Article

Demand for Money in Middle Eastern Countries

Author(s):
International Monetary Fund. Research Dept.
Published Date:
January 1980
Share
  • ShareShare
Show Summary Details

The demand for money has been perhaps the most comprehensively studied of all economic relationships over the past ten years. In addition to the natural academic interest in the determinants of demand for an important financial asset, research has been spurred by the operational importance of the relationship in circumstances where central banks have increasingly come to use rates of growth of monetary aggregates as proximate targets for policy. 1 Improving understanding of the demand for money is particularly important in the formulation of economic programs supported by assistance from the International Monetary Fund. Such programs typically involve performance criteria that set limits on the aggregate domestic lending of the banking system, either directly or through limits on the growth of the monetary base. The formulation of targets in this way is based implicitly on the assumption that there is a stable demand function for money. Given an objective or forecast for the rate of price increase, and an assumption concerning the potential of an economy for real economic growth, a certain rate of monetary expansion will then be consistent with the growth in demand for money implied by these assumptions. If a forecast or target for the likely balance of payments outturn is then added, the appropriate rate of domestic credit expansion can be determined. Credit expansion above or below this rate will cause a divergence of monetary growth from the planned rate, resulting in inflationary or deflationary pressures and tending to bring about a deterioration or improvement in the balance of payments. 2

It is recognized, of course, that the simple model sketched above is no more than a first approximation of complex economic forces that may be at work. In devising a financial program for an economy, modifications must therefore be made to take into account, as far as possible, the special circumstances of individual cases. In particular, it is important to refine the underlying estimate of the demand for money beyond simple (but often useful) heuristic assumptions such as that the demand for money grows in step with nominal income, that is, the assumption of constant velocity.

Since the revival of interest in the demand for money, which could perhaps be considered to have begun with Friedman’s (1956) restatement of the quantity theory, understanding of this basic relationship has improved in a number of respects. In financially sophisticated economies, refinements have been introduced to incorporate the influence of changes in economic structure, demographic shifts, the availability of a complex array of alternative financial assets, and the influence on money demand of different patterns of lagged adjustment to changes in its principal determinants. Understanding of the money demand relationship has also been broadened by studies undertaken for a variety of economies at different stages of development and with different economic structures.

At the risk of oversimplification, the body of evidence now available permits a number of generalizations. First, in many different circumstances a demand for money function can be estimated with sufficient stability to be operationally useful. Second, the two most important determinants of the demand for real money balances are a scale variable providing some measure of the volume of transactions in the economy and an opportunity cost variable measuring the return (expected or actual) on holding money relative to alternative assets, real or financial. Third, the coefficients of the various determinants of the demand for money vary between economies with different structural characteristics. Estimates of the determinants of the demand for money obtained from data for industrial countries cannot, therefore, be extrapolated with confidence for the purpose of making policy prescriptions in developing countries.

The main purpose of this study is to estimate money demand functions for all countries in the Middle Eastern Department of the Fund for which data are available. 3 It needs to be stressed, however, that an exercise of this kind has inevitable limitations. The available statistical estimates for a number of important variables in the demand for money function are generally less reliable than those for more developed economies. In a number of cases, time series are too short to permit a high degree of confidence in the resulting estimates. Furthermore, income data are usually available only on an annual basis, thus precluding the estimation of anything more than the most rudimentary lag structure. Also, many of the economies in the region have been going through a process of rapid structural change, which may well be reflected in changes in the underlying determinants of money demand.

The plan of this paper is as follows. Section I discusses a number of theoretical and practical problems involved in estimating the demand for money in the countries to be studied. Section II presents the basic model and provides individual country estimates for the estimated equations. Section III presents some results derived from pooling data across countries. Section IV discusses the statistical procedures employed and the interpretation of the results. Finally, Section V assesses some of the conclusions that can be drawn.

I. Theoretical and Practical Issues

It has already been noted that there is wide agreement that a demand for money function should contain a scale variable relating to the level of transactions in the economy and a variable representing the opportunity cost of holding money relative to alternative real or financial assets. The principal issues in constructing a demand for money function, therefore, relate to the definition of the money stock and the appropriate specification of the income and opportunity cost variables. In addition, it remains for consideration whether in particular cases other variables might systematically affect the demand for money, and whether provision should be made for the lagged adjustment of actual money holdings to the desired money stock.

DEFINITION OF MONEY STOCK

Neither theoretical considerations nor empirical evidence is conclusive in demonstrating whether a broad definition of the money stock (encompassing all the deposit liabilities of the banking system) or a narrow definition (covering only currency and demand deposits) is likely to be most stably related to the macroeconomic variables whose value it is desired to influence. It is generally accepted that the appropriate definition of the money supply should be that collection of assets among which substitutability is highest and which is most stably related to a small set of determining variables. It is sometimes held, however, that to be operationally useful a money stock definition should comprise an aggregate that the monetary authorities can adequately control. In financially developed economies, this second principle is sometimes adduced in support of a narrow definition of money (Ml), which tends to be more responsive to open market and interest rate policies. In many developing countries, however, available policy instruments apply principally to the volume of credit extended by the banking system, which would tend to make the total liabilities of the banking system (M2) easier to control than a particular component.

However, even if the monetary authorities are able to control M2 better than Ml, in the short-to-medium term, developments in Ml could still be a useful guide to the impact of monetary policy in circumstances where an empirically stable relationship between Ml and total output had been established. Accordingly, it seems desirable not to prejudge the issue of which definition of the money supply is likely to be the most appropriate for the countries under study.

There are also a number of other issues relating to the definition of the money stock that warrant prior discussion. In a number of the countries, the public sector is large and extends into manufacturing and other industries. The conventional definition of the money supply does not include deposits of central and local governments or of entities under direct government control but does include balances held by public economic entities with their own financial resources and budgets separate from those of the central government. However, the distinction between public sector entities along the above lines should depend on the behavior of the entities concerned rather than on their legal or accounting status. 4 In some instances, it may be that deposits of public enterprises should be excluded from the money supply statistics. So, for those countries where the data permit, it seems desirable to conduct tests both including and excluding deposits of public sector entities outside government.

