An Extended Scenario and Adjustment Model for Developing Countries
  • 1 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund

This paper discusses three important extensions to the developing country scenario and adjustment model used in the World Economic Outlook exercises. First, the model is augmented to include fiscal and monetary sectors and now explicitly captures links among government policy, investment, output and inflation. Second, the external sector is modified to allow domestic demand factors to influence imports, as well as allowing flexibility in the financing of imports. Third, the model system is extended to the group of net-creditor countries, and for the oil exporters within this group, oil exports are modeled separately. The revised model is estimated for each of the 95 developing countries and parameter estimates for each of the main equations are presented. The paper also reports the results of four simulation exercises to illustrate how the new model system may be used to quantify the effects of changes in domestic policies and in the external environment.

Abstract

This paper discusses three important extensions to the developing country scenario and adjustment model used in the World Economic Outlook exercises. First, the model is augmented to include fiscal and monetary sectors and now explicitly captures links among government policy, investment, output and inflation. Second, the external sector is modified to allow domestic demand factors to influence imports, as well as allowing flexibility in the financing of imports. Third, the model system is extended to the group of net-creditor countries, and for the oil exporters within this group, oil exports are modeled separately. The revised model is estimated for each of the 95 developing countries and parameter estimates for each of the main equations are presented. The paper also reports the results of four simulation exercises to illustrate how the new model system may be used to quantify the effects of changes in domestic policies and in the external environment.

I. Introduction

A set of developing country models, LDCMOD, developed by Adams and Adams (1989), have been utilized extensively in the World Economic Outlook for analyzing the impact on net-debtor developing countries of changes in their external environment. These models have been used to prepare alternative scenarios which quantify, for instance, the likely effect on developing countries of changes in industrial country growth, international interest rates, and the international price of oil. 1/ In addition, the models have been used for updating of the IMF staff’s projections for individual countries following changes in assumptions regarding the external environment. This is necessitated by the fact that for a large number of mainly medium and small developing countries, staff projections are not available for each semi-annual WEO round, and, in a few cases, may not even be available annually.

This paper reports the results of efforts under way to extend and augment these models in the following three ways. First, to allow an analysis of the effect of changes in domestic policies, and particularly to assess the impact of slippages in the implementation of Fund programs, the models are extended to include fiscal and monetary sectors. The equations for these sectors incorporate error-correction specification which allows policies to have differential short- and medium-term effects. Secondly, a key assumption in the models that imports are residually determined by the foreign exchange available in each country is modified, allowing flexibility in financing imports. Thirdly, the models are extended to the net-creditor countries, most of which are oil exporters. For these countries, oil exports are modelled separately from their non-oil exports, and imports are assumed to be determined by domestic factors.

The models are estimated separately for each of the 95 developing countries. The need to estimate these models for many small- and medium-sized countries for which reliable time-series data on many key variables were not available, imposed a serious constraint in specifying some of the equations. Nevertheless, the estimated models do appear to be robust, and can be highly useful for both simulation and projection adjustment purposes, as well as for analyzing the implications of changes in domestic fiscal and monetary policies.

The rest of the paper is structured as follows: Section II summarizes the main features of the existing model system, and discusses the extensions noted above. Section III specifies the domestic sector including equations for private absorption, government sector, and money, prices and the exchange rate. Section IV specifies the external sector, in which the current and the capital accounts of the balance of payments are modelled separately. Separate equations are specified for exports and imports of goods and nonfactor services, net transfer receipts, net investment income flows, debt and non-debt creating capital flows, and the accumulation of international reserves. For each of the equations in Sections III and IV, estimated parameter coefficients are presented for all developing countries, grouped both by geographical region as well as by analytical group with countries classified according to their predominant export. Section V reports the results of four simulation exercises including the implications of an increase in fiscal deficits as well as monetary expansion, the impact of a fall in world oil prices, and higher growth in industrial countries. Appendix I provides a list of the 95 developing countries and the various country groups and regions referred to in the paper; Appendix II provides a complete listing of the model variables and the equations.

II. Main Features of the Model

The existing model system includes a detailed set of behavioral equations and identities describing the domestic and external sectors of a prototype net-debtor developing economy. 1/ An identical structural model is estimated using historical data for each individual country, but differences across countries are captured by the different parameter estimates. Owing to data limitations and problems with outliers, some of the parameters are constrained to be within a range determined by theoretical considerations.

Output is modeled by disaggregating it into internationally tradable and nontradable goods, but no distinction is made between private and public expenditures. Demand for nontradables is determined endogenously as a function of income, the real exchange rate, and external borrowing. Output is then divided into consumption and investment using historical weights. Inflation is determined by excess demand in the goods market, defined as the deviation of actual from potential output, where the latter is a function of capital stock and the real exchange rate. 2/

LDCMOD has a detailed external sector, including separate equations for import and export volumes, official and private transfers, investment income receipts and debits, non-debt-creating capital flows, new borrowing, and change in reserves. An important feature of the model is the assumption that imports are determined residually by the amount of foreign exchange available. This reflects the external constraints faced by most developing countries following the oil shocks of the 1970s and the international debt problem of the 1980s.

