Information about Europe Europa
Chapter

8. External Sector

Author(s):
International Monetary Fund
Published Date:
November 2005
Share
  • ShareShare
Information about Europe Europa
Show Summary Details

a. Some basic concepts

The economic developments recorded in a country’s balance of payments (BOP) reflect the interplay of forces in the domestic economy and the global economy. If home prices rise faster than world prices, and the exchange rate is held constant, home goods will tend to lose competitiveness on world markets, and the volume of exports will decline; foreign goods will become increasingly attractive on the domestic market, and imports will increase. If there is rapid growth of spending in foreign markets, exports will tend to grow strongly, all the more so if demand growth in the domestic economy is sluggish at the same time.

The influences on external flows that come from the domestic economy in the forecast period should be based on the forecasts for other sectors, especially the growth of real output and income and the change in the average level of the price of domestic output. In principle, one could forecast developments in the rest of the world in the same way, country by country, adding up the projections to provide a forecast of world market conditions faced by the home economy in the forecast period. However, that would be a time-consuming task and probably impractical. A number of international organizations, including the IMF and the OECD, regularly forecast macroeconomic conditions in the world economy in line with their respective roles in monitoring developments in member countries.1 Indicators of output and prices in individual economies are averaged, using the export or import shares of the home economy as weights, to produce measures of expected global developments that reflect the prospects of the country in question. Forecasters in Country A will usually know more about the details of their country than the staffs of international organizations do, and they may also be more current, but the international organizations are likely to be in a better position to have a consistent view of the global economy.

Basic procedures for combining measures of domestic and global activity to produce forecasts of BOP items are described in the following two sections of this workshop. The forecasts that emerge from this process will usually correspond to “underlying” or fundamental conditions. For example, the forecaster is not likely to have reliable information about weather conditions in the forecast period and therefore will assume average rainfall and temperature. If agricultural products make up a significant share of the country’s exports, then the export forecasts will tend to reflect this neutral assumption of neither a bumper crop nor a poor harvest. Similarly, forecasts of financial flows will often focus on longer-run net movements of “capital” (finance) and abstract from sudden short-run shifts of funds whether recorded or included in errors and omissions. On the other hand, periods of strong or weak macroeconomic activity can be forecast if one has foreknowledge of the stance of official policies. BOP forecasts will not necessarily be neutral with respect to policy-induced fluctuations in activity and the price level.

In principle, BOP forecasts will be more accurate if they are more disaggregated. For example, the factors that determine oil imports may be significantly different from those influencing imports of other goods and services; exports of primary products may reflect weather at home and demand abroad, whereas exports of manufactured goods may be uncorrected with rainfall but reflect demand at home—what is “left over” after home demand has been satisfied is exported. But while disaggregation can increase accuracy, it also requires more time, data, and estimation. A high degree of accuracy is not guaranteed by a highly disaggregated approach since accuracy of forecasts also involves whether the determining influences, such as output and prices (the explanatory variables), change as expected in the forecast period. Abrupt and disruptive political, military, weather, technological and institutional events usually cannot be anticipated; errors due to such events usually cannot be avoided. The forecaster will try to balance the benefits and costs of disaggregation in some informal way in deciding whether to use broader or narrower concepts as endogenous variables. In the present forecasting exercise, the degree of disaggregation is also limited by the time available to participants and by the objectives of the course, which involve demonstrating techniques of general usefulness rather than pinpointing all of the influences on an item in the BOP accounts of a particular country.

Like the trade-off between accuracy and effort associated with the degree of disaggregation, there is also a trade-off between more sophisticated and less sophisticated methods. Consider the following hierarchy: (1) “Next year will be like last year.” In this simplest case, the forecaster simply assumes no change in any BOP item. The forecast errors associated with this method may not be very large, but neither can great interest be attached to the outcome; if nothing changes, there is no need for forecasts. Nevertheless, for certain parts of the BOP, this method may be useful. For example, Net Errors and Omissions is often forecast by use of a naive model. (2) “The economy will continue to expand.” The second method assumes that the rate of growth of the item will remain constant, rather than the level. Most forecasters would probably concede that this method gives more useful results if the rate of growth to be used for the forecast period has been estimated on the basis of a past time period that can be considered typical in some sense. (3) “Economic behavior links aspects of the BOP—such as export and import volumes—to output and price developments at home and abroad; by allowing for this behavior, the forecaster can increase the accuracy of his projections.” If output and income in the home country decline, it is very likely that the demand for imports will also decline, other things being equal. Forecasts based on trends or mean historic growth rates will have larger errors when the economy is not growing in the usual way.

Economic forecasts are based on the notion that aspects of past events contain some usable information about what we can expect in the future. In distilling patterns useful for forecasting from past events, two factors strongly influence how far one can go, namely the availability of data and the stability of the structure of the economy. In measuring the effect of a prospective change in relative prices on, say, the volume of imports, one would like to have a measure of the average level of prices of importables (including goods and services that might be imported) and the average level of costs of producing domestic substitutes for those foreign goods. What is usually available, instead, is the level of prices of actual imports (an import price index or unit value index) and the level of prices of domestic output of all kinds (the GDP deflator). These series at best approximate the information that is needed theoretically.

Structural change and the reform of institutions may be carried out piecemeal rather than all at once, depending on political circumstances. Liberalization of the external sector may include removing imports from a restricted list, lowering advance import deposit rates, lowering or eliminating tariffs, removing quantitative restrictions, and dropping licensing fees. This process, which may occur over a number of years, usually proceeds much faster in periods in which there is a positive balance of payments outcome, but may be reversed, if temporarily, when there is a large and unsustainable deficit. Since such measures are often microeconomic (affecting individual commodities), there may not be sufficient information to estimate their impact in terms of an equivalent price increase or decrease of imports on average, which is what is needed to forecast macroeconomic behavior. The essential point is that forecasting methods based on voluntary economic behavior will be inappropriate in an environment in which there is significant use of, or changes in, administrative controls.

Finally, BOP forecasts may be either consistent or diagnostic. If expenditures on foreign goods and services exceed receipts from sales, either the home country must receive financial inflows from investors, it must borrow, it must pay out of its foreign reserves, or it must be in arrears (“involuntary lending”). This is a tautology. It holds in reality, but not necessarily in forecasting, depending on the forecaster’s purpose. Suppose the minister of finance is asking about the consequences of an increase in the fiscal deficit, to be financed by borrowing from the central bank. The BOP forecaster may report that this scenario would tend to produce accelerating inflation at home, lower exports, increased imports (if the exchange rate is not freely floating), and a disincentive to foreign investors and lenders that will tend to shrink financial inflows; the incipient overall deficit may exceed the country’s international reserves, resulting in a hypothetical financing gap. In this case the forecast is diagnostic; in practice, such a large deficit cannot occur. The government would be forced to rein in its expansionary fiscal and monetary policies, or impose administrative controls on imports, or arrange for special loans, or allow a large devaluation of the currency, or fail to make agreed payments. In reality, some factor would have to change allowing the accounting identity underlying the balance of payments to be satisfied (a consistent forecast). The usefulness of a diagnostic forecast is in breaking the policy-making process down into two parts: “What will happen if certain actions are taken (or not taken)?” Then, in a second step, “What options are available to avoid large imbalances and achieve a desirable outcome that is consistent with the BOP constraint?”

b. Forecasting major items

This section discusses methods that use economic behavior and statistical techniques to forecast the balance of payments.2 These methods will work best when applied to short-run or medium-run forecasting (one or a few years ahead) in a small to medium-sized economy with diversified export and import flows, little use of administrative restrictions and controls, limited structural change during the sample period, and reasonable availability of annual data.

For most BOP current account entries, any change from one period to the next is the result of a change in volume and a change in price. Economic theory does not provide inferences about values directly, but about volumes and prices separately. Thus, to forecast imports for a coming year, one will need a forecast of the change in volume and a forecast of the change in price. The following identity is worth keeping in mind:

Note that “one plus the proportional change” in a variable, xt, is equal to xt/xt-1.

