Journal Issue
Share
Article

The World Trade Model: Invisibles

Author(s):
International Monetary Fund. Research Dept.
Published Date:
January 1979
Share
  • ShareShare
Show Summary Details

The purpose of this paper is to present a model of imports and exports of invisibles for the 14 major industrial countries: Austria, Belgium (including Luxembourg), Canada, Denmark, France, the Federal Republic of Germany, Italy, Japan, the Netherlands, Norway, Sweden, Switzerland, the United Kingdom, and the United States. The model is the second of two parts of the world trade model developed by the Fund’s Research Department to explain current account transactions of industrial countries. The first part deals with merchandise trade flows and was presented in Deppler and Ripley (1978). The present model is similar to its counterpart for merchandise trade in that it is essentially short run in nature. Thus, the approach involves attempts to precisely identify cyclical and other short-run factors and to separate them from long-run factors. Applications of the model include short-term forecasting (6 to 18 months ahead) and simulation of the effects of variations in growth rates, rates of inflation, and exchange rates.

To our knowledge, no relatively disaggregated world model of flows of invisibles has ever been constructed. Given the important role such flows have come to play in many countries in recent years,1 this gap can be explained only by the numerous statistical and conceptual difficulties involved. While a number of empirical studies on transactions in invisibles have appeared over the past two decades, these studies have generally focused on international flows of aggregated or disaggregated invisibles for a particular country, or flows of a particular group of invisibles (such as travel or transportation) for one country or a group of countries. 2 A complete multicountry analysis of disaggregated groups of services, however, is necessary if the cross-country interdependence of the flows is to be taken into account. Furthermore, a study that applies the same methodology to all countries is also needed if meaningful cross-country comparisons of results are to be made.

The model presented in this paper is currently limited to 14 industrial countries as a consequence of the paucity of the data for the remaining countries. At a later date, the model may be closed by including all countries or country groupings. Two alternative approaches to the specification of export equations for individual invisible items are adopted in the paper. In the first approach, called the unrestricted approach, separate bilateral functions for a country’s exports of invisibles to each importing country are summed over each importing market to obtain an aggregate export equation. This approach is called unrestricted because imports are not predetermined and therefore are not constrained to equal exports. The second approach, called the market shares approach, takes imports of the various countries as predetermined and determines a country’s export share by such factors as relative prices and exchange rates. The market shares approach is used for categories of invisibles for which information on market shares in a base period is available, and for which the market share structure moves slowly over time according to changing factors, such as relative prices. In all other cases, the unrestricted approach is used.

When specifying the model, the serious data constraints that limit any empirical study on invisibles have to be recognized. Thus, invisibles are subdivided into only six broad groups, namely, freight transportation, travel and passenger transportation, other services, investment income, workers’ earnings and remittances, and transfers. These broad groupings of invisibles are thought to be more appropriate than aggregate invisibles for the following reasons: (a) Invisibles can be dichotomized into items for which an income is received (investment income, workers’ earnings and remittances, and transfers) and items that involve an expenditure (freight transportation, travel and passenger transportation, and other services); variables that determine income items will be different from those that determine expenditure items, (b) Within income and expenditure items, imports and exports of the different individual groups considered are likely to be determined by different exogenous variables.

The organization of the paper is as follows. Section I contains the general specification of the model and the individual basic equations for the six groups of invisibles. Section II presents the parameter estimates, together with a discussion of the model results. Section III contains a summary and concluding remarks. The Appendix provides information on data compilation, use of proxy variables, and the arrangement of country data in the standard framework. Complete documentation of the data sources is available upon request from the author, whose address is Research Department, International Monetary Fund, Washington, D.C. 20431.

I. Model Structure

general specification

The invisibles model is a set of equations comprised of behavioral relationships that determine the volume and/or the value for each group of invisibles. The 14 countries and six groups of invisibles are shown in Table 1. The model differentiates between services (including freight transportation, travel and passenger transportation, other services, investment income, workers’ earnings and remittances) and transfers (including private and official transfers). Workers’ remittances 3 are therefore analyzed as a service item, although for balance of payments purposes they are usually defined as a transfer item.