In this paper, LDCMOD is extended in three main directions. First, to permit the model to analyze the impact of changes in domestic policies, fiscal and monetary sectors are introduced and private and public components of aggregate demand are modeled separately. Government revenue is determined endogenously and takes into account the lagged adjustment of revenue to inflation. Government expenditure is divided into current expenditure, which is endogenously determined, and capital expenditure, which is treated as an exogenous policy variable.

With regard to the monetary sector, the stock of money is endogenously determined (although, for simulation purposes, it is also possible to consider it as a policy variable) and the model incorporates the role of public sector financing in the determination of the money stock. Prices are determined by the interaction of money demand and supply functions. Government policy influences output directly through the effect of changes in expenditures on aggregate demand, and indirectly through changes in the stock of money and prices, which in turn influence domestic absorption and exports. A change in government’s capital expenditure also affects the economy’s productive capacity and the supply of exports.

Second, the external sector is modified to take account of the effect of domestic factors on imports. This is implemented by specifying imports to be externally constrained-- and thus determined residually--only when the option of using reserves or foreign borrowing is not available to the country. The model is, therefore, made more flexible so that, depending on the availability of reserves or new loans, imports may switch from being residually determined by external financial flows and export earnings as in the original version of the model, to being determined in a behavioral manner by a mixture of domestic and external factors. In addition, the nominal exchange rate, rather than being exogenous, now adjusts endogenously to maintain purchasing power parity in the long run. This is important for an analysis of changes in government policies that may involve large changes in domestic prices relative to foreign prices over a number of years.

Finally, the model system is extended to the group of net-creditor countries. A key assumption here is that the imports of these countries are primarily determined by domestic factors. The equation for private absorption is also modified to take account of the absence of external constraints, and for the oil exporters within this group, a specific equation is developed for oil exports. Both export and import equations for the oil exporters are estimated by pooled cross-section time-series data and imposing cross-country equality restrictions on the slope parameters. Using panel data for this group of countries should improve the reliability of estimates, given the similarities between these countries, and homogeneity of their main export. 1/

In the following sections, the extensions to the model system, including parameter estimates for the new or modified equations, are discussed in detail. The model is estimated separately for each of the 95 developing countries (87 net debtor and 8 net creditor). The estimation period is 1973-91 for most countries and equations. 1/ Where the modifications are not significant, the specifications are discussed in less detail and the reader is referred to Adams and Adams (1989). Estimation is carried out by ordinary least squares. Owing to data limitations for many countries, more sophisticated estimation techniques--for example, to take account of simultaneity between equations--were not employed. Moreover, in many cases, stringent constraints had to be imposed on the range of permissible parameter values in view of the limited number of observations and to remove the effects of outlying observations, and extreme estimates. Particular difficulty was encountered in the case of some high-inflation countries in the Western Hemisphere and Africa, especially in estimating the equations for money and prices. Imposing these constraints ensured stability of the estimated models which was important for undertaking the simulation exercises.

III. Domestic Sector

A key extension of the model system relates to the separate modeling of the private and public components of aggregate demand, allowing the model to be used for the analysis of fiscal and monetary policies. A second important innovation in modeling the domestic sector is the attention paid to the interaction between fiscal deficits, monetary growth, inflation, and output. Chart 1 illustrates this interaction, abstracting from other parts of the model. An increase in the exogenous component of public expenditure leads to a higher fiscal deficit and monetary growth, and thus higher inflation. In the short run, as public expenditure increases, the effect on aggregate demand is positive despite crowding out of private absorption. In addition, higher public investment expenditure raises the capital stock, increasing potential output as well as enhancing the supply of exports. In the medium run, further crowding out takes place as higher inflation leads to a reduction in private absorption and, by lowering competitiveness, exports. Inflation also leads to a deterioration of the fiscal balance, both directly due to the lagged response of nominal government revenue to higher prices, and indirectly through lower revenue from taxes and tariffs as GDP and exports are adversely affected in the medium run. These various interrelationships are discussed in detail in the following sections. 2/

Chart 1.
Chart 1.

Dynamics of Budget Deficits, Money Growth, Inflation, and Real Output

Citation: IMF Working Papers 1993, 073; 10.5089/9781451849387.001.A001

1. Private absorption

Modeling the investment and consumption components of private absorption separately proved impractical because of data problems for a number of countries and the difficulty in identifying a stable investment function, owing in part to the structural differences between the 1970s and the finance-constrained environment of the 1980s. Therefore, private absorption is modeled in its aggregate form. In doing so, the degree of crowding out that may result from government activity is taken into account. In the case of many developing countries, evidence suggests that such crowding out may take place directly as resources are claimed by the government, or indirectly, through the price system, in particular through higher interest rates and inflation. 1/ Based on this evidence, the formulation adopted here takes account of both the direct and indirect effects of fiscal expenditure on private absorption. Furthermore, the role of financial wealth and access to external loans in determining private absorption is also taken into account, yielding the following specification:

NAPR=F(RI,GCENL/PGDP,TXP/PGDP,FMB/PGDP,RB)(1)

where NAPR, RI, GCENL, PGDP, TXP, FMB and RB denote, respectively, absorption by the private sector in real terms, real income, total nominal government expenditure, GDP deflator, export prices, nominal money stock, and real external borrowing. 2/ Real government expenditure is expected to influence private absorption negatively. 3/ The ratio of export prices to GDP deflator, as a proxy for the real exchange rate, captures the effect of terms of trade changes on private expenditures and is expected to have a positive coefficient. The real money supply is used as a proxy for financial wealth (or permanent income) in the absence of a more comprehensive measure. Finally, external borrowing is included in order to capture the effect of external constraints on absorption of net debtor countries. It is expected that access to foreign loans would have a positive effect on real absorption. External borrowing is excluded in the case of net-creditor developing countries, which are assumed not to face external financial constraints. The equation is estimated in log-linear form. With total private sector absorption determined by equation (1), output is allocated to private consumption and investment spending according to fixed shares, estimated using historic data.