(1) Merchandise exports

An export equation for forecasting purposes would generally be of the form,

The variables in the equation are defined as follows:

XR: the volume of merchandise exports, measured as an index number or in constant prices;

ER: the nominal exchange rate, units of domestic currency per unit of the currency of a major trading partner (for example, the U.S. dollar);

PX: the average level of prices of exports expressed in foreign currency (the same currency as the exchange rate), in index form;

PD: the average cost of producing exportables in the home country, usually proxied by the GDP deflator, an index;

YR: productive capacity in the export sector, possibly proxied by real GDP;

GAP: excess demand in the domestic economy as measured by real spending minus potential output, possibly proxied by real GDP minus real trend GDP.

While equation (8.1) is intended to represent factors influencing the supply of exports, it is not a simple textbook supply function, which would have quantities on one side of the equal sign and prices on the other. The motivation for the above form is that forecasts will have been made elsewhere of the level of potential output in the forecast period. Other things being equal, the capacity to produce exports will vary with total output. The ratio of export prices to home costs is a second, independent, explanatory factor. The price of exports relative to average production costs in the economy will reflect incentives to increase or decrease resources used in the export sector (thus decreasing or increasing the production of non-exports).3 Third, if demand is strong at home, more exportables will be consumed domestically, and export volume will be less; this is the role of the variable GAP. With total production of exportables and home and foreign prices determined elsewhere, this equation allows implicitly for the amount of productive resources devoted to producing exports and for the proportion of exportable output that is exported instead of being consumed at home. Using the small country assumption (see Box 8.1), one can logically base export forecasts on the special form of supply relation in equation (8.1).

In addition to the variables listed above, other considerations may play a part in an actual application of these concepts. A major one is lags. If exportable primary commodities are harvested late in the calendar year, output in year t will influence exports in t+1. Export incentives or taxes will also influence export volumes. Examples of commercial policies that are hard to incorporate in a forecasting equation include tariffs, license fees, and quotas on individual commodities. According to economic theory, measures of this type may be expressed in terms of a price equivalent, and thus included in the “price” of traded goods. But the size of these subsidies and taxes may differ among commodities, may be changed frequently over time, and may be recorded only in obscure or inconvenient data sources. It is therefore usually very difficult to express the effects in terms of the average price of all exports or imports during a period of time.

The price that is relevant in determining production incentives is the producer price, which may differ substantially from the export price. If agricultural output is sold to a government marketing agency that is the sole exporting agent, there is a possibility that the costs of the marketing board may become inefficiently high, that prices paid to farmers may be held down to prevent windfall gains when the exchange rate is adjusted, or that the board’s revenues may be siphoned off to supplement government revenues from taxes or to make private payments to enrich individuals. In such cases, measured export prices may not help substantially in forecasting export volumes (and the resulting, reduced incentives to produce for export will result in a stagnant export sector).

Box 8.1.Box 8.1. The Small-Country Assumption

In economic theory, volumes of goods and services are analyzed by distinguishing factors related to demand and factors related to supply. The two groups together provide an “equilibrium” level of output, so that we might consider estimating both supply and demand “curves” for the main items in the BOP accounts. Usually, however, a short-cut is possible. If residents of the home country increase their demand for a particular imported good, say coffee, this generally will tend to result in pressure on the world price to rise. However, since the consumers in the country in question are likely to comprise a very small percentage of total world coffee drinkers, and the increase in their demand is an even smaller proportion of the total quantity consumed per unit of time, the increase in the world price that results from more demand in a single country is likely to be small. In fact, it is likely to be infinitesimal. In formal analytical terms, this amounts to saying that the supply curve of coffee is horizontal in the short run and in relation to changes in demand that may occur on the part of consumers in the country in question.

The supply of coffee is not price-insensitive in general, and especially not in the medium run. However, in estimating the volume of coffee imports of Country A we may ignore the effects of A’s demand changes on the world price because the demand changes are small relatively speaking. This is known as the “small country assumption,” although it is likely to hold for countries of all sizes if we are talking about imports in the aggregate—imports of all goods (and services). The significance of the small-country assumption is that one may forecast import volume changes based solely on factors affecting demand. Coffee demand will go up or down in line with the buying tendencies of consumers in A, and there is no need to allow for a reduction in a hypothetical import increase because the import demand caused scarcities and rising prices in the global market. From a formal analytical point of view, this assumption is of great significance because it simplifies the forecasting process considerably.

For exports, the small-country assumption has a similar, but not identical, implication. If suppliers in a particular country increase their output more than any simultaneous increase in domestic demand, so that more output of Country A is available for export, one may ignore any tendency for this supply increase to depress world prices. In fact world prices may change in the forecasting period—due to wars, freezing weather in other countries, the introduction of new hybrids or fertilizers or processing equipment in some or many supplying countries. (These changes are supposed to be already reflected in the price forecasts available from outside organizations or prepared internally; global events will influence the global supply, and hence the global price.) But potential supply changes in Country A are too small a proportion of the world export market to matter much, according to the small-country assumption. The implication is that a country can sell as much as it likes (and is able to produce) without depressing the global market price; there will always be enough demand if it can produce and export a bit more It is as though world demand is infinitely clastic (a horizontal global demand curve) from the point of view of conceivable changes in the quantities supplied by Country A. Therefore, export volume can be forecast without regard to demand factors according to the small-country assumption.

If the small-country assumption does not hold for exports, then changes in the quantity supplied by Country A may influence the average price of its exports. In this case, prices and volumes are determined by the interaction of both demand and supply. In general, demand for A’s exports will tend to vary with income and expenditure in partner countries as well as with relative prices.

To estimate the relevant coefficients, the forecaster would use statistical methods applicable to systems of equations.4

(2) Merchandise imports

In light of the small-country assumption, the equation for import volumes will contain factors influencing the demand for foreign goods. That is, it is assumed that changes in demand are small enough relative to global supply that supply is not a factor limiting the imports of Country A in the forecasting period. In general terms, an equation suitable for forecasting import demand may be written,

The variables in the equation are defined as follows:

  • MR: the volume of merchandise imports, measured as an index number or in constant prices, and excluding insurance and freight (imports on “f.o.b.” basis);
  • YR: aggregate income, possibly proxied by real GDP;
  • ER: as above, the nominal exchange rate, units of domestic currency per unit of a major foreign currency;
  • PM: the average level of prices of imports in foreign currency (the same currency as the exchange rate), expressed as an index;
  • PD: as above, the cost of producing import substitutes in the home country (the “domestic price level”), usually proxied by the GDP deflator, an index;
  • GAP: demand minus potential output in the home economy, possibly proxied by actual minus trend real GDP.

Equation (8.2) includes the hypothesis that as income rises, the demand for foreign goods will increase (along with the demand for domestic goods). If foreign goods become cheaper relative to home-produced substitutes, the demand for imports will also increase for this reason. Finally, if spending at home is high relative to potential output, demand will spill over onto the foreign market; as domestic supplies are bought up, home consumers and producers will buy foreign items because the domestic versions are unavailable.

The list of special factors that may complicate import forecasting is similar to the one presented above for exports. If the average price of foreign goods (measured in domestic currency) rises abruptly, perhaps because the home currency was devalued, it will take time for importers to adjust their orders downward to reflect the smaller level of demand, and for the reduced quantities of foreign goods to be shipped, off-loaded, and sold to domestic buyers. In other words, there may be a lag between the change in a right-hand-side variable in the above expression and the corresponding change in the volume of imports. Just as in the case of exports, it is a simple matter to allow for this by including lagged values of the explanatory variables on the right-hand side of equation (8.2).

Forecasting accuracy may be increased by disaggregating imports by type. Oil imports are often separated from other goods because it may be difficult to substitute domestic sources of energy for imported petroleum and because the world price of oil has changed markedly several times in recent decades relative to the average world price of non-oil commodities and manufactures. However, this disaggregation will be successful for forecasting only if the world price is allowed to pass through promptly to domestic buyers. If the government “cushions” some or all consumers and producers by subsidizing the home oil price, and only slowly or much later adjusts the domestic price of oil toward the global level, then the measured price of oil imports and the actual domestic oil price will differ. Without data on the price actually paid in Country A, it will not be possible to use past consumption patterns to forecast the volume of oil imports in a future year even if oil and non-oil commodities have separate equations.