Table 1.Fourteen Industrial Countries: Specification of Invisibles Model
Country DisaggregationService and Transfer GroupingEndogenous VariablesExogenous Variables
AustriaFreight transportationImport value and export valueReal gross national product
BelgiumTravel and passenger transportationImport volume and export volumeImport volume of commodities
CanadaOther services Private GovernmentImport volume and export volume
DenmarkExport volume of commodities
FranceInvestment incomeIncome inflow and income outflow value for Canada, the United Kingdom, and the United StatesConsumer price index
Germany, Fed. Rep.DirectTransatlantic airfare index
FinancialImplicit deflator
ItalyOtherExchange rates
JapanorFreight rates
NetherlandsNet total flowsNet income flow value for remaining 11 countriesUnit labor costs in manufacturing
NorwayInterest rates
SwedenWorkers’ earnings and remittancesIncome inflow and income outflow value for France, the Federal Republic of Germany, and ItalyForeign assets and liabilities
SwitzerlandForeign labor
United KingdomTransfers Private OfficialIncome inflow and income outflow valueNormal output per man-hour in manufacturing
United StatesOutput per man-hour in manufacturing
Rest of the worldTotal services and transfers

Three of the six groups of invisibles (Investment income, Other services, and Transfers) have been divided into subgroups. “Investment income” was subdivided because separate groups of investors are involved whose behavior may be quite different, and because there is evidence of considerable benefit to be gained from disaggregating in terms of increased accuracy of the results. 4 “Other services” and “Transfers” were subdivided to isolate government transactions that are treated as exogenously determined.

Services

The model is based mainly on a static partial equilibrium framework, for which behavioral relationships are hypothesized on the demand side between the level of imports and exports of services and the relevant explanatory variables. Import demand functions are estimated in volume terms for travel and passenger transportation, and other private services, and in value terms for freight transportation, investment income, and workers’ earnings and remittances. 5 Prices are assumed to be exogenous in the period of observation. 6 Where imports of services of a country are a small proportion of world supply, or the price of inputs into the foreign service industry is determined by local demand and supply and the supplier cannot discriminate between domestic and foreign markets, this is a reasonable assumption. To the extent that disequilibrium situations exist because of nonprice factors, such as incomplete adjustment of the optimizing criteria and long lags in buyers’ responses, a disequilibrium approach would be required.

(1) Imports. In our analysis of imports, services are assigned to three main types—general competitive services, specific one-price services, and factor incomes, which are described separately. Specifically, travel and passenger transportation and other private services are assigned to the general competitive service group, freight transportation to the specific one-price service group, and investment income and workers’ earnings and remittances to the factor income group.

For general competitive service imports, product differentiation of the service is assumed, thus requiring separate demand functions for imported and domestically produced services. The volume equations for competitive services are derived on the assumption that the consumer is faced with a “two-stage budgeting decision,” where, at the first stage, expenditure is allocated among broad groups of goods and services with reference to income and prices and, at the second stage, expenditure 7 is allocated between the domestic and the foreign service with reference to domestic and foreign prices of the service. For competitive services, we are concerned with the second stage only. We therefore assume that each imported service group is weakly separable from other goods and services (both domestic and imported). This requires strong assumptions about the way in which imported travel services, for example, are substituted for durables. The approach assumes that marginal rates of substitution between any two services in the relevant service group are independent of the quantity consumed of any commodity or service from any other group.

A general import demand function for each competitive service item for country i will therefore contain the relative price of the imported and domestic service item plus total expenditure on the service item. A trend term is also introduced to represent smoothly changing factors, such as tastes and quality changes. The structural import demand for competitive services of the ith country is written as

In this equation, MVSi is the volume of imports of the service item by country i; DSi is real total expenditure of country i on the service item (i.e., both domestic and foreign service items); and RMPSi is an index of the import price of the service (MPSi) relative to the domestic price of the service (PSi) and is calculated as RMPSi = MPSi/PSi. The import price, MPSi, is a weighted average of the export price indices of each country from which country i imports the service item.