Table 1 presents the parameter estimates--obtained from estimating separately the absorption function for each of the 95 countries--averaged for the four developing country regions (Africa, Asia, Europe and non-oil Middle East, and Western Hemisphere), four country groups according to predominant export of the countries, and for net creditor countries. These estimates suggest that, for net debtors as a whole, the elasticity of private absorption with respect to income is 0.79, and with respect to relative prices it is 0.24. The average elasticities both for real money and real new borrowing are small and positive, but vary considerably across regions. 1/ The elasticity with respect to government expenditure is negative, suggesting a crowding-out of private expenditures resulting from fiscal expansion. Net creditors as a group have a lower income elasticity but a higher price elasticity compared with the net debtors, reflecting the importance of oil prices in determining absorption among the oil exporters in this group.

Table 1.

Average Estimated Coefficients for Real Private Absorption 1/

article image

Estimation is undertaken for individual countries using annual data for the period 1973-91. The coefficients are averaged using as weights the share of each country’s purchasing power parity (PPP) valuation of GDP in the regional or group GDP.

2. Government sector

In developing countries, general government spending, including expenditures by central and state governments, amounts to on average 30 to 40 percent of GDP. While the share of central government spending is lower, it still amounts to 20 to 30 percent of GDP in the net debtor countries (Table 2). The share among net creditors is approximately 10 percentage points higher, reflecting the relatively more important role played by government amongst the major oil exporting countries. Unlike many conventional models which are based on aggregate government expenditure, a distinction is made here between government current and capital expenditure. Because capital expenditure is more likely to be used as a policy instrument by the authorities, it is treated as an exogenous policy variable. 2/ This seems consistent with the evidence of the last decade on the behavior of public sector investment. 3/

Table 2.

Government Expenditure in Developing Countries 1/

(In percent of GDP)

article image

Data are for central government only.

In addition to the distinction between government current and capital expenditures, a second important feature is to allow the possibility of inflation-induced deficits, which arise from a differential impact of inflation on real government expenditure and receipts. 4/ Existing empirical evidence suggests that in the short run the price elasticity of nominal expenditure tends to be larger than the price elasticity of nominal receipts. That is, a rise in the price level causes a greater short-run increase in nominal government expenditure than in nominal revenue, leading to a widening of the fiscal deficit, which in turn leads to higher prices via an increase in the money supply. 1/ As discussed below, this important two-way relationship between budget deficits and inflation is captured by the equations for government expenditure and revenue, and those for inflation and the money supply.

Nominal government current expenditure, GCE, is therefore modeled as a function of the domestic price level, as well as GDP and lagged government revenue. The last variable is included on the assumption that the government adjusts its expenditure partly in line with its revenue. To capture the short- versus long-run effects of changes in the price level, an error-correction formulation is adopted. The long-run elasticity with respect to prices is set equal to unity to ensure the long-run homogeneity of degree 1 of nominal expenditure with respect to prices. The following specification is utilized for estimation purposes:

Δlog(GCE)=a0+a1Δlog(GDP)+a2Δlog(PGDP)+a3Δlog(GCR/PGDP)δ1[log(GCE1)b1log(GDP1)b2log(PGDP1)b3log(GCR1/PGDP1)](2)

where GCR denotes nominal government revenue. The impact of a change in the independent variables is measured by the ai’s in the short run and by the bi’s in the long run. b2, the long-run elasticity with respect to the price level, is set to equal 1. The difference between the short- and long-run elasticities reflects the speed and extent of adjustment of current government expenditure to each of the explanatory variables. For example the mean-lag of the impact of a change in GDP on GC is equal to (b1 - a1) /b1δ1. Thus, the larger is the error-correction coefficient δ1, or the smaller is the difference between the short-run and the long-run elasticities, the faster will be the speed of adjustment following a shock to the system. 2/

Nominal government revenue, consisting of tax and nontax receipts, is specified to be a positive function of prices and domestic activity, as well as of imports and exports. The rationale for including the latter two variables is provided by the very significant proportion of government revenue derived from trade taxes in many developing countries. 1/ Thus, total nominal revenue is modeled as follows:

Δlog(GCR)=α0+α1Δlog(GDP)+α2Δlog(PGDP)+α3Δlog(M)+α4Δlog(X)δ2[log(GCR1)β1log(GDP1)β2log(PGDP1)]β3log(M1)β4log(X)1](3)

where M and X are imports and exports in real terms. As in the case of government expenditure, long-run elasticity with respect to the price level, β2, was set equal to unity to ensure homogeneity.