Besides disaggregating by type of import, one may disaggregate the source of demand on the right-hand side of equation (8.2). Investment spending in the economy may, for example, contain a higher proportion of foreign inputs (imported equipment) than consumption spending (imported raw materials, semi-processed goods and spare parts, and finished foreign goods). To reflect this, YR would be disaggregated into the main components of spending allowed by the national accounts data of the home country. If, in fact, consumption and investment spending have similar shares of imports, then the disaggregation will not improve forecasting accuracy by much.

Structural changes and reforms are harder to incorporate. If tastes in the importing country are gradually changing, and if the secular shift in tastes tends to favor foreign products over home goods (or the reverse), this factor will escape inclusion or it will be mixed together with other influences that change gradually over time (trend domestic output). If there is a single and sudden change—in port facilities and roads, credit for foreign purchases, or the presence in the country of foreign sellers and distributors—one may be able to capture the influence of the change with a “dummy” (dichotomous) variable.

Hardest of all to include are irregular changes in the levels of tariffs, license fees, advance deposits, or direct quantitative controls on imports or on the allocation of foreign exchange to pay for imports. The demand equation above represents voluntary economic behavior, not administrative controls. Suppose the domestic authorities always ease import restrictions when the BOP is strong and increase restrictions when it is weak; in this case, the level of foreign reserves of the banking system may be included as an explanatory variable in equation (8.2). However, the authorities may ease restrictions even when there is no BOP surplus, for example because they have adopted a program of liberalization of trade and payments; they may postpone tightening restrictions if they are willing to tolerate a devaluation of the nominal exchange rate or if they are able to borrow foreign exchange from foreign creditors. In the extreme, they may simply allow a longer lag in the processing of import license applications instead of tightening controls when reserves fall. For these reasons, use of the level of short-run net foreign assets is often not related in a systematic way to the degree of restrictiveness of administrative measures affecting import volume. To the extent that restrictions are used to control import volume, it is not possible to exploit past patterns of voluntary behavior to forecast the volume of imports in a future year.

(3) Travel receipts and payments

Travel services as recorded in the balance of payments are conventionally defined to include food and lodging and other tourist expenditures that occur after the traveler arrives in a foreign country, but travel does not include transport costs (such as airline tickets). It is as though the home country exports restaurant meals and hotel nights to its foreign visitors, while its own citizens import them from foreigners when they travel abroad. Not surprisingly, methods for forecasting travel credits and debits resemble equations for forecasting merchandise exports and imports. The coefficients of the equations may, of course, be quite different.

For tourists from western European countries, income elasticities for travel expenditures have been estimated in careful studies to be as high as two to three in some post-war decades. That is, as incomes increased 1 percent in real terms, expenditure on foreign travel increased 2-3 percent; there was a strong preference to spend extra income on tourism.5 Price elasticities as applied to a single foreign destination were also quite high, in the range of two to three. If hotels in a country with sunny beaches became expensive because the domestic exchange rate was not maintained at a competitive level, the firms putting together package tours in the tourists’ home countries would simply switch to the next country down, or up, the coast.

The same reasoning applies to travel payments as to receipts. For an economy with lower income than western Europe in recent decades, the income elasticity applicable to payments would likely be between one and two since a larger proportion of travel expenditures would be for business purposes (and probably less income-elastic). The price elasticity would be much lower if the price of travel services refers to foreign countries in general. A leisure traveler leaving home may be deterred from taking his vacation in Country B if its costs are relatively high, but this would not deter him from visiting Country C; if the price of tourism to this potential traveler is high in general (for all destination countries), he is less able to substitute for his planned expenditures since the alternative to traveling abroad is traveling within his home country or engaging in non-travel leisure activities at home.

To prepare forecasts based not on these general ranges of elasticities, but specific to the country in question, one may use an equation of the form,

The variables in the above equations are defined as follows:

  • TCR: travel credits (receipts) in real terms (deflated by PD, the price of domestic travel services;
  • TDR: travel debits (payments) deflated by the price of foreign travel services, PD*, expressed in domestic currency);
  • YR: domestic income in real terms (real GDP);
  • YR*: the sum of real income (real GDP) in the foreign countries from which most of the home country’s tourists originate;
  • PD: the average price of tourist services in the home country (possibly proxied by the GDP deflator);
  • PD*: the price of tourist services in foreign countries (possibly proxied by a weighted average of GDP deflators of partner countries, all expressed in terms of a single foreign currency, such as the U.S. dollar); and
  • ER: the bilateral exchange rate between the home country and the country of origin or destination of travelers, units of home currency per unit of foreign currency; if the exchange rate is in terms of U.S. dollars, then measures of foreign countries’ prices of tourist services must also be expressed in terms of U.S. dollars before they are combined into a weighted average to produce PD*.

The simple equations given above are limited by the use of GDP deflators as proxies for the prices of travel services, and by the omission of any measure of the cost of traveling from one country to another. It is usually necessary to assume that the prices of hotels and restaurants move in line with domestic prices (the GDP deflator) since systematic indices of prices in this sector are not generally available. The average price of substitutes is assumed to be given by an average of foreign GDP deflators. That is, foreigners will spend more on travel services in the home country if the average price of domestic output is low relative to the average price of foreign output, and obversely for domestic residents who travel abroad. In fact, the prices of travel services may move quite differently than prices of the rest of GDP; times of rapid hotel construction in the host country will tend to be accompanied by small increases, or absolute decreases, in the average cost of accommodation to tourists, and hotel managers may raise room rates on a weekend of a major soccer tournament. Air fares (part of Transport in the BOP accounts) have dropped sharply in recent years; omitting this variable will tend to result in estimates of the income elasticity that are biased upward.6

(4) Investment income

A simple approach to forecasting investment income can be derived from the identity that links interest payments and indebtedness,

interest due, IY, is equal to the amount of indebtedness, D, times the rate of interest, r. Writing this same identity for period t-I, subtracting it from the above expression, and re-expressing the result in terms of period-to-period changes (denoted by Δ), yields

This expression, still an identity, says that the change in interest payments may be decomposed into two terms, the level of the interest rate times the change in debt, and the change in the interest rate times the value of debt in the preceding period.

Suppose that the stock of a country’s net foreign debt that was acquired at a fixed rate of interest (“on fixed terms”) is given by (1-α); it follows that the second term in (8.6) should be preceded by the coefficient ± since current changes in interest rates will not affect interest payments that are due on past debt borrowed on fixed terms. Also, since r must be interpreted as the average yield on an economy’s foreign assets and liabilities, let us postulate a relation between this average and observable interest rates on global financial markets,

where ius is an interest rate in the U.S., iDM refers to the German market, and so forth. If an interest rate from a single foreign country is used to represent all of the i’s, or alternatively, some average of foreign rates with weights based on the past distribution of indebtedness among foreign currencies, equation (8.6) may be rewritten as

The expression given in equation (8.8) is suitable for econometric estimation. Since money loaned in period t earns interest only after some time has passed, both current and lagged values of ΔDt and Dt-1 may be included. In this form, the expression can be estimated and used for forecasting investment income payments, with a similar equation for receipts. Since data on gross assets probably cover only the short-term liquid items held by the banking system, this approach effectively omits medium- and long-term foreign assets, a drawback in countries in which they are held in large amounts. Also, implicitly it assumes that all foreign assets are held in the banking system and not by private individuals or enterprises.7

(5) Other services and transfers

To complete the forecast of goods and services, it is a common practice to add together all service receipts except travel, and to apply a similar approach to other service payments. Since the categories, Other Service Credits and Other Service Debits, are made up of diverse components, but together account usually for a small part of changes in the current account balance, it seems efficient to devise an aggregative method for forecasting them. Economic analysis does not apply to values, however; it treats volumes and prices separately. Even if remaining service items are not aggregated into credit and debit residuals, there is still a hurdle because countries usually do not prepare and publish price indices of services, even for total service flows.8

In this case, two alternative forecasting approaches may be used. In one case, the forecaster concocts a proxy price index for service flows—for example, a weighted average of the GDP deflator and the average price of exports of merchandise for service credits. The proxy index is used to deflate the value of Other Service Credits, and to form a relative price variable that can be used in a forecasting equation like the one described above for merchandise exports. A similar approach can be taken for Other Debits. The other approach is to relate Other Service Credits and Debits separately to values of merchandise exports and imports. The obvious motivation is that freight payments and receipts will tend to vary with the volume and price of trade. Moreover, forecasts using this method, although expressed in terms of values, will nevertheless respond to changes in relative prices such as follow from an exchange rate change; the second method will therefore extend to services the effects of exchange rate changes on merchandise trade, and possibly avoid understating the results.