For some service items, only one price exists. The price of such service items is assumed to be determined either by world supply and demand, so that the domestic price does not deviate much from the foreign price, or by monopolistic conference. It is hypothesized here that freight transportation is an example of a one-price service item where prices are determined by world markets. For example, liner prices are determined monopolistically by liner conferences, and tramp prices are determined by a nearly perfect market. Imports of freight transportation are, however, quite likely to be price sensitive, because certain commodities may not be imported if the cost of transporting them becomes too high. The import demand equation for country i for service items with one price may be written

DGSi is real total expenditure of country i on goods and services; WPS is the world price of the service item; and t is introduced to represent smoothly changing factors, such as changes in domestic supply of the service item.

A full analysis of factor incomes would require a consideration of the determinants of the factor level, that is, international investment behavior and migrant labor behavior. For this study, however, the factor level is taken to be exogenously determined, and the analysis will focus on how particular variables affect the demand for services from these factors. Factor incomes are estimated in value rather than in volume terms because of the difficulty involved in finding the appropriate deflators. The general equation for outflows of factor income (imports of the factor service) takes the form

where YOi is the value of the outflow of factor income, RAi is the return to the factor, and FLi is the factor level in country i.

(2) Exports. Two approaches to building export functions are used in this paper—the market shares approach and the unrestricted approach. 8 The market shares approach to specifying export demand functions is an import allocation approach, where changes in market shares over time are explained by relative prices and trends. To specify export demand in this way, use is made of the theoretical formulation derived by Strotz (1957; 1959) and Gorman (1959), which allows for consistency of the two-stage maximization procedures. At the first stage, total import demand for the service of country j (MVSj) is determined, and at the second stage it is allocated among the i exporting countries or markets. Here we assume that the first stage of the allocation has already determined total imports for each country, so that we deal only with the decision as to how these imports should be allocated. We also assume that the utility function is homo-thetic, so that market shares depend only on the relative price of the products in the market, not on the size of the market itself.

With 14 countries included in the model, the number of parameters to be estimated is still quite large; to further reduce the number of responses, we use the theoretical formulation derived by Armington (1969) for products distinguished by place of production. This involves two further assumptions. If the elasticities of substitution between any two competing services in the market are the same as those for any other pair of competing services in that market, then all the price parameters can be expressed in terms of one substitution elasticity, σj and the share of country i in country j’s imports. The general bilateral export function for the market shares approach can therefore be written as

where XVSij is the export volume of the service item from country i to country j; smij0 is the base-year share of the ith country in the jth country’s imports of the service item; PSi is the price of the export service of country i; and MPSj is the average price of the export service to country j, calculated as

and σj is the substitution elasticity in market j. The trend term is added to the equation to account for factors that affect the allocation of imports over time. 9 In this equation, exports of country i to country j are determined by the base-period share of country (market) j’s imports, a change in relative competitors’ prices, and a trend term.

Although equation (4) will not hold exactly in the aggregate unless a first-order linear approximation is obtained, it is assumed here that equation (4) can be used as an adequate approximation of the aggregate relationship, in which XVS represents aggregate exports, and the foreign market variable and foreign prices are appropriately weighted averages. The market-shares estimating equation is

where XVSi is the volume of exports of the service item of country i; FSi is the foreign market variable defined as

and RXPSi is the relative effective price competitiveness variable calculated as

where sxij is the share of the jth country in country i’s service exports. The market shares approach will be used for the two general competitive service items—travel and passenger transportation, and other private services.

In the unrestricted approach to specifying export demand functions, exports are not simply the allocation of predetermined imports. Exports are explained by taking separate unrestricted bilateral import functions 10 for the j countries to which country i exports, and weighting and summing across these j countries. Thus, the export function for service items for which only one price exists is specified as

FDGSi is foreign real total (domestic and foreign) expenditure on goods and services (i.e., weighted aggregate real total expenditure on goods and services of the j countries importing from country i). All other variables are defined in equations (2) and (5). The unrestricted approach is also used to describe inflows of income (YI) for factor services:

RAif is a weighted average of returns to the factor of the j countries importing from country i, and FLFi is the sum of the factor level in the j foreign countries.