Data on government accounts for many developing countries are available only for relatively short periods of time. This meant that reliable estimates of the error-correction coefficients and the long-run elasticities in equations (2) and (3) could not be obtained for each individual country. Instead, based on estimation results for countries with longer data series, and to ensure comparable dynamic properties of the model across countries, the error-correction coefficients in both equations were set equal to 0.30 for all countries. A similar procedure, with appropriate homogeneity conditions, was used to obtain the long-run elasticities. Thus, in the expenditure equation, long-run elasticities with respect to government revenue and GDP, b1 and b3, were set equal to 0.50, and in the revenue equation long-run elasticities with respect to GDP, imports and exports, β1, β3 and β4, were respectively set equal to 0.50, 0.25, and 0.25. These are generally larger than the short-run elasticities reported below.

Equations (2) and (3) were estimated conditional on the above long-run elasticities and the error-correction coefficients. The estimated parameters for these functions are shown in Tables 3 and 4, and confirm that for all regions, short-run price elasticities of nominal revenue are smaller than the corresponding elasticities of expenditures. For the net debtors, in the short run, a 1 percent increase in the price level leads to a 0.93 percent increase in nominal expenditure but a 0.81 percent increase in nominal revenue, indicating that inflation leads directly to a deterioration of fiscal balances. In the long run, as noted above, nominal revenue and expenditure increase in proportion to the price level. Short-run income elasticities average around 0.35 in both equations, except for net creditors where in the revenue function, the elasticity is higher. The coefficients on exports and imports in the revenue equation show some variation across regions, with Africa and Western Hemisphere having relatively high elasticities underlining the important role which trade taxes play in several countries in these regions. Net creditors have a significantly higher coefficient on exports than net debtors, reflecting the importance of oil exports for government revenues.

Table 3.

Average Estimated Coefficients for Central Government Current Expenditure 1/

article image

Estimation is undertaken for individual countries using annual data for the period 1973-91. The coefficients are averaged using as weights the share of each country’s purchasing power parity (PPP) valuation of GDP in the regional or group GDP.

Table 4.

Average Estimated Coefficients for Central Government Revenue 1/

article image

Estimation is undertaken for individual countries using annual data for the period 1973-91. The coefficients are averaged using as weights the share of each country’s purchasing power parity (PPP) valuation of GDP in the regional or group GDP.

The balance on the government’s budget, GCB, is given by the following identity:

GCB=GCRGCENL(4)

where GCENL is the sum of government nominal current and capital expenditures.

3. Money, prices, and the exchange rate

The common practice of estimating an equation for the stock of money as a function of income and the interest rate is justified on the grounds that the money stocks is determined essentially by demand factors. In many low-income developing countries, however, money markets are not well developed; in others, especially in Asia and Latin America, while financial markets have developed and broadened in recent years, supply-side factors, including public sector financing requirements, remain important in determining the stock of money. The approach adopted here, therefore, is to explicitly take account of the supply-side factors in determining the money stock. Given the supply of money, domestic prices are then determined by the demand for real money. 1/

Two main sources of money supply are domestic credit expansion and changes in official foreign exchange reserves. While reserves are determined primarily by the balance of payments (discussed below), domestic credit is essentially determined by government policy. Credit to the public sector, in particular, depends largely on the magnitude of fiscal deficits, reflecting government expenditure and tax policies. In view of these considerations, the following equation was estimated for the stock of money:

ΔFMB/NGDP=c0+c1(GCB/NGDP)+c2(ΔR.e/NGDP)+c3(ΔFMB/NGDP)1(5)

where ΔFMB is the change in the nominal stock of broad money, NGDP is nominal GDP, GCB is the nominal government balance and is used as a measure of credit extended to the public sector, ΔR is change in the nominal stock of foreign exchange reserves, and e is the nominal exchange rate; dividing by nominal GDP scales the variables appropriately. The short-run elasticities with respect to the explanatory variables are given by c1 and c2, while the long-run values are c1/(1-c3) and c2/(1-c3). The estimated coefficients are presented in Table 5. The coefficient on the government balance indicates that a 1 percent increase in deficit for net debtor countries leads to an increase in broad money of 0.58 percent in the short run and 0.78 in the long run. As expected, the change in foreign reserves has a small positive effect.

Table 5.

Average Estimated Coefficients for Broad Money 1/

article image

Estimation is undertaken for individual countries using annual data for the period 1973-91. The coefficients are averaged using as weights the share of each country’s purchasing power parity (PPP) valuation of GDP in the regional or group GDP.

The price equation is derived as an inverted demand-for-money function. Thus, from the long run relationship: FMB/PGDP = GDP/v, where v is the velocity of money, the following error-correction specification is obtained: 1/

Δlog(PGDP)=d0+d1Δlog(FMB)+d2Δlog(GDP)δ3[log(PGDP1)log(FMB1)+log(GDP1)](6)

The long-run elasticities with respect to money and GDP are set equal to 1 and -1, respectively, consistent with the assumption of a constant long-run velocity of money. Based on preliminary estimates and an examination of simulation properties of the model, the error-correction coefficient was set equal to -0.70 for all countries, implying a relatively fast adjustment of prices to changes in the stock of money. 2/

This specification of the price equation captures the inflationary dynamics associated with government policy as discussed earlier; in particular, policy affects prices through monetary growth resulting from the monetization of budget deficits. The estimated parameters for the price equation are shown in Table 6. These short-run elasticities with respect to money and GDP are markedly lower than the long-run elasticities of 1 and -1, respectively. For net debtor countries as a whole, a 1 percent increase in money supply leads to a 0.21 percent increase in prices in the short run. This effect is similar across different regions except for the Western Hemisphere where it is larger, reflecting in part the higher variability of inflation.