Official and private transfers do not seem to be amenable to economic analysis in ways that result in simple forecasting equations. Here is where one may justifiably use naive models. If the value of transfers expressed in foreign currency does not vary much from year to year, one may base a forecast on the assumption that future flows will tend to equal an average of the recorded values of recent years. (“Typical” years should be used in computing the average, not drought years marked by large receipts of temporary humanitarian aid.) If transfers are growing or declining, one may substitute a trend for the assumption of no change. Alternatively, net transfers may be expressed as a fraction of GDP, to explore whether this ratio tends to be constant over time. What is recorded as private transfers may actually include some financial flows, and the resulting series may be highly variable. One may use an average or a trend, even in this case, in the absence of a fully articulated model of speculative short-term finance.

(6) Financial flows9

According to economic theory, financial flows are influenced by relative rates of return in various countries after allowing for expected changes in associated exchange rates. Savings will move to countries where interest rates are higher if the difference is enough to compensate for certain fees, information costs, and the risks of exchange rate change and default. It is usually difficult to translate these analytical relations into a statistical equation. In the first place, if the forecast period is a year, much of the variation in short term capital flows will not be observed—finance may flow in response to high rates early in the year and then flow out later in the year if domestic rates fall or other factors change, so that annual statistics hide the activity that has occurred. For a second reason, the confidence and expectations of asset holders, presumably a major determining factor in speculative flows, cannot be measured directly (and if it could be, behavior would likely change in response to the information becoming well known).

It is nevertheless useful to make a forecast of the less variable part of financial flows, usually long-term finance including direct foreign investment. Here the approach is to use a naive model, as in the case of transfer flows described in the preceding subsection. The value of net longer-term financial inflows (subtracting amortization payments from gross inflows or outflows, and excluding highly variable flows, usually short-term) can be compared with the value of domestic GDP (measured in the same currency). If this share has some tendency to remain stable, then an average of the ratio over recent, normal periods provides a method of forecasting inflows in the forecasting period. If policies in the forecast year are likely to be more dedicated to achieving macroeconomic stability than in the past, a judgmental adjustment may be made to the forecast to allow for the affect of a change in stability on investors’ expectations, and conversely if stability is expected to decline. If nominally short-term net flows, and Net Errors and Omissions, appear to be about as regular as long-term finance, it would be logical to include these flows in the total net amount to be forecast in the absence of an expected sudden change in the orientation of macroeconomic policy.

If the country has already accumulated substantial foreign debt and has had difficulty servicing it, there may be little or no foreign finance flowing in spontaneously. In this case, the nature of inflows may relate more to the government’s need to borrow. If so, the only reasonable forecasting method is actual knowledge of the government’s borrowing needs and of the likely creditors and their attitudes toward further lending.

c. Regression equations for goods, services, and income

(1) Merchandise exports

Coefficients of export forecasting equations and related statistics are given in Table 8.4. The first two lines in the table are based on a simple linear trend, and the other four incorporate behavioral relationships from economic theory as discussed in the preceding section. The dependent variable is real exports (merchandise exports f.o.b. in millions of U.S. dollars, deflated by an export price index equal to one on average in 1984-8610).

The trend equations show that it possible to account for around 85 to 90 percent of the variation in real exports simply by the assumption of a constant yearly increase. The estimates based on behavioral relations account for a higher percentage, 96 to 98 percent. Shortening the sample for trend estimation from 1970-95 to 1984-95 leads to a somewhat higher estimate of average growth of export volume, consistent with the goal of the reforms undertaken during the early 1980s. Performance of the naive equation and a behavioral equation (8.11), both for shorter and longer samples, is shown in Figure 8.1.

Table 8.1.Turkey: Merchandise Exports, 1975-95
19911992199319941995
(U.S. dollars, millions)
Export values
Agricultural products2,7252,2592,3812,4712,314
Mining and quarrying286264238272406
Manufactures10,58312,19012,72615,36318,916
Food industry1,1141,2001,2681,6151,940
Textiles4,3285,3125,4526,3668,264
Iron and steel1,4521,5582,0112,3692,257
Refined petroleum277231172235277
Other3,4123,8893,8234,7786,178
Total exports, f.o.b., customs basis13,59414,71415,34518,10621,636
Non-customs exports73177266284339
Total exports, BOP basis13,66714,89115,61118,39021,975
Source: Republic of Turkey, Prime Ministry, State Institute of Statistics Monthy Bulletin; Central Bank, Quarterly Bulletin; and staff calculations.

Shares of total exports on customs basis.

1975-791980-841985-891990-941995
AverageAverageAverageAverage
(Figures in percent)
Export shares 1/
Agricultural products61.836.920.316.410.7
Mining and quarrying6.23.83.21.91.9
Manufactures32.059.476.581.887.4
Food industryn.a.8.77.08.09.0
Textilesn.a.20.826.734.238.2
Iron and steeln.a.5.611.112.010.4
Refined petroleumn.a.4.32.81.61.3
Othern.a.20.028.926.028.5
Source: Republic of Turkey, Prime Ministry, State Institute of Statistics Monthy Bulletin; Central Bank, Quarterly Bulletin; and staff calculations.

Shares of total exports on customs basis.

Source: Republic of Turkey, Prime Ministry, State Institute of Statistics Monthy Bulletin; Central Bank, Quarterly Bulletin; and staff calculations.

Shares of total exports on customs basis.

Table 8.2.Turkey: Merchandise Imports, 1980-95
19911992199319941995
(U.S. dollars, millions)
Import values
Intermediate goods12,08513,12715,74613,59620,807
Crude oil2,4562,6322,5492,4323,017
Other9,62910,49513,19711,16417,790
Investment goods6,0516,7739,5656,89410,488
Consumer goods2,9102,9714,1162,7804,414
Total imports, c.i.f., customs basis21,04722,87129,42823,27035,709
Nonmonetary gold1,1611,4301,8814801,322
Transit trade64151229251301
Freight, insurance-1,265-1,371-1,767-1,395-2,145
Total imports, f.o.b., BOP basis21,00723,08129,77122,60635,187
Source: Republic of Turkey, Prime Ministry, State Institute of Statistics, Monthly Bulletin; Central Bank, Quarterly Bulletin; and staff calculations.

Shares of total imports on customs basis.

1980-841985-891990-941995
AverageAverageAverage
(Figures in percent)
Import shares 1/
Intermediate goods73.065.257.258.3
(Crude oil)35.819.111.48.4
(Other)37.246.145.849.8
Investment goods24.326.629.529.4
Consumer goods2.78.313.312.4
Source: Republic of Turkey, Prime Ministry, State Institute of Statistics, Monthly Bulletin; Central Bank, Quarterly Bulletin; and staff calculations.

Shares of total imports on customs basis.

Source: Republic of Turkey, Prime Ministry, State Institute of Statistics, Monthly Bulletin; Central Bank, Quarterly Bulletin; and staff calculations.

Shares of total imports on customs basis.