Transfers

Private transfers include transfers between individuals, between nonofficial organizations, or between individuals and nonofficial organizations. 11 Included are gifts, dowries, inheritances, alimony, and other support remittances, such as contributions to religious, scientific, cultural, and charitable associations. All private transfers are unrequited. Private transfers are assigned to the factor income group. Equations (3) and (7) are used to estimate, respectively, outflows and inflows of private transfers.

Official transfers comprise military and economic grants, subscriptions to international organizations, technical assistance, and scholarships and pensions. For the purposes of this paper, official transfers are treated as exogenously determined.

Identities

A complete system of world trade in invisibles would require that all flows of the kth invisible item be completely determined for all η countries that trade in the invisible item, so that world imports are identically equal to world exports. To build such a world model for invisibles would be extremely difficult, however, in view of the paucity of the data, so that for the rest of the world invisible items are taken to be exogenously determined. In the estimation of the model, the constraint that world imports equal world exports is not taken into account.

individual group equations

Freight transportation services

The transportation account of the balance of payments contains three separate items: (i) receipts from and payments for passenger transportation; (ii) receipts from and payments for cargo transportation; and (iii) receipts from and payments for port services. 12 In addition, the transportation account is defined to include insurance on merchandise trade. 13

The first item—passenger transportation—has been separated from the other categories of transportation and is included in the equation for international travel. The second item—receipts from and payments for international cargo transportation—covers international receipts from and payments for all vessels and vehicles used for international transportation of goods, excluding port services. The third item of the transportation account—receipts from and payments for port services—includes charges for landing services, maintenance facilities, fuel charges, and expenditure ashore by crews.

Special allowance must be made for payments and receipts for port services. As a country’s expenditure to foreign carriers for the transportation of its imports and passengers increases, the port services of foreign carriers in the country in question must also rise; these port service receipts are effectively an offsetting exported component of imported cargo transportation and passenger transportation services. Similarly, when transportation receipts from freight earnings on a country’s carriers increase, that country’s payments to foreign countries on port services also increase; these port service payments are therefore an offsetting imported component of exported cargo transportation and passenger transportation services. 14

For the individual country, imports of freight transportation services (MVFRi) have two main components: (i) imports of cargo transportation (MVCTi); and (ii) imports of port services (MVPSi). A country’s demand for imports of cargo transportation (MVCTi) is a derived demand, stemming from the demand for imported commodities. The demand for imported cargo transportation is therefore determined by imported commodities (MVGi) and the price of cargo transportation to country i (MPCTi).

Country i’s import demand for port services (MVPSi) is related to country i’s total cargo transportation services abroad and country i’s total passenger transportation services abroad. (Total is defined here to include both exports of these services and services provided for domestic residents by domestic operators.) It is assumed that a country’s total cargo transportation services abroad are proportional to its exports of goods, and that a country’s total passenger transportation services abroad are proportional to its exports of travel services. For the majority of countries in this study, this is a reasonable assumption, since third-country trade (from foreign port to foreign port) in cargo transportation and passenger transportation is negligible. Therefore, country i’s import demand for port services is determined by its exports of goods and travel services (XGTVi) and the price of port services (MPPSi).

Country i’s demand for imported freight transportation services (MVFRi) is the sum of imports of cargo transportation and imports of port services, and may be written as

where

and spmi is the share of country i’s spending on cargo transportation in country i’s spending on cargo transportation and port services.

The trend term is added to the equation to represent factors that change slowly over time. Important factors that may be accounted for by the trend term are technological changes in the shipping industry, commodity and geographic changes in patterns of trade, growth of flag discrimination in some parts of the world, and changing shares in the country’s active world tonnage.