Table 6.

Average Estimated Coefficients for GDP Deflator 1/

article image

Estimation is undertaken for individual countries using annual data for the period 1973-91. The coefficients are averaged using as weights the share of each country’s purchasing power parity (PPP) valuation of GDP in the regional or group GDP.

With regard to the determination of the. Exchange rate, the modeling difficulties are well-known. In the case of many developing countries, the modeling problem is compounded by the fact that the nominal rate is often kept fixed for long periods of time at a level that may not be justified by the fundamentals. 1/ In view of this, one option would have been to assume that for the duration of any simulation exercise, the exchange rate is exogenous. This would have been inappropriate, however, when a simulation entailed a significantly different inflationary path compared to the baseline. To overcome this problem, it is assumed that purchasing power parity holds in the long run, although in the short run significant deviations from it may occur. 2/ Ideally it would be appropriate to take account of the exchange rate regime in modeling the exchange rate. But quite apart from the modeling difficulties, it is not clear that for simulation purposes such an approach would have any additional benefits over what is attempted here. This is because over the medium term, large deviations from purchasing power parity are not expected, in general, to be sustainable, regardless of exchange rate regime. Thus, the following error-correction formulation for the nominal exchange rate is adopted:

Δlog(e)=αΔlog(PGDP/PGDP*)δ4[log(e1)log(PGDP1/PGDP*1](7)

where e is the nominal exchange rate and PGDP* is foreign prices (defined as the GDP deflator for industrial countries in U.S. dollars). 3/ Estimating this equation for high-inflation countries, or for countries where the exchange rate is fixed with occasional realignments, was particularly problematic, and the parameter estimates varied considerably across countries. This resulted from the high variability in the response of the exchange rate to prices in different time periods across different countries. To avoid convergence problems, and to ensure that model properties conform to prior expectations, we set the short-run coefficient, α, equal to 0.50, and the error-correction coefficient, δ4 equal to 0.20 for all countries. These values imply that the nominal exchange rate adjusts to a price shock by 50 percent after one year and by nearly 80 percent after five years.

IV. External Sector

The current and the capital accounts of the balance of payments are modeled separately. The current account balance is disaggregated into exports and imports of goods and nonfactor services, net transfer receipts, and net investment income flows. For the capital account, separate equations are specified for non-debt-creating capital flows, net external borrowing, and the accumulation of international reserves.

1. Exports of goods and nonfactor services

Exports are disaggregated into non-oil, oil, and nonfactor services. The specification for net debtors assumes that non-oil exports are determined by both supply and demand factors. Demand is assumed to depend on export prices relative to world prices and world income, while supply depends on export prices relative to the domestic price of nontradables and the capital stock in the tradable sector:

TXQNd=F(TXP/TXP*,Y*)(8)
TXQNs=F(TXP/PGDP,K)(9)

where TXQN is non-oil exports, d and s denote demand and supply respectively, TXP is export price, K is the domestic capital stock, and * denotes world. From these structural equations, the reduced form equations for non-oil export volume and prices are derived. In the case of net creditor oil exporters, which have only negligible non-oil exports, both non-oil exports and their prices are treated as exogenous.

The estimation results are reported in Tables 7 and 8. These indicate that for net debtors, the elasticity of non-oil export volume with respect to world income is, on average, above unity, underlining the high sensitivity of non-oil exports to external demand. The elasticity has the lowest value in the case of Africa, and the highest for Europe and the non-oil Middle East. When China and India, where the adoption of export oriented policies is relatively recent, are excluded, the average income elasticity in Asia increases to 2.47 and for exporters of manufactures, to 2.21. Thus, in general, exporters of manufactures have a much higher foreign income elasticity than primary product exporters. The elasticity of export volumes with respect to relative prices is also broadly as expected with the elasticity for the exporters of manufactures markedly higher than that for primary product exporters. Capital stocks influences non-oil export volumes for most groups, with the highest impact for exporters of manufacturers.

Table 7.

Average Estimated Coefficients for Non-Oil Export Volumes 1/

article image

Estimation is undertaken for individual countries using annual data for the period 1973-91. The coefficients are averaged using as weights the share of each country’s purchasing power parity (PPP) valuation of GDP in the regional or group GDP.

Table 8.

Average Estimated Coefficients for Non-Oil Export Price 1/

article image

Estimation is undertaken for individual countries using annual data for the period 1973-91. The coefficients are averaged using as weights the share of each country’s purchasing power parity (PPP) valuation of GDP in the regional or group GDP.

The estimation results for the price equation are similar to those in Adams and Adams (1989). They indicate a very small impact of capital stock and foreign demand on export prices in all regions. The elasticity with respect to world prices, however, is not significantly different from unit in most regions, indicating that changes in world prices are largely transmitted into export prices.