Table 8.3.Turkey: Exchange Rates, 1975-95
Nominal, BilateralNominalReal
Liras perLiras pereffective, 1/effective, 1/2/
U.S. dollar,US. dollar,indexindex
Yearend-of-periodperiod average(1990=100)(1990=100)
197515.1514.44
197616.6616.05
197719.4418.00
197825.2524.288,450.4
197935.3531.086,394.3159.9
198090.1476.042,655.9124.4
1981133.62111.222,112.0123.5
1982186.75162.551,611.0113.1
1983282.80225.461,272.0109.3
1984444.73366.68872.7103.8
1985576.86521.98652.3105.8
1986757.79674.51417.287.9
19871,020.90857.21293.482,0
19881,814.841,422.35178.781.2
19892,313.692,121.68128.288.5
19902,930.072,608.64100.0100.0
19915,079.924,171.8166.62102.2
19928,564.436,872.4240.3298.6
199314,472.5210,984.6328.64107.5
199438,726.0029,608.6811,9580.7
199559,650,0045,845.066.6786.1
Sources: IMF,International Financial Statistics; and staff calculations.

Period average. A decrease in the index corresponds to a depreciation of the lira. The average includes all Fund member countries supplying relevant data; weights are based on Turkish import and export shares. (An alternative index, based on eight industrial country partners, is given in Table I.1).

A decrease corresponds to a depreciation in real terms as measured by relative changes in Turkish and partners’ consumer price indices.

Sources: IMF,International Financial Statistics; and staff calculations.

Period average. A decrease in the index corresponds to a depreciation of the lira. The average includes all Fund member countries supplying relevant data; weights are based on Turkish import and export shares. (An alternative index, based on eight industrial country partners, is given in Table I.1).

A decrease corresponds to a depreciation in real terms as measured by relative changes in Turkish and partners’ consumer price indices.

Table 8.4.Turkey: Regression Estimates for Forecasting Export Volume 1/
CoefficientsSummary indicators
EquationConstant
NumberTrendGDPRt-1GAPRPX2XRt-1termR2SEEDWSample
(8.9)586-1,4370.911,4240.51970-95
(15.7)(2.5)
(8.10)731-4,0710.861,0991.51984-95
(8.0)(2.1)
(8.11)0.256-0.039247.7-16,6780.978331.01971-95
(26.8)(0.5)(4.9)(8.9)
(8.12)0.177-0.055634.40.351-11,8900.986882.01972-95
(5.2)(0.9)(3.5)(2.5)(4.5)
(8.13)0.301-0.081161.8-22,3750.966592.21984-95
(9.3)(0.9)(3.1)(4.0)

The dependent variable is XR, defined below. Figures in parentheses are t values. The algebraic derivations of independent variables included in these regressions are as follows:

  • XR: X_D/(PX_D/100);
  • TREND: 1 in 1970, 2 in 1971, and so forth;
  • GAP: GDPR - exp[l0.428 + 0.0415558 TREND];
  • PGDP: GDP/GDPR; PGDP = 1 in 1987;
  • RPX: PX_D/(PGDP/(NERA/1000));
  • RPX2: RPXt + RPXt-1.

In GAP, the expression within brackets results from regressing the log of GDPR on TREND over the period 1970-95. See Chapter 6.

The dependent variable is XR, defined below. Figures in parentheses are t values. The algebraic derivations of independent variables included in these regressions are as follows:

  • XR: X_D/(PX_D/100);
  • TREND: 1 in 1970, 2 in 1971, and so forth;
  • GAP: GDPR - exp[l0.428 + 0.0415558 TREND];
  • PGDP: GDP/GDPR; PGDP = 1 in 1987;
  • RPX: PX_D/(PGDP/(NERA/1000));
  • RPX2: RPXt + RPXt-1.

In GAP, the expression within brackets results from regressing the log of GDPR on TREND over the period 1970-95. See Chapter 6.

Figure 8.1.Turkey Merchandise Trade

(Monthly data; 1990 = 100)

Figure 8.2Turkey Real Effective Exchange Rate

(Value in millions of US dollars)

Source: International Monetary Fund, staff estimates.

Figure 8.3.Turkey Export Volume, Actual and Fitted Values

(In millions of average 1984-86 U.S. dollars

Source: International Monetary Fund, staff estimates.

In the behavioral equations, exports in real terms are related to the lagged value of real GDP, a measure of excess demand, and export prices relative to the GDP deflator. The representation of the relative price of exports (RPX2) is actually the sum of the unlagged and lagged terms (RPXt + RPXt-1) since, when they were included separately, the estimated coefficients were almost exactly the same magnitude. Excess demand is measured as the difference between real GDP and its trend using the same estimate as introduced for potential output in the workshop on output and prices.

All of the estimated coefficients in the behavioral equations have the expected signs although the effect of excess demand on export volume is weak. The Durbin-Watson Statistic indicates a variable or variables may have been omitted from equation (8.11); including the lagged dependent variable as a regressor (equation (8.12)) suggests lagged effects are missing in (8.II).11 This allowance for lags is not needed if the sample for estimation is restricted to the post-reform period, 1984-95, as in equation (8.13). (If the lagged value of real exports is included in this case, it has an insignificant coefficient, not shown in Table 8.4.) In fact, the sizes of coefficient estimates in equations (8.11) and (8.13) are rather similar, casting some doubt on the transforming effects of payment reforms and special incentives on export performance. Since there is little shift in the equation itself due to reforms, the higher growth in exports can be attributed to the values of the independent variables—the more depreciated exchange rate and the growth in export supply.

(2) Merchandise imports

Estimation results are given in Table 8.5. The trend equation for the shorter, recent period has a substantially steeper slope than the results for the longer sample; import growth has been higher in recent years. A linear trend can account for close to 85 percent of the variation in real imports, but the regression equations based on behavioral hypotheses are able to account for more than 90 percent. The effort to describe past variation in volumes with a simple regression model is less successful for imports than for exports. This is shown in Figure 8.4.

There is a significant positive forecast error in 1995, the final historical year, according to Figure 8.4.b. This can pose a problem in forecasting 1996. If the unusual event or events that boosted imports in 1995 are temporary, then the regression estimates of the coefficients of the behavioral relation will provide a reasonable forecast of real imports in 1996. If, however, the positive error in 1995 signals some permanent structural change beginning in 1995 (one that necessarily is not incorporated in the specification of the equation), then the coefficient estimates will under-forecast again in 1996. Some judgment is required in applying the regression results in an actual forecasting situation.

There are no important lags in the behavioral results with the exception of relative prices; when the lagged value of the dependent variable was included on the right-hand side, its coefficient was not significant (not shown in Table 8.5). The behavioral equations incorporate the analytical links described in the preceding section: import demand in real terms varies with real income; and when total demand in the economy is high, relative to a measure of potential output, there will be extra “spill-over” demand for foreign goods to supplement the domestic supply. In equation (8.16), the relative-price term has the correct sign though the coefficient is not significant: when the price of imports rises relative to the prices of import substitutes produced at home, the volume of imports demanded is smaller.

In equation (8.17), the same as (8.16) but estimated with a shorter sample of post-reform years, the coefficient on the relative price of imports becomes both positive and significant. (Testing indicated that only the lagged import-price term was significant in the sample of recent years; in the long sample, the measure of relative prices has been made equal to the simple sum of unlagged and lagged values, as in the case of exports and for the same reason.).

It is known that administrative restrictions on imports (including tariffs, the list of items requiring licenses, and advance deposit percentages) were varied frequently in the more recent sub-sample; both increases and decreases occurred (see Part I, Section 2). As already discussed, since these measures are often specified to affect individual commodities or groups of items for irregular time periods, it is a nearly impossible task to construct an index of the price-equivalent effects of such microeconomic changes for total real imports. One hypothesis suggested by the coefficient on RPMt-1 in equation (8.17) is that years of major exchange-rate devaluations were accompanied by significant liberalizations, so that increases in imports followed increases in the domestic price of foreign goods. A further limitation of these empirical results involves exchange rate expectations; the response of importers and exporters to the current level of the real rate depends on whether they expect it to persist.