The export demand equation is defined in a way that is similar to the import demand equation. Denoting XVFRi as export volume of freight transportation services of country i, one may write

where

spxi is the share of country i’s receipts from cargo transportation services in country i’s receipts from both cargo transportation and port services; XVGi is the export volume of goods in country i; and MGTVi is the import volume of goods and travel services in country i.

For a few countries—such as Denmark, Japan, Norway, and the United Kingdom—the amount of third-country trade transported by domestic carriers is large, so that exports of cargo transportation are likely to be related to world volume of exported goods as well as to domestic exports, and imports of port services are likely to be related to world volume of exported goods and travel services as well as to domestic exports. 15 For these countries, the trade variables MXVTi and XMVTi are defined as

WXVGTV is the volume of world exports of goods and travel services, WXVG is the volume of world exports of goods, sdmi is the share of country i’s port service payments from own country payments in country i’s payments on port services from own country and third-country payments, and sdxi is the share of country i’s cargo receipts from own country exports in country i’s cargo receipts from own country exports and third-country exports. Import and export prices of freight transportation also include weights for third-country exports.

Equations (8) and (9) are defined in volumes of freight transportation imports and exports. It was decided, however, to multiply both sides of the equation by the price of freight transportation services, so that data on values of freight transportation imports and exports appear on the left-hand side of the equation, as suggested by Hemphill. This is done because XPFRi and MPFRi may be subject to larger measurement errors than are deflators for the other service groups. Equations (8) and (9) can therefore be rewritten for estimation purposes as

Travel and passenger transportation services 16

The international travel item of the balance of payments statistics of a country covers expenditure in the country by nonresidents and payments to foreign countries by residents of the country in question. The passenger transportation item covers passenger fares paid by foreigners to domestic carriers (receipts) and passenger fares paid by domestic passengers to foreign carriers (payments).

It is postulated here that the demand for imports of foreign travel and passenger transportation services for country i (MVTVi) is determined by the price of domestic travel and passenger transportation services relative to the foreign price of these services plus expenditure on domestic and foreign travel and passenger transportation services. It will also depend upon destinational characteristics, such as climate, cultural attributes, and scenery. 17

Total expenditure on domestic and foreign travel and passenger transportation services is not generally available as an aggregate variable for the majority of the countries in this study, so that a proxy variable has to be found. It is asserted here that the expenditure variable is closely associated with the permanent component of income. A permanent income index (YPi) was therefore derived from real gross national product (Yi), measured in constant 1970 U.S. dollars, through a geometrically declining moving average truncated after six periods, namely,

where

and k refers to semiannual frequencies.

The total foreign price variable is a weighted average of the prices of travel and passenger transportation for all j countries that export to country i. Each of these prices should include the price of travel services in country j, PTVj, 18 and the price of transportation from country i to country j, Tij. It is assumed here that the price of transportation services per mile, T, is the same across countries, but that the share of transportation services in each country’s travel and passenger transportation price index may differ from country to country because the distances involved differ. The foreign price of travel services, MPTVi (which is a weighted average of the travel prices of the j countries that export to country i), and the transportation price variable, T, have been separated for estimation purposes because the weights on each price variable in the total foreign price variable are unknown. The travel price variable also enters the equation in lagged form because knowledge of foreign travel price changes may be acquired by travelers from previous trips to the country. Knowledge of transportation prices, on the other hand, is likely to be more readily available to travelers, so that no lag is involved.

The import equation for travel and passenger transportation flows can be written

T is the world index of the average price of passenger transportation; and RMPTVi is the index of import prices of travel services (MPTVi) relative to domestic prices of travel services (PTVi, and is calculated as MPTVi/PTVi.