Another question surrounds the treatment of balances held in foreign currency. Conventionally, foreign currency balances held by residents with domestic banks are treated as part of the money supply, since, given a fixed exchange rate and convertibility into domestic currency, they contribute to the liquidity of the holder no less than do balances of domestic currency. On the other hand, foreign currency balances may be held for somewhat different purposes than local currency balances, 5 so that it may be worthwhile to conduct experiments, where possible, with the exclusion of foreign currency deposits from the money supply. 6

scale variable

Although a variety of techniques have been suggested to enable the demand for money to be regressed against a wealth variable (such as using permanent income), the available data are neither of sufficient quality nor in sufficiently long time series to permit these techniques to be applied for the countries being considered here. Relying simply on income variables, the aggregate that is most commonly used in studies for other countries is gross domestic product (GDP). For several reasons, however, this may not be the best measure of transactions in the economies of a number of countries in the region. Many have important oil sectors, which in a number of cases account for as much as half of total GDP. Decisions on the volume of oil that is produced and on its price neither affect nor are affected by monetary creation, and there is no direct effect on the liquidity of the private sector. The oil sector of these countries is perhaps best regarded as an “enclave” within the non-oil economy, and is best disregarded when constructing a time series of income that is considered to influence the demand for money.

Another consideration is the treatment of remittances by expatriate workers. These have grown rapidly in recent years, representing as much as one half of non-oil GDP in one recipient country (the Yemen Arab Republic) and constituting a significant part of GDP in the countries where the workers reside. Conventionally, GDP is a measure of output within a given economy, so that the earnings of expatriate workers constitute part of the GDP of the countries in which they reside and not of their countries of origin.

To decide whether this is an appropriate definition of the scale variable influencing the demand for money, one needs to consider the reason for holding money balances. If it is to finance transactions taking place in the domestic economy, then it is the total income or wealth available to domestic residents that is the relevant determining factor for money demand. In the case of workers’ remittances, in most of the countries being studied, funds are normally sent back by workers to their families, who remain in the countries of origin. These funds then become part of normal income receipts, against which transactions balances are held. In consideration of this, it seems appropriate to define national income for purposes of estimating the demand for money as including remittance receipts.

Where official national income statistics are nonexistent or known to be unreliable, some alternative must be found. In these instances, we have relied on Fund staff estimates of output generated on the basis of qualitative and quantitative information gained in the process of preparing staff consultation reports, and on estimates provided in World Bank reports. 7

opportunity cost variable

An opportunity cost variable in a demand for money function is intended to measure the yield on money against other assets that might be held. In financially developed economies, this variable is usually an interest rate, though there is no general agreement on which particular interest rate should be used. Furthermore, it is also often argued that inventories of real assets are an alternative form in which wealth can be held, and, hence, that the expected inflation rate should enter as a determinant of money demand. Among developing countries, it is quite widely accepted that an interest rate is in practice an unsatisfactory measure of the opportunity cost of holding money. In the first place, financial markets outside the banking system are not well developed, so that the possibilities of substitution between money and other financial assets are limited. Second, a more practical objection is that observed interest rates are often centrally determined and remain unchanged for long periods. Thus, there is often insufficient variation in this interest rate to enable its influence on the demand for money to be estimated with confidence. Furthermore, in certain of the countries under study the payment of interest is legally prohibited.

For both theoretical and practical reasons, therefore, it seems appropriate to estimate money demand functions using a measure of expected inflation as the opportunity cost of holding money. Instead of experimenting extensively with different distributed lag proxies for expected inflation, we have (arbitrarily) assumed that the expected inflation rate equals the actual rate. This was done in order to limit the already substantial quantity of empirical work to be conducted and may be a legitimate approximation in view of the fact that annual data are being employed.

The next question is which measure of price inflation should be used. The GDP deflator is the most general measure of prices within an economy but may be inappropriate for a number of reasons. First, the aggregate GDP deflator reflects changes in the price of oil exports, and, as argued already, the oil sector should be regarded as essentially an enclave within the non-oil economy. Second, the GDP deflator captures changes in the price level of domestic output only. Since we are using the inflation rate as a measure of the opportunity cost of holding money compared with buying goods, the inflation measure should also include changes in the prices of imported goods. For this reason, the consumer price index (CPI) may be a theoretically preferable construct to the GDP deflator as a measure of the opportunity cost of holding money balances.

The CPI, however, is itself an imperfect measure in many of the countries under study. The consumption baskets used often reflect the purchasing patterns of only a particular income segment and may employ data relating only to the capital city. A further difficulty is introduced by the practice in a number of countries of subsidizing the purchase of commodities that play a key role in the CPI. Although an accurate consumer price index might provide inflation rate estimates that for our purposes would be preferable to those from a GDP deflator, the quality of the data used to generate the CPI may be poor enough to counteract this advantage. Consequently, we have used inflation rates from both the CPI and the GDP deflator (or non-oil GDP deflator for the oil exporting countries) in our empirical work, with the ultimate choice resting on the empirical results.

other variables

Since the purpose of the present study includes cross-country comparisons, and is intended to pool data across countries, it is desirable not to include country-specific determinants of money demand. Where the number of observations is small, there is the further consideration that statistical fits can be artificially improved by the inclusion of additional variables, without necessarily improving the forecasting properties of the model or its usefulness as a guide to policy prescription. Since much previous work on the demand for money has discovered important lagged responses, it seems desirable to conduct experiments with simple lag structures.

Before proceeding to the results of the statistical estimates, it is worth discussing the results that the estimated coefficients might be expected to have, on the basis of theoretical considerations and empirical work performed for other countries. Concerning the income elasticity of the demand for money, there are two competing theoretical propositions. The first is that the use of money balances is subject to economies of scale, and, therefore, that as real income increases the observed velocity of circulation will increase. On these grounds, an income elasticity of demand for money somewhat less than unity could be expected. The second proposition is that money is a “luxury” good and that both the desire and the capacity to hold liquid balances increases more than proportionately with income and wealth. A third, and related, proposition is that the gradual absorption of the nonmarket sector into the market sector of a developing economy results in the demand for real balances increasing more than proportionately with real income; that is, the estimated income elasticity of the demand for money may be greater than unity because of the upward bias induced by monetization.

