes in the stock of foreign exchange held by the private sector are, as in Lizondo (1991), determined by the excess of the supply of dollars from non-oil exports to the parallel market, ϕqx, over imports allowed in that market: 3/

f˙=φqxαθω(mρ1αdαθ1+f).(12)

In the steady state, the market for the nontradable good clears, and the parallel market premium, the real money supply, the stock of foreign exchange in the parallel market, and the price level are constant [ḋ=ṁ=ḟ=ṗ=0]. 1/ The equilibrium condition for the nontraded good market is: 2/

qn=gn+ω(1α)(mραdαθ+fρ1d).(13)

The above equilibrium condition determines the price of the nontraded good and, consequently, the value of the real exchange rate.

Solving the system of equations (7) and (11)-(13) provides the steady-state values of the model. The steady-state value of d is reported below and those for m, f, and the equilibrium value of ρ are provided in Appendix II: 3/

d*=θ(1φ)φ[qn(1+1αα(1θ))gn].(14)

It should be noted that if the functional relationship between the variable θ on the total foreign exchange available to the banking system, S, is ignored, oil revenues and the volume of public sector imports, gm, will have no effect on the steady-state value of the parallel market premium. The stability properties of the dynamical system (11)-(13), which exhibits saddlepath properties, are discussed in Appendix II.

Some Comparative Static Results

The sign of the effect of changes in oil revenues, non-oil exports, and the parameter β are as follows: 1/

(A)d*H<0,(B)d*qx<0,(C)d*gx>0,(15)
(D)d*qn<0,(E)d*β>0,(F)d*φ0.(16)

Results A and B, not surprisingly, state that the parallel market premium declines as oil and non-oil exports rise. An increase in such exports would raise the level of foreign exchange available to the foreign exchange board, and lead to an extension in the range of goods that could be imported by the private sector at the official exchange rate. This would induce a lower demand for foreign exchange in the parallel market, and the premium would therefore decline. 2/ Also, an increase in non-oil exports implies that the supply of foreign exchange in the parallel market will increase, reducing the premium. Result C indicates that an increase in government expenditures on the nontraded good would lead to an increase in its price, and an increase in domestic credit to the government. The parallel market premium would then increase as private agents try to restore their portfolios to the desired level. An increase in the output of the nontraded good would lead to a decrease in the price of the nontraded good, and the premium (result D), since such an output increase would set in motion forces opposite to those caused by an increase in government expenditures on the nontraded good as discussed under result C. Result E states that if the government increases its own imports and lowers the amount of foreign exchange made available to the foreign exchange board for allocation to private sector importers at the official exchange rate, the demand for foreign exchange in the parallel market would rise, thereby leading to a higher premium. Result F suggests that a decrease in the parameter ϕ, which implies an increase in the surrender requirements on non-oil exports, would have an ambiguous impact on the premium. On the one hand, an increase in the surrender requirement rate would increase the foreign exchange available to the foreign exchange board and allow for more goods to be imported at the official rate, which would then lower the demand in the parallel market. On the other hand, a higher surrender requirement rate implies that non-oil exporters would reduce their supply of foreign exchange to the parallel market. The net impact of the lower demand and supply of foreign exchange on the premium is, therefore, ambiguous.

III. Empirical Examination

This section first examines the statistical properties of the parallel market exchange rate for the U.S. dollar in Iran. The model proposed in Section II is then econometrically examined.

1. Preliminary statistical analysis

The distribution of changes in the rial/dollar parallel market exchange rate

The basic statistics of the distribution of the rate of change of the rial/dollar parallel market exchange rate for the period covering January 1978 to December 1990 are provided in Table 1. It is worth noting that the reported statistic for skewness indicates some asymmetry in the distribution of the rate of change in the parallel market exchange rate. On the other hand, the distribution is characterized by leptokurtosis (fat tails), which indicates the presence of numerous large changes in the parallel market rate in Iran. This feature is also observed in other exchange rates. 1/

Table 1.

The Statistical Properties of the Rate of Change in the Rial/Dollar Parallel Market Exchange Rate

(in percent, January 1978 - December 1990)

article image
Note:The maximum monthly increase in the parallel market rate occurred in June 1980 while the maximum decline in the parallel market rate occurred in December 1980.

Chart 2 illustrates the monthly movements of the rate of change of the rial/dollar parallel market exchange rate. It indicates that the most volatile period in the parallel exchange market was between 1979 and 1982, which coincided with pronounced political upheavals. Since 1982, however, while the parallel market rate rose steadily, its volatility diminished. Chart 3 presents the frequency distribution of the rate of change in the rial/dollar parallel exchange rate. 1/

Chart 2:
Chart 2:

Volatility of the Rial/Dollar Parallel Market Rate

(percent monthly change: January 1978 - December 1990)

Citation: IMF Working Papers 1995, 069; 10.5089/9781451849035.001.A001

Chart 3:
Chart 3:

Distribution of the Rate of Change in the Parallel Market Exchange Rate

(In percent: Jan. 1978-Dec. 1990)