Once imports of various countries’ foreign travel and passenger transportation services have been determined, the market shares approach can be used to determine the exports of country i; these are determined by base-period shares in country (market) j’s imports, a change in country i’s travel price relative to competitors’ travel prices, the price of transportation,19 and a trend term. The export equation for travel and passenger transportation flows for country i is as follows:

RXPTVi is a double-weighted index reflecting the relative effective price competitiveness and is calculated as

where stvxij is the share of the jth country in country i’s travel and passenger transportation exports, and FTVi is the foreign market share variable for travel services calculated as

where stvmij0 is the share of the ith country in country j’s travel and passenger transportation imports in the base period.

Flows of investment income 20

These flows arise from international transactions and can be broadly classified into three groups: (i) direct investment, (ii) financial investment, and (iii) other investment. A full model of international investment income flows should consist of the foreign demand for assets and the domestic supply of liabilities for each country for each of the three groups. The factors that influence these supply and demand equations are likely to be complex, so that no attempt has been made to explain them here. Two approaches to the specification of investment income flows are adopted—a gross flows approach and a net flows approach.

(1) Gross flows approach. Separate relations are estimated for three types of income flow—direct investment (DI), financial investment (FI), and other investment (OI). Flows of direct investment income arise from the ownership of direct investment capital. Flows of financial investment income are composed of income from private holdings of foreign bonds, foreign equities, long-term loans to foreigners, and short-term claims on foreigners. Flows of other investment income include official investment income, such as income receivable or payable by the country’s government or central bank, by a foreign government or central bank, or by an international organization. Official investment income includes income received by the official sector on reserve holdings.

The general earnings equation for country i for all three groups—DI, FI, and OI—is given in equation (14). The investment income outflow equation for country i (imports of capital services) is then presented for outflows of direct investment income and for outflows of financial and other investment income. The investment income inflow equation (exports of capital services) is then explained by summing across separate bilateral income outflow equations for the j countries in which country i holds investment assets.

Foreign earnings on liabilities in country i (Ekith) during period t on each k-type liability are the product of the average actual rate of return (RAkith) and the average value of the liability holding (LIAkit) in country i: 21

where k refers to the liability type (DI, FI, OI), and h indicates that the earnings and rate of return variables are those appropriate to the home country (i.e., country i). The rates of return, flow of earnings, and income outflows resulting from these earnings are considered separately for direct investment and for financial and other investment.

(a) Outflows of direct investment income. The rate of return on direct investment liabilities in country i (RADIith) is likely to be related to the long-term interest rate (RPDIith) that should reflect the inflation rate in country i, and to cyclical fluctuations in the level of economic activity (CYDIith) that result in quasirents and losses. The rate of return on direct investment liabilities. in country i is

Recorded outflows of direct investment income differ from earnings because the proportion of earnings that is typically retained and reinvested in the host country is not recorded together with income outflows. To determine outflows of direct investment income, it is necessary to determine the factors that affect the proportion of earnings that is reinvested. Important factors that affect the proportion of earnings repatriated include tax policies or restrictions on profit remittances of host countries, the expected exchange rate at which domestic currency earnings will be converted at the time of repatriation, and the relative profitability of domestic investment compared with foreign investment. The first two factors are difficult to model, so that it is assumed here that the proportion of earnings repatriated (resulting in income outflows) will depend upon the relative profitability of domestic and foreign investment (PRi). Where tax policies, restrictions, and “leads and lags” with respect to exchange rates exist, and are well documented, dummy variables are included in the equation.

In addition to these effects, firms or individuals often adjust their income outflows (YODIit) slowly to higher or lower levels of earnings. When the change in earnings is accepted as permanent, rather than temporary, adjustment will be complete. Outflows of permanent direct investment income (YOPDIit) are given by

and actual income outflows adjust gradually to permanent income outflows over a period of time. The adjustment process can be explained by a stock adjustment model of the form

where η is the coefficient of adjustment. Substituting equation (15) into (14), (14) into (16), and (16) into (17) gives

The coefficients of equation (18) are related to the underlying parameters in the following manner:

(b) Outflows of financial and other investment income. Financial and other investment liabilities are a mixture of longterm and short-term investments, and for these two categories, earnings are