Table 8.5.Turkey: Regression Estimates for Forecasting Import Volume 1/
CoefficientsSummary indicators
EquationConstant
NumberTrendGDPRGAPRPM2RPMt-1termR2SEEDWSample
(8.15)0.7371,1450.842,5121.01970-95
(11.2)(1.1)
(8.16)1.2549,2060.832,1623.01984-95
(6.9)(2.4)
(8.17)0.3120.129-14.02-6,2770.951,4771.41971-95
(17.8)(0.9)(1.3)(4.1)
(8.18)0.4870.243120-32,7170.961,1141.81984-95
(10.9)(1.8)(3.1)(5.1)
(8.19)0.3780.170-13,9590.921,5621.91984-95
(9.6)(0.9)(4.4)

The dependent variable is MR, defined below. Figures in parentheses are t values. The algebraic derivations of independent variables included in the equations are as follows (see also the footnote to Table 3.4):

  • MR: M_D/(PM_D/100);
  • RPM: PM_D/(PGDP/(NERA/1000));
  • RPM2: RPMt + RPMt-1.

The dependent variable is MR, defined below. Figures in parentheses are t values. The algebraic derivations of independent variables included in the equations are as follows (see also the footnote to Table 3.4):

  • MR: M_D/(PM_D/100);
  • RPM: PM_D/(PGDP/(NERA/1000));
  • RPM2: RPMt + RPMt-1.

Figure 8.4.Turkey Import Volume, Actual and Fitted Values

(In millions of average 1984-86 U.S. dollars)

Source: International Monetary Fund, staff estimates.

The same pattern in price coefficients between longer and more recent samples repeated itself in most of the variants of the import demand specification that were estimated. The results from the recent, shorter samples (the positive sign on RPM) cannot be used for forecasting unless one is prepared to posit that there will always be a link between devaluations and liberalizations. Accordingly, in equation (8.18) the relative price term is omitted. A judgmental allowance may be made for the effects of relative price changes in cases in which restrictions do not change. An elasticity of about 0.5 is reasonable for agricultural goods (except tree crops), 1.0 for manufactured goods, in the short run (effects occurring in the same year); longer-run elasticities should be half again as large.12

(3) Travel receipts and payments

Regression estimates are given in Table 8.6. They are generally reasonable but not strong, and forecasts may be based instead on the discussion of probable elasticities given in the preceding section.

Equation (8.19) is in terms of levels of the variables (that is, not in logarithms, unlike subsequent credit equations in the table). The fit is reasonable but none of the coefficients is significantly different from zero at a high confidence level except the lagged dependent variable.13 In the case of travel debits in level form (equation (8.22)), the income variable is significant as well as the lagged dependent variable (moderately significant). The logarithmic form of these equations was also estimated in order to estimate elasticities directly; using logs results in more stable coefficient estimates for the receipts equation, weaker estimates in the case of payments.

The log equation for travel credits estimated from the longer sample, equation (8.20), includes the lagged value of real travel credits as a regressor. The customary transformation to obtain an estimate of long-run elasticity yields a value of about three for income14 and about -22/3 or relative prices, as anticipated in the discussion in the preceding section. When the sample is shortened to the post-reform period (equation (8.21)), the lagged value of the dependent variable is no longer significant. The elasticity on income increases to about four, and the elasticity on relative prices drops (in absolute value) to about 1½. Neither is as plausible as the coefficients from the longer sample,15 although they may be used for forecasting.

Table 8.6.Turkey: Regression Estimates for Forecasting Travel Volume 1/
CoefficientsSummary indicators
Equation NumberDependent VariableLagged Dependent VariableIncomeRelative PriceLagged Relative PriceConstant termR2SEEDWSample
(8.19)Travel credits, levels0,6051,698-381-1,3050.933212.11975-95
(2.8)(1.3)(1.0)(0.7)
(8.20)Travel credits, logs0.5591.37-1.172.890.980.1732.61975-95
(4.4)(1.5)(3.9)(4.4)
(8.21)Travel credits, logs4.01-0.669-0.8185.880.920.1772.71984-95
(5.3)(1.4)(1.7)(15.0)
(8.22)Travel debits, levels0.3280.00722-170-95.80.651151.91975-95
(1.5)(2.2)0.7(0.9)
(8.23)Travel debits, logs0.4550.971-0.0432-7.600.610.3921.91975-95
(2.2)(1.2)(0.1)(0.9)
(8.24)Travel debits, logs1.15-0.512-6.850.490.1782.01984-95
(2.9)(1.1)(1.5)

Travel credits and debits and domestic and foreign income are in real terms. Figures in parentheses are t values. When the dependent variable is in logs (see column 2, equations (21), (22), (24), and (25)), the associated independent variables are also in logs. Variables have been defined as follows:

real travel credits: TCR = (TC_D * (NERA/857.2))/PGDP; (NERA/857.2)/PGDP = 1 in 1987;

real travel debits: TDR = -TD_D/(lCPGDP_D/3.776); ICPGDP_D/3.776 = 1 in 1987;

real domestic income: GDPR;

real foreign income: ICGDPR;

relative price of travel credits: RPTC = (PGDP/(NERA/857.2))/(ICPGDP_D/3.776);

relative price of travel debits: RPTD = 1/RPTC.

Travel credits and debits and domestic and foreign income are in real terms. Figures in parentheses are t values. When the dependent variable is in logs (see column 2, equations (21), (22), (24), and (25)), the associated independent variables are also in logs. Variables have been defined as follows:

real travel credits: TCR = (TC_D * (NERA/857.2))/PGDP; (NERA/857.2)/PGDP = 1 in 1987;

real travel debits: TDR = -TD_D/(lCPGDP_D/3.776); ICPGDP_D/3.776 = 1 in 1987;

real domestic income: GDPR;

real foreign income: ICGDPR;

relative price of travel credits: RPTC = (PGDP/(NERA/857.2))/(ICPGDP_D/3.776);

relative price of travel debits: RPTD = 1/RPTC.

The results for real travel debits are noticeably weaker although this flow is also smaller. As expected, the long-run elasticities for travel expenditure by Turkish residents are smaller than the elasticities in the credit equations. In the log regression with the longer sample (8.23), the income elasticity is 1.8 after transforming to long-term basis, and the price elasticity is virtually zero. In results from the post-reform sample, the two elasticities are about 1 and -0.5, respectively. As in the case of merchandise imports, imports of travel services may be highly influenced by payments restrictions so that a strong response to measurable changes in the relative price is not observed. If travel abroad by Turkish residents is largely for business rather than pleasure, this demand for travel services may be price inelastic.

(4) Investment income

The equations introduced for investment income in the preceding section are identities and may be used directly to generate forecasts. Forecast errors will result from lags, errors in the data (if IY includes dividends as well as interest), and an unstable or unobservable ratio of average yields on gross foreign assets and liabilities to world interest rates. Regression estimation will tend to minimize two of these sources of error; the econometric equations can be specified to allow for lags and for a constant ratio of implicit average yields to observable interest rates. Regression estimates are given below:

R2 = 0.75 SEE = 95.11 D.W. = 3.6 1984-95.

R2 = 0.56 SEE = 133.8 D.W. = 2.4 1981-95.

The independent variables in the above expressions are defined as follows:

RDD: (USINT/100.)*ΔDEBT_D; INT is in percent, and DEBT is in U.S. dollars, millions;

DRD:(ΔUSINT/100.)*DEBT_Dt-1;

RDA.(USINT/100.)*ΔFA_D;FA_D is in U.S. dollars, millions;

DRA:(ΔUSINT/100.)*FA _Dt-1.

The logic of the above equations suggests that the constant terms be suppressed, but they are not highly significant, and forcing the regression plane through the origin will reduce the goodness of fit. A constant difference in percentage points between the interest rate paid on foreign borrowing (or lending) and the reference U.S. three-month deposit rate (see equation (8.7) in the preceding section) would provide a basis for a constant term.

Lags are significant in the equation for investment income payments, and in fact the coefficients on the lagged terms have similar magnitudes to the unlagged terms; they could be added together to reduce the number of regressors. In the equation for investment income receipts, the lag coefficients are not significantly different from zero.