The available statistical evidence from other studies permits two tentative generalizations to be made. First, transactions balances, defined as Ml, appear to be more subject to economies of scale than money broadly defined. That is, income elasticities tend to be lower for a narrow definition of money than for a broad one. Second, broad money appears to be more of a luxury good in developing economies than among countries whose economic and financial systems are more developed. While observed income elasticities in developed economies are sometimes above and sometimes below unity, developing countries tend to exhibit an income elasticity closer to 1.5. In a study of six Asian economies, Aghevli, Khan, Narvekar, and Short (1979) found elasticities ranging from 1.33 to 1.85 for a broad definition of money. Mackenzie (1979), in a study of money demand in Egypt, found elasticities for broad money of about 1.5. Morgan’s (1979) estimates for five oil exporting countries (including two in the Middle East), incorporating a Koyck-type lag structure, were of long-run income elasticities ranging from 1.41 to 1.82. However, Galbis (1979), in a study of nineteen Latin American countries, found elasticities generally well below unity though dispersed over quite a wide range, which is perhaps not surprising in view of the more disturbed monetary experience of a number of Latin American countries. 8

As far as the influence of opportunity cost variables is concerned, theory suggests that an increase in the expected rate of inflation would reduce the attractiveness of money balances. This effect should be more pronounced for narrow money, which conventionally has a zero nominal yield, than for broad money, which includes time and savings deposits, whose yield can be adjusted to offset inflationary expectations. Empirical work on developing countries has been less successful in discovering significant and stable coefficients for inflation elasticities than for income elasticities. In the study by Galbis (1979), only sporadic evidence of significant negative inflation elasticities was found; while Morgan (1979), despite experiments with various specifications of the expected price change variables, was unable to discover any significant relationship with the appropriate sign. Therefore, while the expected sign of the inflation coefficient is negative, there can be no strong presumption about the value it might be expected to assume. 9

Empirical work on developed countries has established quite clearly that the demand for money responds with a lag to changes in its underlying determinants. On a priori grounds, there seems little reason to doubt that a similar pattern of lagged response exists in developing countries. However, it is more difficult to capture lagged patterns of response where only annual data are available, and what evidence there is from previous statistical studies on developing countries suggests that most of the adjustment takes place in the first year.

II. Empirical Results

The basic estimated model is as follows:

where M = money stock, P = price level, P-1 = price level in previous period, and Y = income.

The model is estimated in double logarithmic form, so that the coefficients may be directly interpreted as elasticities. The specification is in constant price terms to reflect the constraint imposed by economic theory that the demand for money not be subject to money illusion (i.e., the elasticity of demand for nominal money balances with respect to changes in the price level is assumed to be unitary). This procedure has the advantage of limiting the degree of spurious correlation introduced by common inflationary trends in both dependent and independent variables.

In all cases the model was estimated for as long as permitted by the data, except for Lebanon and Ethiopia, for which the estimates presented used data truncated to exclude the period of military disturbances. The precise definition of the variables employed varies somewhat from country to country as a result of taking into account the considerations discussed in more detail in the previous section. The experiments conducted with excluding balances of public sector enterprises and/or foreign currency deposits did not appear to affect the results substantially. Where it seemed desirable to make such adjustments and where the data permitted, the sample size was considerably shortened, and the results differed little from those using the longer unadjusted series. Consequently, all results reported here are in terms of the conventional definitions of the money supply as employed by the Fund’s International Financial Statistics (IFS). Adjustments to the scale variable, however, either by removing the influence of the oil sector or by including the effects of factor payments from abroad, were much more useful.

Non-oil GDP performed substantially better than did total GDP for the oil exporting countries, and real gross national product (GNP), that is, including remittances, performed better than did real GDP in those countries (principally Afghanistan, the Yemen Arab Republic, and the People’s Democratic Republic of Yemen) where remittances play an important role.

Complete country-by-country estimates are provided in the Appendix. In general, the equations for broad money are better determined than those for narrow money, a result which may be attributable to a growing substitution from demand deposits to time and savings deposits. For all but one of the countries, the equations are reasonably well determined in the sense of having high explanatory power (R¯2 close to or higher than 0.8, and in many cases 0.9), reasonably small standard errors of estimate (generally lower than 10 per cent, often around 5 per cent), and only limited evidence of serial correlation—although such inferences are difficult in small samples.

Given the broadly satisfactory statistical properties of the equations, attention can be focused on the coefficients of the independent variables. As was to be expected, the coefficient on income turned out to be highly significant in nearly all cases, with the notable exception of Pakistan (Table 1). Although the coefficient estimates vary quite widely, they are generally between 1 and 2, with the largest number in the range of 1.00 to 1.50 (Table 2).

Table 1.Summary of Income Elasticities
CountryIncome Elasticity of Narrow Money95 Per Cent Confidence IntervalIncome Elasticity of Broad Money95 Per Cent Confidence IntervalSample Period
Afghanistan1.629920.898 to 2.3621.731831.018 to 2.4451969-77
Bahrain0.73700-0.215 to 1.6891.148210.499 to 1.7981973-78
Egypt1.674031.261 to 2.0881.828801.386 to 2.2721965-78
Ethiopia0.9283290.672 to 1.1841.535181.399 to 1.6721962-73
Iran1.144930.968 to 1.3221.442331.234 to 1.6501960-78
Iraq1.364831.252 to 1.4771.462561.355 to 1.5701965-77
Jordan0.8718610.635 to 1.1091.099320.843 to 1.3561967-78
Kuwait1.298190.921 to 1.6761.4110111.219 to 1.6011972-78
Lebanon0.558790.303 to 0.8141.284351.220 to 1.4601965-73
Libyan Arab Jamahiriya1.183350.961 to 1.4061.339740.931 to 1.5191970-78
Oman1.082570.524 to 1.6410.9179720.175 to 1.6611973-78
Pakistan1.02512-1.868 to 3.9181.37241-1.438 to 4.1831974-78
Qatar0.9479020.737 to 1.1590.8460080.588 to 1.1041973-78
Saudi Arabia1.485411.377 to 1.5941.310001.248 to 1.3721968-78
Sudan1.298341.018 to 1.5791.478291.160 to 1.7971964-77
Syrian Arab Republic1.209570.977 to 1.4431.236190.998 to 1.4751970-77
United Arab Emirates1.725960.438 to 3.0141.21289-0.895 to 3.3201975-78
Yemen Arab Republic2.200701.385 to 3.0172.199321.411 to 2.9881974-78
Yemen, People’s Dem. Rep. of1.083880.103 to 2.2061.187380.034 to 2.3411973-76
Table 2.Estimated Income Elasticities of Demand for Money(Number of countries)
< 1.001.00-1.251.25-1.501.50-1.751.75-2.00> 2.00
M156431
M2258211

Given the limited number of observations, and the general unreliability of the basic data, too much significance should not be attached to the point estimates of income elasticity. It is interesting to note that, in addition to the ten countries where the point estimate of income elasticity of demand for narrow money falls in the range of 1.00 to 1.50, a further eight countries have estimates that fall in this range at the 95 per cent confidence interval. That is, only one country (Lebanon) has an income elasticity whose 95 per cent confidence interval lies entirely outside the range of 1.00 to 1.50. For broad money, only one country (Qatar) exhibits an income elasticity that is outside the range of 1.25 to 1.50 at the 95 per cent confidence interval.