γki is country i’s share of long-run liabilities of type k (i.e., financial liabilities and other liabilities) in its holdings of both long-run and short-run liabilities of type k; and RPkih is the long-run interest rate, RTkih the short-run interest rate, on liability of type k in country i. 22

Outflows of financial and other investment income are assumed to be a constant fraction of earnings, so that

Many long-run financial and other investment liabilities have fixed rates of return, so that past interest rates have an effect on current income earned. A lagged dependent variable is therefore added to the equation to reflect the effect of past interest rates on outflows of current income. Substituting equation (19) into (20) and adding a lagged dependent variable gives

Income inflow (YI) equations can be explained in a way that is similar to income outflow equations except that the exogenous variables are weighted and summed across the j foreign countries in which country i holds assets. The income inflow equation for direct investment can be written as

and for financial and other investment

αki is country i’s share of long-run assets of type k in its holdings of both long-run and short-run assets of type k;RPkif is a weighted average of foreign long-run interest rates, and RTkif a weighted average of foreign short-run interest rates, on asset of type k, where the weights are the share of country j in country i’s total direct investment assets; and ASSki is the average value of country i’s total holding of assets abroad over the period.

(2) Net flows approach.23 In this approach, investment income is defined in net rather than in gross terms. The advantages to be gained from using this method to estimate investment income flows is that it eliminates the need for data on stocks of international assets and liabilities. The assumptions on which the approach is based are rather more restrictive than the assumptions in the gross flows approach: (i) For each country, the rate of return on all assets at time t is assumed to be equal to the rate of return on all liabilities at time t. (ii) The parameters for income inflow equations are assumed to be identical to those for income outflow equations, (iii) The current account is assumed to be equal to net capital flows (including reserve changes), (iv) Income inflows and outflows are assumed to be the same as earnings.

For countries with large errors and omissions items representing an error in an item on the current account, the net change in the investment position will be misstated. It will also be misstated where balancing items are funded partly (or totally) by gold, which does not earn a rate of return. Furthermore, for those countries that report income flows rather than earnings, there is no information on what proportion of earnings is actually repatriated.

Using these assumptions, the equation for net flows of investment income (YN) for all categories of investment for country i is

Assuming that RAitf = RAith = RAit, equation (25) can be rewritten as

where NASSit = ASSitLIAit. NASSit is the net average value of the foreign asset position of country i at time t, and RAit is the average rate of return on these net foreign assets. At present, equation (26) is merely a definition. It becomes an estimating equation when we find proxy variables for RAit and NASSit. The estimating equation for net flows of investment income can be expressed as

where RA^it and NASS^it are proxy variables for RAit and NASSit. Clearly, the closer the proxy variables are to the real variables, the closer c0 will be to zero and c1 to one.

Other services

This category consists of two subgroups: (i) an account that covers transactions in services between private residents of a country and overseas residents that cannot be included in other categories; and (ii) miscellaneous government transactions. The latter subgroup is treated as exogenously determined in this study.

Because it includes items that cannot be identified elsewhere, the group “Other private services” is extremely heterogeneous. For countries for which workers’ earnings are important and can be distinguished separately, this item has been treated separately. The remainder of the other service group is made up of commissions and agency fees connected with selling of imports and exports; professional and technical services; earnings and payments for films and television programs; communications services, including telecommunications and postal and telephone services; financial and allied services, such as nonmerchandise insurance; royalties and management fees; and other property income not included elsewhere. Where the data permit, royalties and management fees are estimated together with direct investment income flows.

Import demand for other private services for country i is postulated to be determined by economic activity of that country as measured by gross national product (GNP), 24 and by the price of domestic other private services relative to the foreign price of these services. Export demand for other private services is explained by the market shares approach. Other private services import and export equations for the ith country are as follows:

MVOPi and XVOPi are import and export volumes of other private services for country i, GNPi is real gross national product in country i in constant 1970 U. S. dollars, RMPOPi is the weighted index of import prices of other private services (MPOPi) relative to the domestic price of other private services (POPi) for country i; RXPOPi is a double-weighted index reflecting the relative effective price competitiveness, and FOPi is the foreign market share variable for other private services.