By the interpretation suggested in the preceding section, the average interest rate paid by Turkey on foreign borrowing is only about half of the reference rate (the coefficients on the RDD terms sum to about 0.5 in equation (8.25)). Such a low figure is unlikely, but these coefficients are sensitive to concessional rates and grace periods on official loans. The other coefficients in the payments equation suggest that about a quarter of Turkey’s foreign borrowing is at variable terms (adding the coefficients on lagged and unlagged DRD), whereas this figure is reported elsewhere to be closer to one third. On the asset side, the coefficient on RDA suggests a low yield on assets, a third or so of the reference interest rate, which is plausible. The second coefficient can be interpreted as representing the share of foreign assets at variable interest rates, a little higher than 40 percent.

(5) Other services credits and debits

Simple regression equations are given below. The coefficients and other regressions statistics are quite satisfactory. The equations utilize the second of the methods introduced in the preceding section—the miscellaneous groupings contain transport services and have been related to the values of merchandise exports and imports.

R2 = 0.97 SEE = 523 D.W. = 2.1 1975-95.

R2=0.98 SEE=165 D.W.=2.0 1975-95.

All of the variables in these equations are as defined in the workshop database. Miscellaneous service receipts, including transport receipts, tend to average about a quarter of the value of merchandise exports, and miscellaneous debits are about a tenth of imports in the short run. If the coefficients are transformed to long-term basis, OSC tends to increase by 0.54 of the increase in export value, OSD by 0.14 of the increase in imports. (Approximately the same magnitudes were estimated when the OS variables were regressed on current and lagged values of the respective trade series and the coefficients were summed.)

d. Exercises and issues for discussion

  • (1) Prepare a forecast of the balance of payments for 1996, completing the main items in Tables 8.7, 8.9, 8.10, and the first section of 8.11. The projections should be consistent with those developed in other workshops, in particular Prices, Output and Expenditure. Use the information on world market developments provided below in 8.13. The reduction in tariffs on imports that will come into effect at the beginning of the forecast year will cause a decrease in the average lira price of imports of about 1.1 percent.
  • (2)List the main assumptions and policies underlying your forecasts.
  • (3)Review the acceptability of the balance of payments projection prepared for 1996. To the extent that the forecasts indicate problems (for example, the financing requirement in unrealistically high, pointing to the emergence of a financing gap), what are the major options available to the authorities?
  • (4)Comment on the possible impact of changes in the following items on the forecasts for the current and financial accounts:
    • domestic credit;
    • domestic consumption;
    • the exchange rate;
    • foreign or domestic interest rates.
Table 8.7.Turkey: Balance of Payments, 1993-96(In millions of U.S. dollars
1996
199319941995Sectoral ForecastBaseline
Trade balance-14,160-4,216-13,212
Exports, f.o.b.15,61118,39021,975
Imports, f.o.b.-29,771-22,606-35,187
Services, net 1/6,7047,0199,582
Credits10,65210,80114,606
Debits-3,948-3,782-5,024
Net investment income-2,745-3,264-3,205
Transfers, net3,7683,0924,496
Private3,0352,7093,425
Official7333831,071
Current Account Balance-6,4332,631-2,339
Financial Account, net 2/8,963-4,1944,722
Long-term finance5,9099332,417
Direct investment622559772
Portfolio investment3,9171.1581,724
of which:3,9171.1581,724
Equity431994120
Other investment1,370-784-79
Short-term finance3,054-5,1272,305
Net Errors and Omissions-2,2221,7692,275
Overall balance3082064,658
Total change in reserves-308-206-4,658
IMF purchases0340347
Official reserves 3/-308-546-5,005
Foreign exchange assets-314-625-5,032
Monetary gold67927
Memorandum items
Liras per U.S. dollar, end-of-period14,47338,72659,650
Liras per U.S. dollar, average period10,98529,60945,845
Source: IMF, Turkey—Recent Economic Developments, 1996.

Formerly called “nonfactor services.”

In Turkey’s BOP accounts, capital account items (new definition) are zero or not shown separately.

A negative sign signifies an increase (inflow).

Source: IMF, Turkey—Recent Economic Developments, 1996.

Formerly called “nonfactor services.”

In Turkey’s BOP accounts, capital account items (new definition) are zero or not shown separately.

A negative sign signifies an increase (inflow).

Table 8.8.Balance of Payments, 1993-96(In billions of Turkish liras)
1993199419951996
Sectoral ForecastBaselineProgram
Trade balance-155,548-124,832-605,704
Exports, f.o.b.171,487544,5101,007,444
Imports, f.o.b.-327,034-669,341-1,613,148
Services, net 1/73,643207,826439,287
Credits117,012319,807669,612
Debits-43,369-111,981-230,325
Transfers, net41,39191,551206,119
Private33,33980,211157,019
Official8,05211,34049,100
Net investment income-30,154-96,644-146,933
Current Account Balance-70,66777,901-107,231
Financial Account, net 2/98,459-124,180216,480
Long-Term Capital64,91027,625110,807
Direct investment6,83316,55135,392
Portfolio investment43,02834,28779,037
Other investment15,049-23,213-3,622
Short-Term Capital33,548-151,805105,673
Net Errors and Omissions-24,40952,378104,297
Overall balance3,3836,099213,546
Total change in reserves-3,383-6,099-213,546
IMF purchases010,06715,908
Official reserves 3/-3,383-16,167-229,454
Foreign exchange assets-3,449-18,506-230,692
Monetary gold662,3391,238
Memorandum item:
Liras per U.S. dollar:
End of period14,47338,72659,650
Period average10,98529,60945,845
Source: Based on Table 8.7.

Formerly called “nonfactor services.”

“Finance” and “capital” are used interchangeably.

A negative sign signifies an increase (inflow).

Source: Based on Table 8.7.

Formerly called “nonfactor services.”

“Finance” and “capital” are used interchangeably.

A negative sign signifies an increase (inflow).

Table 8.9.Turkey; Trade in Services and Income, 1993-96(In millions of U.S. dollars)
1993199419951996
Sectoral ForecastBaselineProgram
Service credits10,65210,80114,606
Travel3,9594,3214,957
Other credits6,6936,4809,649
Of which: transport1,2411,2211,712
Investment income credits 1/1,1358901,488
Transfer credits3,8003,1134,512
Official7654041,087
Private3,0352,7093,425
Of which: workers’ remittances2,9192,6273,327
Service debits-3,948-3,782-5,024
Travel-934-866-911
Other debits-3,014-2,916-4,113
Investment income debits 1/-3,880-4,154-4,693
Transfer debits-32-21-16
Official-32-21-16
Private
Of which workers’ remittances
Source: Balance of Payments Statistics Yearbook, 1996.

Excluding reinvested earnings, for which data are not available.

Source: Balance of Payments Statistics Yearbook, 1996.

Excluding reinvested earnings, for which data are not available.

Table 8.10.Turkey: Foreign Merchandise Trade: Value and Volume, 1993-96(Percentage change from previous year, in U.S. dollars)
1993199419951996
Sectoral ForecastBaselineProgram
Exports 1/
Value, customs basis4.318.019.5
Price 2/2.8-12.819.2
Volume1.535.30.3
Imports 3/
Value, customs basis28.7-20.953.5
Price 2/-7.68.215.1
Volume39.2-26.933.4
Oil and oil products imports
Value3.7-5.325.0
Price-14.8-2.17.6
Volume21.7-3.316.2
Terms of trade11.2-19.43.6
Source: IMF, International Financial Statistics.

Excluding transit trade.

The trade-price indices are Fischer-type, equal to the geometric mean of Laspeyres and Paasche indices.

Excluding transit trade and nonmonetary gold.

Source: IMF, International Financial Statistics.

Excluding transit trade.

The trade-price indices are Fischer-type, equal to the geometric mean of Laspeyres and Paasche indices.

Excluding transit trade and nonmonetary gold.