The countries that exhibit “abnormal” income elasticities (i.e., point estimates outside the range of 1.00 to 2.00) are Qatar and the Yemen Arab Republic (for both broad and narrow money), Bahrain, Ethiopia, Jordan, and Lebanon (for narrow money only), and Oman (for broad money only). In none of these cases is the data base satisfactory. For Oman, Qatar, and Jordan, the dependent and scale variables were deflated by price indices of dubious quality because of the absence of a GDP deflator. Preliminary regressions on other countries (not reported here) using a price index instead of a national accounts deflator showed that the estimated income elasticity varied enormously from one procedure to the other. Of course, if it is true that a consumer price index is an unsatisfactory deflator, this casts doubt on the results achieved for Sudan and the People’s Democratic Republic of Yemen, where a consumer price index was also used as a deflator.

For the Yemen Arab Republic, the financial system was until recently at a very early stage of development, and the non-monetized sector was large. Under the influence of large remittances from abroad, it is reasonable to expect that the process of monetization would be more rapid than in countries where rising levels of income derive from domestic development. Thus, it is perhaps not surprising that the estimated income elasticity of demand for money exceeds 2. Confirming this interpretation is the fact that the other country in the region where remittances dominate economic developments (Afghanistan) records an income elasticity of money demand of about 1.7, a high figure compared with the estimates for other countries.

Bahrain, Ethiopia, and Lebanon all have puzzlingly low income coefficients in the demand for narrow money equations; however, since in each case the corresponding broad money equation gives a result which is more in conformity with those obtained for other countries, the explanation may lie in special factors affecting the substitution between broad and narrow money.

Three experiments with lag structures were conducted. 10 One included current and lagged real GDP (or non-oil GDP or GNP) as explanatory variables. In general, the results closely resembled those reported in Table 1, the sum of the two coefficients being in all cases close to the single income elasticity reported there. However, the explanatory power of the equation (as measured by the R¯2 or adjusted SEE) was reduced. Experiments were conducted with the Koyck transformation as a means of introducing the lagged dependent variable. The results had better fit, but the implied long-run income elasticities of money demand were somewhat more volatile. White (1978) has suggested that the stock adjustment transformation should be applied to nominal rather than real balances (i.e., to use In (M-1/P) instead of the lagged dependent variable). Experiments along these lines also produced better overall fit, but the implied income elasticities were also more volatile and in some cases assumed implausible values. Both of these approaches to capturing the effects of lagged adjustment suffer from the disadvantage that the measured adjustment coefficient may be reflecting an autoregressive structure rather than genuine lags. That this may be the case is suggested by the fact that estimated elasticities are more closely clustered around a central value using a simple (unlagged) specification than when a lag mechanism is introduced.

Concerning the influence of inflation on the demand for money, the results are much less clear-cut than for income elasticities. When perverse (positive) signs were obtained, the equations were rerun with the opportunity cost variable excluded. In some cases, the coefficient was negative and significant but implausibly large (in absolute value). 11 Generally, these results coincide with an implausible income elasticity and a very short sample. So, in these cases, the equations were rerun excluding the opportunity cost variable, resulting in income elasticity estimates that were more plausible on a priori grounds. Of the equations reported, only three countries (Afghanistan, Iran, and Saudi Arabia) include opportunity cost variables (Table 3).

Table 3.Summary of Opportunity Cost Coefficients
CountryOpportunity Cost Coefficient of Narrow Money95 Per Cent Confidence IntervalOpportunity Cost Coefficient of Broad Money95 Per Cent Confidence Interval
Afghanistan–0.398271–1.304 to +0.508–0.469023–1.352 to +0.414
Iran–0.705891–1.734 to +0.322–0.696293–1.499 to +0.107
Saudi Arabia–0.266592–0.489 to –0.044–0.166877–0.363 to +0.030

In a further attempt to capture the cost of holding real balances, equations were estimated for all countries with the London Eurodollar rate of interest as an explanatory variable. The argument is essentially that Eurodollar deposits may be a relevant alternative to holding real balances, at least for the business sector of Middle Eastern countries. However, the results were poor, with the Eurodollar interest rate coefficient taking a perverse (positive) sign more often than not, and never proving to be both negative and significant for any country for either narrow or broad money.

These results are perhaps disappointing but are similar to other studies of money demand in nonindustrial countries which have also had difficulty uncovering significant opportunity cost effects. It may well be that in the absence of liquid asset alternatives to holding money for much of the population in many of these countries, the opportunity cost effect is indeed insignificant.

III. Pooled Cross-Section and Time-Series Data

One of the shortcomings of the estimates presented in the previous section (and in more detail in the Appendix) is the shortness of the time series available for most countries. As one means of overcoming this limitation, data may be pooled across countries to provide a greater number of observations. One of the requirements of such a procedure is that the data being pooled come from a homogeneous underlying population.

To attempt to meet this requirement, several transformations were performed on the data, and a priori choices of country groupings were made. The transformations involved converting monetary and income data to constant dollar terms, to eliminate unit-of-account differences, and deflating individual country variables by population estimates. This second adjustment, which amounts to estimating the demand for money in per capita terms, is necessary to eliminate income differences that derive from country size rather than from income levels. For the country groupings, estimations were performed pooling data for all countries in the sample and separating major oil exporting countries from the others. (The precise lists of countries used in the pooling exercise are provided in the footnote to Table 4.)

Table 4.Income Elasticities: Cross-Section Estimates
Country Grouping 1Sample PeriodNarrow Money ElasticityBroad Money Elasticity
Group A1974-780.8550.895
Group B1974-780.7940.944
Group C1974-780.8730.925
Group D1970-780.9310.961
Group D1970-731.0951.155
Group D1973-780.9180.973
Group E1970-780.9030.938
Group E1970-730.8490.887
Group E1973-780.6900,946

The country groupings are

Group A: Afghanistan, Bahrain, Egypt, Ethiopia, Iran, Iraq, Jordan, Kuwait, Lebanon, Libyan Arab Jamahiriya, Oman, Pakistan, Qatar, Saudi Arabia, Sudan, the Syrian Arab Republic, the United Arab Emirates, and the Yemen Arab Republic.

Group B: Eight oil exporting countries—Iran, Iraq, Kuwait, Libyan Arab Jamahiriya, Oman, Qatar, Saudi Arabia, and United Arab Emirates.