The volume of oil exports by net debtors as a whole is assumed to be exogenous and is allocated across individual countries according to historical shares:

Netdebtors:¯TXQO=F(Totaloilexportsofnetdebtorcountries)(10)

where TXQO is oil exports by individual net debtor countries.

Oil exports by net creditor oil exporters on the other hand, are assumed to be determined by oil prices relative to world prices, (industrial countries) world GDP, and a time trend (which captures the long-run tendency for both a decline in energy intensity of production, and a substitution away from oil). To allow for dynamic adjustment over time, the lagged value of oil exports is also included:

Netcreditoroilexporters:¯  TXQO=F(TXQO1,TXPO/P*,Y*,t)(11)

where TXPO is the price of oil, P* is industrial country GDP deflator, Y* is industrial country GDP, and t is a time trend. Oil export equations for the oil exporting countries are estimated by imposing cross-country equality restrictions on the parameters (except for the intercepts). This should give more reliable estimates than using data on individual countries, given the homogeneity of oil and similarity between net creditor oil exporters. The parameter estimates, presented in Table 9, all have the expected signs and magnitudes, and suggest that holding world income and prices constant, exports of oil exporters fall, on average, by 2 percent a year. 1/

Table 9.

Panel Data Parameter Estimates for Oil Export Volumes: Oil Exporting Net Creditors 1/

article image

Estimation is undertaken using a panel of cross - section time series data (for the seven oil exporting countries for the period 1973-91).

World oil prices are assumed to be determined exogenously and the dollar export price of oil in each individual net debtor or net creditor country, TXPO, is assumed to follow the world oil price:

TXPO=F(Worldoilprice)(12)

Exports of nonfactor services, XNFS, comprise receipts from tourism, banking, and other services. For net debtor countries, these exports, in current dollar terms, are assumed to depend on the level of dollar GDP in industrial countries:

XNFS=F(IndustrialcountrycurrentdollarGDP)(13)

In the case of oil exporting net creditors, these exports are relatively small and thus are treated exogenously.

2. Imports of goods and nonfactor services

A key feature of the existing model system is the assumption that imports are determined residually by the foreign exchange available in each country. Foreign exchange, in turn, is a function of export earnings, transfer receipts, and net capital flows. The assumption reflects the constraint on a majority of developing countries in the 1980s that resulted primarily from the external debt crisis, and the virtual cessation of new commercial bank lending. This feature of the model is modified in the new version to allow flexibility in financing imports which lets domestic factors play a role.

It is assumed that for each of the net debtor countries, total imports switch between being fully constrained by external financing and being determined by a mixture of domestic and external factors, depending on the size of a country’s foreign exchange reserves. 1/ Chart 2 illustrates the regime switch and the various factors affecting the determination of imports. When reserves are sufficiently high, the left panel is the relevant one and actual imports are determined by desired imports. When reserves are low, the right panel is the relevant one and imports are constrained. Imports may be further limited in this case when external borrowing is also constrained. These different regimes and constraints are amplified in the equations below.

Chart 2.
Chart 2.

Determination of Imports

Citation: IMF Working Papers 1993, 073; 10.5089/9781451849387.001.A001

If the reserves-to-import ratio is low, then imports are constrained by the amount of foreign exchange available: 2/

Netdebtors:¯Mc=(X*TXP+TΔR)/TMP(14)

where Mc denotes total constrained imports in real terms, X*TXP is total export revenue, T is the sum of net transfer receipts, net investment income receipts and net capital inflows (which will be discussed later), ΔR denotes change in international reserves, and TMP is the price of imports.

When imports are equal to constrained imports, foreign reserves are exogenously determined. One of the key items in net capital inflows is new borrowing, which, as discussed later, could be constrained depending on the debt/GDP ratio. Therefore, two types of potential constraints, in effect, operate on imports; one results from the availability of reserves and the other from the access to new borrowing from the international capital markets.

In contrast, when the reserves - to - import ratio is sufficiently high, imports are determined by the following behavioral equation:

Netdebtors:¯Mu=F(TMP/PGDP,RAN,(R/M)1,X)(15)

where Mu is unconstrained imports and RAN is expenditure on home goods. Exports are included so that external factors are to some extent taken into account, even when the country is not considered to be externally constrained. The presence of the lagged reserves - to - import ratio also generates a response of imports to the change in external environment in the medium term. 1/ The results of estimating this unconstrained equation over the historical period are reported in Table 10, and indicate an average elasticity of expenditure on home goods of 0.74 and an average relative price elasticity of -0.37. Both real export earnings and external reserves also have positive effect on imports.

Table 10.

Average Estimated Coefficients for Unconstrained Merchandise Imports 1/

article image

Estimation is undertaken for individual countries using annual data for the period 1973-91. The coefficients are averaged using as weights the share of each country’s purchasing power parity (PPP) valuation of GDP in the regional or group GDP.

Given total imports, division of imports between oil and non-oil imports for the simulation period is determined by their relative shares computed from historical data.