Citation: IMF Working Papers 1995, 069; 10.5089/9781451849035.001.A001

Unit roots test

In order to examine the stationarity of the parallel market exchange rate series, unit root tests were conducted by estimating Dickey-Fuller and augmented Dickey-Fuller equations. 2/ The test results are reported in Table 2. Given that the ττ (the test statistic corresponding to δ) is -2.63, the hypothesis of a unit root in the logarithm of the parallel market exchange rate series cannot be rejected. Hence, the evidence indicates that the parallel market exchange rate series was nonstationary [I(1)]. Furthermore, the Φ3 statistic is equal to 3.89 which suggests that the null hypothesis that the logarithm of the parallel market exchange rate follows a random walk with drift cannot be rejected at the 95 percent level of confidence. Dickey-Fuller and augmented Dickey-Fuller tests were also used in order to test for the existence of a unit root in the first difference of the series. The null hypothesis of a unit root in the first difference of the series was rejected (Table 2). 1/

Table 2.

Unit Root Test Results for the Rial/Dollar Parallel Market Exchange Rate

(January 1978 - December 1990)

article image
Notes:

The equation estimated is:

ΔBt=α+βT+δBt1+Σi=19γiΔBt1,

where B is the natural logarithm of the parallel market exchange rate, and T is a time trend. The lag length was chosen according to the Akaike Information Criterion [See Harvey (1990)]. The figures in parentheses are t-statistics. Initially, eleven monthly dummy variables were included in the regression equation in order to test for seasonality effects, yet all of their coefficients proved statistically insignificant.

The equation estimated is:

Δ2Bt=α+βT+δΔBt1+Σi=15γiΔ2Bt1.

The critical region for the Dickey-Fuller test at the α = 0.05 level is −3.4398.

The Φ3 statistic is the Dickey-Fuller statistic for testing that the series follows a random walk with a drift (β = δ = 0).

Significant at the α = 0.05 level.

The autocorrelation function

The autocorrelation function of changes in the logarithm of the parallel market exchange rate was estimated in order to obtain descriptive information about the parallel market exchange rate in Iran. Estimated autocorrelation coefficients are presented in Table 3. All estimated coefficients, except those for lag lengths 1, 2, 3, and 5, proved to be statistically significant, indicating the presence of serial correlation in the above series at high lag orders. Of particular interest is the sharp jump between the autocorrelation coefficients for lag lengths of 5 and 6 months, suggesting that the sixth order lagged value of the rate of change in the parallel market rate has some power in explaining the movements in the parallel market exchange rate in the current period. 2/

Table 3.

Autocorrelation Coefficients of the Rate of Change in the Rial/Dollar Parallel Market Exchange Rate

(January 1978 - December 1990)

article image
Notes:

Q(p) refers to the Ljung-Box statistic:

Q(p)=n(n+2)Σj=1p(nj)1rj2,

where n is the number of observations, p is the lag length, and rj is the estimated autocorrelation coefficient. The Ljung-Box statistic is distributed approximately as a chi-square variate with p degrees of freedom.* All autocorrelation coefficients, except for those for n = 1, 2, 3, and 5, were significant at the α = 0.05 level.

Runs test

A nonparametric test of the randomness of changes in the parallel market exchange rate is possible through a runs test. 3/ A run, R, is defined as a group of consecutive price changes of the same sign: +, -, or 0. For large samples, the sampling distribution of R is approximated by a normal distribution and, under the null hypothesis of the independence of price changes, the expected value and the standard error of R are:

E(R)=[N(N+1)Σi=13ni2]N,(17)

and

σR=[Σi=13ni2[Σi=03ni2+N(N1)]2NΣi=13ni3N3N2(N1)],1/2(18)

where ni is the number of observations of each sign (+, -, or 0), and N is the total number of observations (N = Σni). The test statistic is:

Z=R+1/2E(R)σR,

which, under the null hypothesis, has a standard normal distribution in large samples. The 1/2 in the numerator is for discontinuity correction.

Table 4 provides details of actual runs of different length of the rate of change of the parallel market exchange rate around its mean. 1/ While the total number of positive and negative runs were roughly equal, as Table 4 indicates, the negative runs were longer in duration. Total expected and actual runs are reported in Table 5. Given a Z-score of −16.83, the hypothesis of the independence of changes in the parallel market exchange rate is soundly rejected. This result is in accordance with the Box-Ljung test results reported above, which indicated the presence of serial correlation in the parallel market exchange rate series. However, the presence of serial correlation differs from most of the results obtained by Gupta (1981) in his examination of parallel market exchange rates in India, South Korea, and the Taiwan Province of China. The contrasting results might be due either to a significant pattern of serial correlation in the fundamental determinants of the rial/dollar parallel market exchange rate (e.g., oil and non-oil export earnings), or possibly the presence of inefficiency in the Iranian foreign exchange market.

Table 4.

Runs Analysis of the Rate of Change in the Rial/Dollar Parallel Market Exchange Rate

(January 1978 - December 1990; 156 months)

article image
Table 5.