Group C: Ten non-oil countries—Afghanistan, Bahrain, Egypt, Ethiopia, Jordan, Lebanon, Pakistan, Sudan, the Syrian Arab Republic, and the Yemen Arab Republic.

Group D: Four oil exporting countries with longer series—Iran, Iraq, Libyan Arab Jamahiriya, and Saudi Arabia.

Group E: Six non-oil countries with longer series—Afghanistan, Egypt, Ethiopia, Jordan, Lebanon, and Sudan.

The country groupings are

Group A: Afghanistan, Bahrain, Egypt, Ethiopia, Iran, Iraq, Jordan, Kuwait, Lebanon, Libyan Arab Jamahiriya, Oman, Pakistan, Qatar, Saudi Arabia, Sudan, the Syrian Arab Republic, the United Arab Emirates, and the Yemen Arab Republic.

Group B: Eight oil exporting countries—Iran, Iraq, Kuwait, Libyan Arab Jamahiriya, Oman, Qatar, Saudi Arabia, and United Arab Emirates.

Group C: Ten non-oil countries—Afghanistan, Bahrain, Egypt, Ethiopia, Jordan, Lebanon, Pakistan, Sudan, the Syrian Arab Republic, and the Yemen Arab Republic.

Group D: Four oil exporting countries with longer series—Iran, Iraq, Libyan Arab Jamahiriya, and Saudi Arabia.

Group E: Six non-oil countries with longer series—Afghanistan, Egypt, Ethiopia, Jordan, Lebanon, and Sudan.

In no case did the opportunity cost (inflation) variable enter the equations with a sign that was appropriate and significant, and the equations were rerun using only a real income variable. In all cases, the coefficient on this variable was highly significant, and the other statistical properties of the estimated equations were satisfactory. A summary of the results is provided in Table 4, and more complete information on the estimates is given in the Appendix.

The puzzling feature of these results is that the estimated income elasticities were all distinctly lower than for the individual countries. For narrow money, the estimated income elasticity coefficients were generally in the range of 0.8 to 0.9, and, for broad money, in the range of 0.9 to 1.00. This divergence between individual and pooled results casts doubt on the pooling procedure and suggests that the countries that were grouped together may not be homogeneous enough to permit pooling.

A possible explanation may lie in the process of monetization. If differences in per capita income among countries are broadly associated with proportionate differences in per capita money balances, then the observed growth in the money/income ratio for individual countries has to be explained by increasing monetization. Put another way, in the pooled cross-section and time-series estimates, the cross-sectional (i.e., between countries) variation may be more important than the time-series (i.e., within countries) variation, with the result that the pooled coefficient estimate may more closely resemble a “pure” income elasticity of money demand, without a monetization factor.

The fact that the individual country estimates are all quite similar further suggests that monetization is proceeding at a similar pace in all the countries under study. 12 At first sight this seems strange, since it is to be expected that monetization would be more rapid in the least developed countries, where the financial system had the greatest scope for expansion. However, it is possible that technical progress in financial intermediation has led to a reduction in its relative cost in all countries, and therefore in increased use of the banking system.

The significant discrepancy between the individual country and pooled results raises the question of which estimates are more relevant for policy analysis and in the formulation of financial programs. The answer is that it should be a figure comprising both a pure income elasticity and a monetization factor, since both contribute to the observed increase in money demand. It may be, however, that the two components should be kept conceptually distinct. While year-to-year fluctuations in the rate of income growth are likely to be associated with corresponding fluctuations in money demand, monetization is likely to be more of a trend factor, relatively independent of cyclical fluctuations. Thus, where real income growth is below its long-run trend, velocity may tend to decline by more than is implied by the individual country income elasticity coefficients; where real income is growing more rapidly than the long-run average, velocity will tend to decline by less.

IV. Statistical Notes

individual country estimates

The dependent variable in all regressions was an annual (five-quarter) average of money supply statistics, obtained from IFS and deflated by the implicit GDP (or non-oil GDP or GNP) deflator, or a price index. The beginning and end points for the averaging procedure were determined on the basis of the fiscal year used in the calculation of income figures. The independent scale variable was real GDP in about one half of the countries but was defined to exclude the oil sector for countries where this was a major component of national income. GNP was the scale variable in countries where remittance inflows are important. Nominal income levels were deflated by the associated national income deflator where this was available or by the consumer price index where it was not.

Statistical procedures and statistical inference are difficult in small samples, and some caveats should be mentioned. All equations were estimated by ordinary least-squares, with a Cochrane-Orcutt first-order correction for autocorrelation when necessary. Obviously, one would not wish to maintain that real GDP and the inflation rate are exogenous with respect to real balances; hence, the usual problems of simultaneous-equation bias and inconsistency arise. Also, it may be that the opportunity cost of holding money is a significant influence, which we have been unable to capture (in many cases) because of the inadequacy of the data. This would introduce further potential bias.

A further difficulty is introduced by the problem of serial correlation. The behavior of the Durbin-Watson statistic is not well established for very small samples (less than fifteen observations), so that it was used as a statistic on a purely heuristic basis. Although serial correlation would not of itself lead to bias in the coefficient estimates, it would produce bias in the estimated standard errors, so that some of the confidence intervals for income elasticity coefficients (particularly for Ethiopian narrow money and Libyan narrow money) may be too narrow.

The stability tests conducted were of the type outlined by Chow (1960) and Fisher (1970). The test was restricted to the stability of the income elasticity coefficient. First, a regression was run for the full available sample in the usual way, and then one was run with the scale variable up to a breakpoint (usually up to end-1973) as one explanatory variable, and the scale variable from then on, as another. Then an F-test was constructed in the usual way, using the sums of squares of residuals from each regression. The difficulties with this procedure are, first, that it relies on an arbitrary choice of the breakpoint, and, second, that it is not powerful when the sample is very short. Nevertheless, the results are useful, and the conclusion that for all of the oil exporting countries the income elasticity of money demand is stable between the pre-1974 and post-1974 periods is indeed interesting.

pooled cross-section and time-series equations

To enable comparability of the series used in pooling, the monetary and income data for each country were converted to a common constant price scale and deflated by estimated population in 1975. It should be noted that a certain degree of non-comparability between countries still exists after these transformations. For some countries the scale variable is real GDP, for others real GNP, and for still others real non-oil GDP. The price index may be a GDP deflator, a non-oil GDP deflator, or, for countries for which constant price national accounts are not available (Jordan, Oman, Qatar, and Sudan), an index of consumer prices.