Non-oil and oil import prices, TMPN and TMPO, are determined as follows:

TMPN=F(Worldpricesofmanufacturedgoods,andofnonoilprimarycommodities)(16)
TMPO=F(Worldoilprice)(17)

When total imports are unconstrained, international reserves are residually determined: 2/

ΔR=X*TXPM*TMP+T,whenM=Mu(18)

Imports of net creditors in the short run are assumed not to be externally constrained and are determined primarily by domestic factors, while foreign reserves adjust residually. Foreign reserves, however, continue to play a role in the determination of imports in the medium term. These assumptions are reflected in the following specification:

Netcreditors:¯TMQN=F(TMP/PNTD,Y,(R/M)1,M1)(19)

where TMQN denotes non-oil imports, TMP is import price, PNTD is the price of nontradables, and Y is income. 1/ As in the case of oil exports, slope coefficients were restricted to be equal across all net creditor oil exporting countries. Moreover, the effect of private and government income is separated to allow different absorption elasticities which could result from the fact that the government is a major importer, and oil revenue is largely taxed (and spent) by it. 2/ The estimation results indicate a small price elasticity, a small but positive response to foreign exchange reserves, and a significantly larger absorption elasticity for the government than for the private sector (Table 11).

Table 11.

Panel Data Parameter Estimates for Non-Oil Import Volumes: Oil Exporting Net Creditors 1/

article image

Estimation is undertaken using a panel of cross-section time series data (for the seven oil exporting countries for the period 1973-91).

3. Transfers and net investment income

The equations for private and official transfers, BTRP and BTRG, are as follows:

BTRP=F(ExportearningsofMiddleEastoilexports,EuropeanGDP).1/(20)
BTRG=F(Officaildevelopmentassistance).(21)

Equation (20) reflects the fact that a bulk of private transfers in the Asian and African countries have been remittances mainly from the industrial countries in Europe and the high income oil exporters in the Middle East.

Investment income credits, BXSI, are specified to depend on a country’s stock of foreign assets and their rate of return, which is proxied by the London interbank offered rate (LIBOR) on six-month dollar deposits. Two categories of investment income debits are considered: direct investment debits, BMSDI, which are modeled as a function of GDP, and other debits, BMSNDI, which comprise interest payments on external debt and are modeled as a function of debt-service payments:

BXSI=F(LIBOR*R)(22)
BMSDI=F(GDP)(23)
BMSNDI=F(Debtservicepayments)(24)

4. Capital account

The capital account is disaggregated into three components: non-debt-creating capital flows, net external borrowing (defined as new borrowing less amortization), and the accumulation of international reserves. The determination of reserves was noted above, and the other equations, discussed below, are essentially unchanged from the existing version of LDCMOD.

Non-debt-creating capital flows, NDKFD, comprise mainly direct foreign investment and portfolio equity flows, but also include new allocation of SDRs, valuation adjustments, and a balancing item necessary to reconcile the discrepancies between the current and capital account of the balance of payments. While in theory there are a large number of factors that affect these flows, in practice, domestic economic performance proxied by the change in domestic GDP appears to be a key determining variable. Thus, the following specification is used:

NDKFD=F(ΔGDP)(25)

New external borrowing is assumed to depend on whether or not a country has access to the international financial markets. If the debt-to-GDP ratio is high (the average over 1989-92 is above 2), then borrowing is assumed to be constrained and is determined by

Bc=Averageamountofdollarfinancingreceivedovertheperiod1988921/(26)

where Bc denotes constrained borrowing. If, on the other hand, external debt is not too large (debt-to-GDP ratio is below 2), B is determined by real LIBOR, terms of trade changes (ΔTOT), domestic GDP, and lagged new borrowing:

Bu=F(RealLIBOR,ΔTOT,GDP,B1)(27)

where Bu is unconstrained borrowing.

Amortization payments, AP, are assumed to be related to external debt, lagged one year:

DSPT=F(D1)(28)

Total stock of external debt, D, is then determined by net external borrowing, B, and last period’s stock of debt adjusted for valuation effects:

D=(1+u)*(1v)*D1+B+debtreductionoperations(29)

where u is the proportional change in the dollar multilateral exchange rate (MERM) between t-1 and t, and v is the share of debt in non-dollar currencies. Debt-reduction operations are exogenous.

Finally, interest payments on outstanding external debt are modeled as a function of current and past LIBOR in order to allow differential interest rates on borrowings of different vintages:

Interestpaymentsdue=ΣaiLIBORtiD1(30)

where ai’s are weights that sum to unity.

V. Simulation Experiments

The results of four simulation exercises are reported here to illustrate how the new model system may be used to assess the effects of changes in domestic policies and in the external environment. Two of the simulations examine the implications of an increase in government expenditure and excessive monetary expansion. Two other simulations consider the impact on developing countries of a fall in world oil prices and higher growth in industrial countries. Before discussing the simulation results reported below, some long-run properties of the model are worth emphasizing.

In the long run, nominal variables, including money, prices, and the exchange rate move proportionally for any given level of output. Nominal shocks, however, may have long-run real effects to the extent that the capital stock (and thus productive capacity) or the composition of output (between the private and public sector) are affected by short-run dynamics. For instance, the private absorption equation allows for an increase in government expenditure to crowd out private absorption to such an extent that output is not neutral to the composition of final demand. In addition, government nominal expenditure and revenue are homogeneous of degree one in prices in the long run. Although in the long run a balanced budget is not imposed, the inclusion of revenue in the expenditure functions ensures that these two variables do not deviate from each other significantly.