Summary Results of the Runs Analysis of the Rial/Dollar Parallel Market Exchange Rate

(January 1978 - December 1990)

article image

2. Regression results

The analysis in subsection 1 indicates that changes in the parallel market exchange rate exhibit predictable patterns, which in turn suggests the possibility of gaining some insight into the serial dependence of the changes in the logarithm of the parallel market exchange rate through regression analysis. 1/ The steady state value of the parallel market premium, which is given in equation (14), is determined by the proceeds from oil and non-oil exports, and by the output and the government consumption of the nontraded good.

Our efforts at empirically examining the relationship suggested by equation (14) is hampered since data on government expenditures and on the production of nontraded goods are available only on an annual basis. Hence, it was possible to conduct only a partial empirical examination using monthly series on oil and non-oil export earnings, which were available for the period January 1978-March 1990. Chart 4 illustrates both the decline and volatility in oil and non-oil export earnings in the period since the 1979 revolution. The autocorrelation tests reported above indicated that high order lags (lags higher than six periods) of the parallel market rate might have some explanatory power. Hence, lagged values of the premium were included in the regression equation. Lagged changes in the official exchange rate were also included in the regression in order to examine their impact on the parallel market premium; it would be expected that a devaluation of the official exchange rate might, on impact, lead to a decline in the parallel market premium, but that the premium would return to its previous level as steady state is regained.

Chart 4:
Chart 4:

Indices of Real Oil and Non-oil Export Receipts

(January 1978=100, deflated by import prices)

Citation: IMF Working Papers 1995, 069; 10.5089/9781451849035.001.A001

Preliminary examination of the order of integration of the real oil and non-oil export earnings indicates that these series were stationary, while the time series for the parallel market premium was nonstationary. Such differences in the order of integration, as well as the absence of data on a number of the variables in equation (14), precluded an empirical examination using cointegration tests. Consequently, the estimation was done simply in first differences. The overall fit of the regression results was not good:

Δlog(dt)=0.02(3.26)0.24(2.93)Δlog(dt6)0.03(2.04)Δlog(OXt1)0.02(1.84)Δlog(OXt2)+0.01(1.24)Δlog(NOXt1)+0.44(0.83)Δlog(et1)AdjR2=0.06,D.W=2.07,σ^=0.09,

where d is the parallel market premium, OX is real oil export receipts, NOX is real non-oil export receipts, e is the rial/dollar exchange rate, and σ is the standard error of the regression. The numbers in parentheses are t-statistics. While there is some evidence of the inverse relationship between oil exports and the premium, the coefficient of non-oil exports was of the wrong sign; however, it was statistically insignificant. The coefficient of changes in the official exchange rate was statistically insignificant. The sixth order lagged value of the change in the premium proved statistically significant; however, lower order lags of the same variable proved insignificant. Re-estimation of the above equation for shorter subperiods provided stronger support for the model, but the overall fit of the equation remained weak. The weakness of the results are likely to be due in part to the large volatility not only in the series on the parallel market premium, but also the series for oil and non-oil exports, as evidenced in Chart 4.

In addition to the paucity of data on pertinent variables, such as output of and government expenditures on nontraded goods, the poor overall fit of the equation may be due to several factors. First, the series for the parallel market exchange rate is very likely to contain considerable measurement error. Second, the poor statistical fit of the equation may be partly accounted for by differences in the calendars used for the measurement of the different variables. 1/ Third, the results might be due to the fact that the empirical analysis was done under the assumption that Iran has been operating under a system of dual exchange rates, when in fact the country operated under a multiple exchange rate system with frequent changes in the coverage of different exchange rates. Fourth, it is very likely that the movements in the parallel market exchange rate have been influenced by the numerous and significant political developments during the period under study. 2/

IV. Concluding Remarks

In this paper we provided a simple model for the determination of the ratio of the parallel to the official exchange rate in an oil exporting economy. After examining some of the empirical properties of the rial/dollar parallel market exchange rate, we examined econometrically developments in the parallel market premium. While the econometric results did not provide strong support for the model, there is some evidence for the role played by oil exports on developments in the parallel foreign exchange market. It is important to emphasize that the model presented in this paper has abstracted from the wealth effects of oil resources. If the government of an oil producing country can readily resort to external borrowing against its oil reserves, indeed some of the motivation for the rationing of imports and the other exchange controls that engender a parallel market for foreign exchange will be attenuated. The model presented in this paper could be extended to provide for a framework for the analysis of such considerations.

The Parallel Market for Foreign Exchange in an Oil Exporting Economy: The Case of Iran, 1978-1990
Author: Mr. Adnan Mazarei
  • View in gallery

    The Parallel Market Premium for the U.S. Dollar

    (In percent)

  • View in gallery

    Volatility of the Rial/Dollar Parallel Market Rate

    (percent monthly change: January 1978 - December 1990)

  • View in gallery

    Distribution of the Rate of Change in the Parallel Market Exchange Rate

    (In percent: Jan. 1978-Dec. 1990)

  • View in gallery

    Indices of Real Oil and Non-oil Export Receipts

    (January 1978=100, deflated by import prices)