The statistical procedures employed assumed that the underlying structure was cross-sectionally heteroskedastic and autoregressive over time. First, the model was estimated using ordinary least-squares, and a separate serial correlation coefficient was calculated for the observations from each country. These were used to apply a Cochrane-Orcutt adjustment to the data, whereupon the equation was re-estimated. This procedure was iterated until the coefficients converged. Then the residual standard deviation was estimated for each country, and the (Cochrane-Orcutt) adjusted data were divided by these standard deviations to remove the influence of heteroskedasticity. 13 Then a final regression was run to provide the results reported here.

Inspection of the results at the intermediate stages (prior to the autocorrelation adjustment, and after the autocorrelation adjustment but before the adjustment for heteroskedasticity) suggested that the adjustments both improved the goodness-of-fit and resulted in a more sensible coefficient estimate. However, interpretation of the summary statistics is more difficult. The R2 refers to the adjusted data, not the original; the SEE is difficult to interpret because the measurement units are not well defined (each observation has been divided by a country-specific residual standard deviation). The Durbin-Watson statistic is not reported, since it has no meaning in a cross-section context—its value will vary with the (unimportant and arbitrarily chosen) ordering of countries.

V. Conclusions

This study has presented estimates of demand for money functions for the nineteen countries in the Middle Eastern Department of the Fund. Comprehensive coverage has been attempted not only to obtain the maximum number of individual country estimates but also to permit comparisons among countries and to provide some indication of the robustness of the estimates for individual countries. In consideration of this latter purpose, a relatively simple specification has been proposed, which can be estimated consistently for all countries (with, however, some minor differentiation in the precise definition of the variables employed).

The statistical properties of the estimated equations varied widely, as was to be expected. In only one case, however (that of Pakistan), were the results clearly unsatisfactory, in that it proved impossible to estimate an equation with significant explanatory power. This may be due in part to the fact that a relatively short time series is available following the secession of Bangladesh, and that adjustments following this event may have taken time to work themselves out before a stable money demand relationship could be re-established. For all other countries, however, a reasonably well-determined equation was estimated, with an income elasticity of demand for money that was statistically significant and within the range of a priori expectations.

Perhaps more significant than the properties of the equations for individual countries is the common pattern that seems to emerge from the study as a whole. The great majority of countries exhibit an income elasticity in the range of 1.00 to 1.50, and, even where the point estimate falls outside this range, the results are consistent with the assumption of an elasticity within the range. On the whole, the estimates for equations using the broad definition of money are better determined than those using narrow money. For broad money, roughly one half the countries have a point estimate for income elasticity in the range of 1.25 to 1.50, and all but one of the others have an elasticity in this range within the 95 per cent confidence interval.

The influence of inflation on the demand for money proved difficult to detect. Only in three cases out of nineteen was the inflation variable significant and of the correct sign. Although these results cannot be taken to contradict the hypothesis that the demand for money is sensitive to expected inflation, they do not provide any firm basis for conclusions about its quantitative importance.

The evidence from pooling data is at variance with the conclusions suggested by the estimates for individual countries. Income elasticities are generally much lower, in the range of 0.8 to 0.9 for narrow money and of 0.9 to 1.00 for broad money. It is tentatively suggested that the explanation for this may lie in a common trend to increased financial intermediation across all countries in the region. If this is true, it suggests that the observed income elasticity for individual countries may be composed of a “pure” income elasticity close to unity and an independent “monetization” factor. Attempts to distinguish these two factors, say by the inclusion of a trend variable, are not generally satisfactory because the trend factor tends to dominate other variables. In analyzing developments in individual countries, however, and particularly in making policy prescriptions, it needs to be borne in mind that the income elasticities may be reflecting a combination of two separate phenomena.

Finally, it needs to be re-emphasized that a statistical exercise such as the one reported in this paper is subject to a number of limitations. The data employed are frequently of poor quality (particularly the national accounts data), and the time series are generally short. Tests for stability of the resulting estimates are, therefore, of doubtful reliability. Autocorrelation may be a problem, and, therefore, the reported standard errors may be underestimated. In any case, the residuals that emerge from the equations are large enough that precise guides to policy could not be deduced.

Nevertheless, the high degree of similarity between the experience of different countries in the region suggests that as an operating guide it would be reasonable to assume that the income elasticity of demand for broad money probably falls in the range of 1.25 to 1.50.

APPENDIX: Data Sources and Definitions and Estimation Results

NOTATION

M1=narrow money (IFS definition)
M2=broad money (IFS definition)
GDPR=real gross domestic product
GNPR=real gross national product
NOGDPR=real non-oil gross domestic product
NOGNPC=current price non-oil gross national product
NOGDPC=current price non-oil gross domestic product
GDPC=current price gross domestic product
GNPC=current price gross national product
X=total exports of goods
Imp=total imports of goods
Oil X=oil exports
PGDP=implicit GDP deflator
NOPGDP=implicit non-oil GDP deflator
CPI=consumer price index
P=index of import prices
In=natural logarithm
R¯2=coefficient of determination, adjusted for degrees of freedom
SEE=standard error of estimate
D-W=Durbin-Watson statistic
N=Population
POD=post office deposits
PM=an index of imported prices t-statistics are in parentheses

INDIVIDUAL COUNTRY RESULTS

afghanistan

Narrow money

Broad money

Money supply observations for 1978 were obtained by extrapolating the growth rate for the first ten months to obtain December 1978 and March 1979 observations, which were then averaged in the way outlined in the body of the paper. The real GNP data were obtained from a World Bank report on Afghanistan, with the final observation being derived from Fund staff estimates. The current price GDP data (needed to obtain the GDP deflator) were obtained from the Fund’s Bureau of Statistics, with the final two observations constructed using the growth rates of real GNP and the CPI. All other data were obtained from IFS.

The overall goodness-of-fit of the equations is reasonable, with the estimated income elasticities sensible, especially in light of the substantial importance of remittances from abroad in this economy. Tests of stability of the income elasticities between 1969-75 and 1975-77 revealed no evidence of instability at the 95 per cent significance level. Money series from 1972-78, with deposits of public sector corporations removed, were available. Regressions from this shorter but better-quality series supported the estimates reported above.

bahrain

Narrow money

Broad money

The money supply series were obtained from IFS. National accounts data for Bahrain were obtained from a World Bank report. The non-oil GDP series were derived by subtracting from total GDP the value added in the mining and refining of oil, with the final observation in these series being constructed from Fund staff estimates.