1. Increase in government expenditure

In this scenario, government capital expenditure in nominal terms is assumed to rise 20 percent above the baseline for the period 1993 to 1998. 1/ In the initial year, reflecting this stimulus, real output rises by 3/4 of 1 percent above the baseline in the net debtor countries, and by 2 percent in the net creditor countries (Table 12). The increase in capital expenditure also raises potential output and augments the capital stock, which in turn increases export capacity. This short-run beneficial impact on output is smallest in the Western Hemisphere and largest in Africa, reflecting, in part, the differing share of government expenditure in GDP. By 1995, the gain in output is virtually eliminated in all regions, reflecting the adverse effects both on private absorption and exports of higher inflation stemming from faster money growth and the increased burden of government deficit as expenditure adjusts faster than revenue to higher prices. Across regions, this Is most notable in the Western Hemisphere and in Africa, where there are a number of large high inflation countries; the Asian and the Middle Eastern and European regions experience a lower increase in prices, reflecting the high weight of countries with low inflation. By 1998, net debtor countries’ real GDP falls by 1/2 of 1 percent below the baseline, as domestic prices and the exchange rate adjust further to higher fiscal deficits.

Table 12.

Medium-Term Implications of Simulations: 20 Percent Increase in Government Capital Expenditure, 1993-98

(Difference from the reference scenario in percent)

article image

Ratio to nominal GDP

Ratios to exports of goods and services.

For net creditors, the output increase, as in the case of the net debtors, has a declining trend. However, the short-run positive impact is larger owing to the higher stimulus reflecting the higher share of government expenditure in GDP. The positive effect, furthermore, lasts longer since exports do not decline (oil prices are exogenously determined) and since imports are not constrained.

2. Monetary expansion

A rise in the money supply has a short-run positive effect on the private sector’s real wealth before it is transmitted into higher prices. This positive effect increases private absorption temporarily, leading to higher activity and GDP. In the medium term, however, through the rise in prices, there are negative effects on aggregate demand, similar to those described in the case of fiscal expansion. The effects are manifested in a loss of competitiveness which leads to a fall in exports, and a decline in private absorption due to the fall in the terms of trade, real income, and wealth. Over time, the effect on GDP diminishes as prices and the exchange rate adjust to the increase in money supply.

The simulated effects of an increase of 10 percent in the stock of money are reported in Table 13. It can be seen that for net debtors as a whole there is a short-run positive effect on GDP of 1 3/4 percent above the baseline after the first year, but by 1998, the effect is reversed to 1/2 of 1 percent below the baseline. The GDP deflator rises by about the same amount as monetary expansion by the end of the period. In some regions, depending on the speed of adjustment of prices to money and the nominal exchange rate to prices, small cyclical movements are also observed. In the case of net creditors, the effect on prices is similar to that for net debtors; however, the effect on output is relatively muted in the initial year and is virtually negligible subsequently.

Table 13.

Medium-Term Implications of Simulations: 10 Percent Increase in Broad Money, 1993-98

(Difference from the reference scenario in percent)

article image

Ratio to nominal GDP

Ratios to exports of goods and services.

3. A fall in the nominal price of crude oil

This scenario considers the effect of a fall in the nominal price of oil, below the baseline, sustained over the period 1993-98. Interest in such a scenario stems from the possibility of excess supply developing in the world oil market over the medium term. 1/ Countries that are dependent on oil imports and are finance-constrained would benefit from a fall in oil prices as this allows a rise in non-oil imports and a higher level of activity; even countries that are not finance-constrained would benefit from a terms of trade gain. Oil exporting countries, on the other hand, will have lower oil revenue (as long as the price elasticity of world demand for oil is less than unity), lower income and lower imports. However, since the volume of oil exports will rise, the effect on GDP (which by definition does not take account of movements in the terms of trade) would depend on the relative size of the fall in domestic activity and the rise in export volumes. Among oil exporters, the effect of a fall in oil prices is likely to be most noticeable in the case of net creditor oil exporters where oil constitutes a large proportion of GDP.

The simulation results indicate that a 10 percent fall in oil prices has a rather small effect on net debtor countries as a whole (Table 14). Some variation, however, is observed across different regions depending on the presence of major oil exporters. In Africa, for example, primarily because of Nigeria, there is a small negative effect on GDP. In the Western Hemisphere, on the other hand, by 1998, the combined positive effects on large oil importers, such as Brazil, offset the negative effects on countries like Venezuela and Mexico. As expected, the effects of an oil price fall is most significant in the case of net creditor oil exporters. 2/ To highlight the differential impact on income and GDP in this case, the simulated effects on both of these variables are reported. Real GDP falls below the baseline initially as the positive effect on export volume is dominated by the negative effect on domestic absorption, but rises above the baseline as export volumes increase. Income, on the other hand, has a more substantial initial fall and remains below the baseline throughout the period. The current account worsens, and the reserves-to-import ratio falls significantly.

Table 14.

Medium-Term Implications of Simulations: 10 Percent Fall in Oil Prices, 1993-98

(Difference from the reference scenario in percent)

article image

Ratio to nominal GDP

Ratios to exports of goods and services.