Table 8.11.Turkey: External Debt, 1993-96(In millions of U.S. dollars; end of period)
1993199419951996
Sectoral ForecastBaselineProgram
(By maturity)
Total outstanding67,35665,60173,278
Medium- and long-term48,82354,29157,577
Short-term18,53311,31015,701
(By borrower)
Medium- and long-term48,82354,29157,577
Government36,23739,55039,175
Central bank6,6188,59710,486
Private sector5,9686,1447,916
Short-term38,53311,31015,701
Central bank667828993
Deposit money banks11,1274,6846,659
Other sectors6,7395,7988,049
(By creditor)
Multilateral agencies8,6749,1839,081
Of which: IMF344573
Bilateral lenders (countries)18,15320,67821,558
Deposit money banks, other private40,52935,74042,639
(By type of credit)
Medium- and long-term48,82354,29157,577
Of which: Project, program credits21,76025,21927,298
Bond issues12,62313,78810,486
Private credits11,34912,95017,438
Short-term18,53311,31015,701
Of which: Deposits3,0973,2664,471
Foreign exchange deposits2,4312,4433,498
Dresdner scheme666823973
Convertible deposits
Total Debt Service8,2279,99311,897
Interest3,5743,9234,303
Medium-, long-term repayments4,6536,0707,594
Debt Service Ratio 1/273129
Memorandum Items:
Average interest rate, medium-
and long-term debt7.37.27.5
Average maturity of medium- and
long-term debt (years)12.512.512.7
Currency composition of debt (in % of total)
U.S. dollar38,333.034.0
Deutsche mark29.933.934.8
Japanese yen19.820.519.2
Source: IMF, Turkey—Recent Economic Developments, 1996.

Interest plus medium- and long-term debt repayments as percent of current account receipts excluding official transfers.

Source: IMF, Turkey—Recent Economic Developments, 1996.

Interest plus medium- and long-term debt repayments as percent of current account receipts excluding official transfers.

Table 8.12.Turkey: Financial Flows, 1993-96(In millions of U.S. dollars)
1993199419951996
Sectoral ForecastBaselineProgram
Long-term finance5,9099332,417
Direct investment622559772
Portfolio investment3,9171,1581,724
Other1,370-784-79
Official sector (incl. central bank)-930-1,461-537
Drawings525365723
Dresdner deposits9251,3151,462
Amortization-2,380-3,141-2,722
Deposit money banks193-282273
Other sectors (incl. private)2,107959185
Short-term finance3,054-5,1272,305
Assets-3,2912,423-1,791
Of which:
Loans extended-289-38293
Deposit money banks-2,8942,451-2,030
Liabilities6,345-7,5504,096
Central bank193168279
Of which:
Dresdner deposits144115101
Bankers’ credits
Deposit money banks4,302-6,7711,700
Foreign exchange deposits520-170899
Foreign exchange credits3,782-6,601801
Other sectors1,850-9472,117
Of which:
Trade credits2,244-8161,671
Foreign exchange credits-394-131446
Source: IMF, Turkey-Recent Economic Developments, 1996.
Source: IMF, Turkey-Recent Economic Developments, 1996.
Table 8.13.Turkey: Selected Indicators of World Demand and Prices, 1994-96(Percentage change unless otherwise indicated)
199419951996
Industrial countries
Real GNP2.52.2
GNP deflators (in U.S. dollars)10.31.9
Trading partner countries
Real GDP 1/2.42.9
Volume of imports1/
Goods and services8.26.1
Goods5.75.6
Goods excluding petroleum5.85.3
World prices (in U.S. dollars)
Manufacturing export prices9.60.6
Petroleum spot price8.0-4.3
Turkey’s trade prices (in U.S. dollars)
Exports 1/9.30.4
Nonfuel primary commodities 1/2.1-2.2
Imports 2/9.4-0.5
Nonfuel primary commodities13.5-3.7
Interest rates (percent per year)
6-month LIBOR on U.S. dollar deposits6.15.6
U.S. prime rate7.18.88.3
Memo items:
Nominal effective exchange rate of U.S. dollar 3/-6.0
Scheduled repurchases from IMF (U.S. dollars, millions)00
Scheduled amortization of external debt (U.S. dollars, millions)
Official sector (including central bank)27222964
Private sector and others48722721
Source: IMF, International Financial Statistics (IFS); World Economic Outlook data base, “Global Economic Environment Indicators” table, end-1995; and IMF, Turkey - Recent Economic Developments (Washington: IMF, 1997).

Export weighted (shares of Turkey’s exports).

Import weighted (shares of Turkey’s imports).

Period average, from IFS. Decrease indicates devaluation.

Source: IMF, International Financial Statistics (IFS); World Economic Outlook data base, “Global Economic Environment Indicators” table, end-1995; and IMF, Turkey - Recent Economic Developments (Washington: IMF, 1997).

Export weighted (shares of Turkey’s exports).

Import weighted (shares of Turkey’s imports).

Period average, from IFS. Decrease indicates devaluation.

1

International Monetary Fund, World Economic Outlook (Washington: IMF, biannual); Organization for Economic Cooperation and Development, OECD Economic Outlook (Vans. OECD, biannual).

2

More comprehensive surveys are presented in Edward E. Leamer and R. M. Stern, Quantitative International Economics (Chicago: Aldine, 1970), and David F. Heathfield, ed., Topics in Applied Macroeconomics (London: Macmillan, 1976), pp. 144-163. See also Morris Goldstein and Mohsin S. Khan, "Income and Price Effects in Foreign Trade," Chapter 20 in Ronald W. Jones and Peter B. Kenen, eds., Handbook of International Economics, Volume II (New York: North Holland, 1985).

3

As implied by the general expression, Country A’s export volume is affected in the same way whether the global prices of its exportable products change (measured, say, in U.S. dollars), or the global price remains constant while the nominal exchange rate changes. The numerator of the fraction expresses world prices in terms of domestic currency, and the two factors have equivalent effects.

4

See, for example, Morris Goldstein and Mohsin S. Khan, “The Supply and Demand for Exports: A Simultaneous Approach,” Review of Economics and Statistics, 1978, pp. 413-27.

5

Jacques R. Artus, “An Econometric Analysis of International Travel,” IMF Staff Papers (Washington: International Monetary Fund, November, 1972).

6

This assertion holds if air fares in real terms had a generally negative trend during recent years, and real personal incomes had a positive trend.

7

Interest paid on foreign-currency-denominated bank deposits (FCD’s) is included in investment income in the balance of payments only if the deposits are held by nonresidents.

8

Implicitly there must be deflators somewhere in the national accounts system since external trade in goods and services comprises a part of real GDP. However, these deflators in many cases may simply be “borrowed” from the price indices used for merchandise alone, or all of services may be deflated with a simple average of a small sample of hotel and restaurant quotations. It would be quite expensive to collect data and construct price indices of foreign trade in services especially since in some cases it is difficult to distinguish the quantity of service from its value (for example, book royalties).

Besides travel, services include insurance and freight on air, land, and sea shipments of goods, the cost of loading and unloading carriers, passenger fares of all kinds, expenditure while ashore of flight and marine crews, purchases of fuel by ships and aircraft, local expenses of embassies and consuls, telecommunications and postal services, construction services, financial services, business services including computers and information technology, the rent due to or from owners of land and buildings, royalties, commissions, and leasing fees. See International Monetary Fund, 1993, Balance of Payments Manual (Washington: International Monetary Fund, 5th ed.).

9

Formerly called “capital” flows.

10

Algebraic transformations of variables are given in the footnotes to the tables, or in the text, in this section. Definitions of the time series used in the transformations, including unit and currency, are given in the data appendix to this volume.

11

To compare coefficients in equations (8.11) and (8.12) it is appropriate to express both sets as long-run values. To do so, divide the coefficients on behavioral independent variables in equation (8.12) by one minus the coefficient on lagged export volume, (1 - 0.35).

12

Compare Goldstein and Khan (already cited), Table 4.1 on page 1,079.

13

These results imply that the variables are colinear. They tend to increase and decrease together, and their separate influences cannot be distinguished.

14

The coefficient on income is divided by one minus the coefficient on the lagged dependent variable, 1.37/(l-.559), which is equal to 3.1.

15

In equation (8.21), both lagged and unlagged relative-price terms were retained, although neither is highly significant, because of plausibility; it seems likely that a price change affecting tourist services in the current year may not influence those persons whose travel plans for the coming vacation season are already made, perhaps even prepaid, but will be fully taken into account by the following year.

    Other Resources Citing This Publication