The narrow money equation is poorly determined, perhaps because of substitution toward quasi-money over the sample period. The latter rose from 40 per cent to 55 per cent of total broad money in that time. The broad money equation is more satisfactory and has a more reasonable income elasticity coefficient. When the sample period was split at 1973, evidence of instability was not significant at the 95 per cent confidence level. Other regressions using real non-oil GNP as the scale variable were less successful. Inclusion of an opportunity cost variable gave a better overall fit but produced implausibly low income elasticities and implausibly high opportunity cost elasticities.

egypt

Narrow money

Broad money

Broad money plus post office deposits

All money supply series were obtained from IFS. The GDP series are also from IFS, with the 1978 current price GDP and the 1975-78 real GDP figures coming from Fund staff estimates. Series on nongovernment public deposits and foreign currency deposits (used in equations not reported here) were obtained from estimates by Mackenzie (1979).

The results achieved are consistent with those reported by Mackenzie. At no time did an opportunity cost variable enter significantly. When the sample period was split at 1973, there was evidence of instability in the estimated coefficients, with the results indicating a significant tendency for the income elasticity of money demand to rise over the sample period. This may reflect a gradually increasing pace of monetization or, as suggested by Mackenzie, an increasing understatement in the official statistics of the underlying inflation rate, or simply a shift in the nature of the function.

ethiopia

Narrow money

Broad money

All series used in the estimations are from IFS. Equations estimated on the full sample period were not worth reporting, the income elasticity being negative in some instances. The explanation for this may lie in the disturbances of the post-1974 period, something that is supported by the evidence of instability that appeared when the sample period was split at 1973. Thus, the results presented here use a sample period truncated at 1973. Measures of opportunity cost, being insignificant, were discarded.

iran

Narrow money

Broad money

The money supply series were obtained from IFS, with the 1978 observations being derived by extrapolating growth for the first seven months of 1978 and then averaging in the usual way. The non-oil GDP series were obtained from Economic Trends of Iran (March 1978), with the 1975-78 figures from Fund staff estimates.

Both equations are well determined, and the inflation coefficient is significantly different from zero at the 90 per cent confidence interval in the case of broad money. A stability test showed no evidence of instability (at the 95 per cent confidence level) when the sample period was split at 1973.

iraq

Narrow money

Broad money

The money supply series were obtained from IFS. The non-oil GDP series for 1972-78 were obtained from Fund staff estimates. The non-oil real GDP series for 1965-71 were generated by multiplying real GDP for those years by the average ratio of real non-oil GDP to total GDP for the years 1972 to 1975. The current price non-oil GDP data for 1965-71 were obtained by multiplying current price total GDP for those years by the average ratio of the two in 1972 and 1973. The resulting equations are well determined, and there was no evidence of instability at the 95 per cent confidence level.

jordan

Narrow money

Broad money

The basic data were obtained from IFS. However, since no constant price national accounts data are available for Jordan, both the scale and the dependent variables were deflated by the CPI as reported in IFS. Current price GNP deflated by the CPI performed substantially better than did current price GDP similarly deflated. The 1978 observation was based on Fund staff estimates. There was no evidence of instability at the 95 per cent confidence level between the subperiods 1967-73 and 1974-78.

kuwait

Narrow money

Broad money

Both money supply series are from IFS. The non-oil GDP series were obtained from Fund staff estimates. Stability tests did not indicate evidence of instability when the series was split at 1973.

lebanon

Narrow money

Broad money

The money supply series were obtained from IFS, and the GDP series from the Fund’s Bureau of Statistics, with more recent observations (not used in the equations reported) obtained from Fund staff estimates.

Equations estimated including data up to 1978 had implausible results, probably because of the war-torn conditions in recent years. Splitting the sample period at 1973 revealed strong instability in the estimates between the two periods, so that the estimates presented are based on time series truncated at 1973. The broad money equation performed better than that for narrow money, probably because quasi-money rose from about one half to more than two thirds of total broad money between 1965 and 1973.

libyan arab jamahiriya

Narrow money

Broad money

The money supply series were obtained from IFS. The non-oil GDP series were obtained from Fund staff estimates. There was no evidence of instability at the 95 per cent confidence level between 1970-73 and 1974-78. An earlier equation for narrow money is not reported, despite better overall fit, because the significant negative opportunity cost coefficient seemed implausibly large at -1.78.

oman

Narrow money

Broad money

All money supply series were taken from IFS. The non-oil GDP series were obtained from a World Bank report. The price series are a weighted average of Oman’s partner country export prices, converted from U.S. dollars into rials Omani, and may be a reasonable proxy for domestic prices because of the high import content of Oman’s domestic consumption.

The goodness-of-fit of the equations is only fair, and the estimated elasticity for broad money is low compared with the results achieved for other countries. This may be because of poor-quality data; in particular, the price series used may understate the rate of inflation and thus overstate real income.

pakistan, 1961-71

Narrow money

Broad money

pakistan, 1974-78

Narrow money

Broad money

All series used are from IFS. Results are reported for two separate sample periods, 1961-71 and 1974-78.

As might be expected, equations estimated on the full sample period (1%1-78) performed poorly, and failed stability tests dramatically, with respect to the periods before and after the secession of Bangladesh. The goodness-of-fit of the earlier equations is much better than that of the latter. However, the income elasticities seem somewhat high, and their relevance to Pakistan today is questionable. Although the fit for the 1974-78 equations is poor, the point estimates of the income elasticities are very similar to those obtained for other countries in the study and have the additional advantage of being based on recent data. Hence, these are the results reported in the summary tables, presented in the body of the paper.

qatar, 1973-78

Narrow money

Broad money

qatar, 1967-73

Narrow money

Broad money

The money supply series used in these equations are from IFS. National income data are not available after 1973, and the equations for 1973-78 use non-oil exports plus imports (as reported in IFS) as a proxy for non-oil GDP. The equations for 1967-73 use national accounts data for real GDP obtained from the Fund’s Bureau of Statistics. The price index used for the 1973-78 equations is a weighted average of Qatar’s partner country export prices, converted from U.S. dollars to Qatar riyals.

Despite poorer-quality data, the equations for the more recent period have more satisfactory statistical properties than those for the earlier period, an I it is the more recent results that are reported in the summary tables of the main paper. Other regressions, which included an opportunity cost variable, achieved better overall fit but are not reported because the opportunity cost elasticities were implausibly high (–2.5 and –1.8 for narrow and broad money, respectively).

saudi arabia

Narrow money

Broad money

The monetary data are from IFS. The non-oil GDP series were obtained from Statistical Summary, published by the Saudi Arabian Monetary Agency, by subtracting figures for the production of crude petroleum and natural gas and the refining of petroleum from total GDP. There was no evidence of instability at the 95 per cent confidence level between the subperiods 1968-73 and 1974-78.

sudan

Narrow money

Broad money

All series used are from IFS. No constant-price national accounts data are available, and, therefore, current price variables were deflated by the CPI. The results are reasonable, with no evidence of instability at the 95 per cent confidence level between 1964-73 and 1974-77.

syrian arab republic

Narrow money

Broad money

Money supply series were obtained from IFS. The non-oil GDP series were constructed on the basis of Fund staff estimates, and proved to work much better than total GDP as the scale variable. The estimated income elasticities seem reasonable, and there was no evidence of instability at the 95 per cent confidence level when the sample was split into the subperiods 1970-73 and 1974-77.

united arab emirates

Narrow money

Broad money

The monetary data were obtained from IFS. The non-oil GDP series were obtained from Fund staff estimates. The narrow money equation performed much better than that for broad money; there was no evidence of instability in either income elasticity coefficient between 1975-76 and 1977-78. However, little reliance should be placed on estimates derived from so short a sample.

yemen arab republic

Narrow money

Broad money

The money supply series were obtained from IFS. The real GNP series were generated from Fund staff estimates by adding a remittances figure to current price GDP and dividing by the GDP deflator. In neither equation was there evidence of significant instability in the coefficients between 1974-76 and 1977-78. Estimates and stability tests based on so small a sample should, of course, be treated with extreme caution.

people’s democratic republic of yemen

Narrow money

Broad money

The money supply series were obtained from IFS. Current price GNP was generated by adding a remittances figure to current price GDP, both obtained from Fund staff estimates. As no constant price national accounts data are available, both the dependent and the scale variables were deflated by the CPI. While the estimated elasticities are plausible, little weight should be placed on estimates from so short a sample.

pooled cross-section and time-series equations

Table 5 lists the variables employed for each country in constructing the dependent and scale variables. In no instance did an opportunity cost variable enter as negative and significant, so that the scale variable is the only explanatory variable in all equations reported.

Equations for Afghanistan, Bahrain, Egypt, Ethiopia, Iran, Iraq, Jordan, Kuwait, Lebanon, Libyan Arab Jamahiriya, Oman, Pakistan, Qatar, Saudi Arabia, Sudan, the Syrian Arab Republic, the United Arab Emirates, and the Yemen Arab Republic for 1974-78.

Narrow money

Broad money

Equations for eight oil exporting countries (Iran, Iraq, Kuwait, Libyan Arab Jamahiriya, Oman, Qatar, Saudi Arabia, and the United Arab Emirates) for 1974-78.

Narrow money

Broad money

Table 5.Variables Used in Pooled Cross-Section and Time-Series Equations
CountryScale Variable 1DeflatorLast Observation
AfghanistanGNPR/NPGDP1977
BahrainNOGDPR/NNOPGDP1978
EgyptGDPR/NPGDP1978
EthiopiaGDPR/NPGDP1978
IranNOGDPR/NNOPGDP1978
IraqNOGDPR/NNOPGDP1977
JordanGNPR/N/CPICPI1978
KuwaitNOGDPR/NNOPGDP1978
LebanonGDPR/NPGDP1978
Libyan Arab JamahiriyaNOGDPR/NNOPGDP1978
OmanNOGNPC/N/PMPM1978
PakistanGDPR/NPGDP1978
Qatar(X + Imp - Oil X)/PM·NPM1978
Saudi ArabiaNOGDPR/NNOPGDP1978
SudanGDPC/N/CPICPI1977
Syrian Arab RepublicNOGDPR/NNOPGDP1977
United Arab EmiratesNOGDPR/NPGDP1978
Yemen Arab RepublicGNPR/NPGDP1977

See page 564 above for an explanation of these variables.

See page 564 above for an explanation of these variables.

Equations for ten non-oil countries (Afghanistan, Bahrain, Egypt, Ethiopia, Jordan, Lebanon, Pakistan, Sudan, the Syrian Arab Republic, and the Yemen Arab Republic) for 1974-78.

Narrow money

Broad money, 1970-78

Equations for four oil exporting countries with longer series (Iran, Iraq, Libyan Arab Jamahiriya, and Saudi Arabia)

Narrow money, 1970-78

Narrow money, 1970-73

Narrow money, 1973-78

Broad money, 1970-78

Broad money, 1970-73

Broad money, 1973-78

Equations for seven non-oil countries with longer series (Afghanistan, Bahrain, Egypt, Ethiopia, Jordan, Lebanon, and Sudan).

Narrow money, 1970-78

Narrow money, 1970-73

Narrow money, 1973-78

Broad money, 1970-78

Broad money, 1970-73

Broad money, 1973-78

REFERENCES

Mr. Crockett, an Advisor in the Middle Eastern Department, is a graduate of Cambridge University and Yale University.

Mr. Evans, a summer intern in the Middle Eastern Department when this paper was prepared, is a graduate of the University of Sydney, Australia. He is currently pursuing dissertation research at the University of Pennsylvania.

For a comprehensive review of the theoretical and empirical literature relating to the demand for money, see Laidler (1977).

This basic framework was developed by Polak (1957).

The countries included in this study are Afghanistan, Bahrain, Egypt, Ethiopia, Iran, Iraq, Jordan, Kuwait, Lebanon, Libyan Arab Jamahiriya, Oman, Pakistan, Qatar, Saudi Arabia, Sudan, Syrian Arab Republic, the United Arab Emirates, the Yemen Arab Republic, and the People’s Democratic Republic of Yemen.

See Mackenzie (1979) for a discussion of this issue for Egypt.

For example, foreign currency balances may be held as savings by residents spending prolonged periods outside the country or by firms exercising foreign exchange retention privileges.

It is of interest to note that the monetary authorities of the United Kingdom, for example, express their principal objective for monetary expansion in terms of the domestic currency liabilities of the banking system only.

A broad description of the data sources used is provided in the Appendix; a complete set of data is available on request from the authors, whose address is Middle Eastern Department, International Monetary Fund, Washington, D.C. 20431.

See, however, Khan (1977) for evidence on high-inflation countries.

The results of these tests are not reported here; they are available on request from the authors (see n. 7).

Money demand studies in developed countries suggest an opportunity cost elasticity of between—0.2 and—0.8, in general.

Strictly speaking, it suggests that monetization is proceeding at the same pace relative to per capita income growth.

For a more detailed discussion, see Kmenta (1971).

Other Resources Citing This Publication