Workers’ earnings and remittances

Workers’ earnings are defined as repatriated labor income of seasonal, border, or temporary workers who have been non-residents of the country for less than one year. Labor income includes wages, salaries, and other compensation to these workers on a gross basis without deduction for taxes, pensions, or other contributions that are included as offsets to labor income in unrequited transfers. Worker’s remittances are included in the private transfers group and are defined as unrequited transfers of migrant workers. These migrant workers are persons who stay, or are expected to stay, in the host country for more than one year. In practice, it is often difficult to draw the distinction between workers’ earnings and workers’ remittances.

A full model of workers’ earnings and remittances should consist of the foreign demand for migrant labor and the domestic supply of migrant labor for each country. The factors that influence these supply and demand equations are likely to be complex, however, so that no attempt has been made to explain them here; instead, the level of migrant labor is taken to be exogenously determined.

Earnings of foreign workers in the host country i (EWRit) during period t are the product of the average wage paid to foreign workers in country i over the period t (WFit) and the number of foreign workers in country i over period t (FLit):

The average wage paid to foreign workers in country i (WFi) is the product of the average hourly wage rate in country i (AHWi) and the number of hours worked in country i (HHWi). The number of hours worked by foreign workers is likely to be related to the cyclical fluctuations in country i; foreign workers are rarely employed in jobs, such as government employment, that are protected from cyclical movements, and the number of hours worked are quite likely to vary with economic activity in the host country.

Only a proportion of earnings of foreign workers in country i will be repatriated. To determine the proportion of earnings repatriated, it is necessary to consider changes in the exchange rate, changes in taxation in the host country, and many other factors. Because of numerous difficulties involved in deriving a model for speculative activity, these activities are included as dummy variables where they have been well documented in other sources. Therefore, YOWRi = EWRiθ where YOWRi is income outflows, and the estimating equation may be written as

A bilateral outflow of workers’ earnings and remittances from country i to country j represents an inflow of workers’ earnings and remittances for country j, so that the preceding analysis explains both flows. Hence, inflows of workers’ earnings and remittances (YIWRi) for a particular country is merely the sum of outflows of foreign countries to which the domestic country sends workers, so that the estimating equation for inflows will be

FFLi is the number of country i’s workers in the labor force of foreign countries j, calculated as

FHWi is the number of hours worked by country i’s workers in the labor force of foreign countries j, calculated as

sweij is the share of country j in country i’s total workers abroad, namely, sweij = FLij/FFLi; and AFWi is the average hourly wage rate paid to workers from country i in the labor force of foreign countries j and is calculated as

Private transfers

Outflows of private transfers of country i (YOTPi) are generally assumed to be related to the level of GNP in country i, plus a trend factor, and the equation takes the form

Inflows of private transfers of country i(YITPi) are related to foreign GNP (FGNPi). This variable is a weighted average of the GNP of the j countries that send private transfers to country i, where the weights are the share of country j in country i’s private transfer inflows. The inflow equation for private transfers is

II. Parameter Estimates

Information on the data compilation, the use of proxy variables, and the arrangement of individual country data in the standard framework is contained in the Appendix. All the variables in equations (8) to (34) are expressed in billions of U. S. dollars. Balance of payments data in nominal (current) dollars are deflated prior to estimation for travel and passenger transportation and for other private services by use of deflators based at 1970 = 1.0. For freight transportation, the dependent variable is in nominal dollars and the deflator is included as an extra variable on the right-hand side of the equation, because doubt exists as to the appropriateness of the deflator. Factor services and private transfers are estimated in current dollars because no appropriate deflator could be found. Import and export volumes of goods are on an f.o.b. basis. All price indices are in U. S. dollars based at 1970 = 1.0.

Estimates of the parameters for the six groups of invisibles were obtained from semiannual data for the 14 industrial countries by means of ordinary least squares. The estimates are shown, together with their standard errors, the R2, the Durbin-Watson statistic, and the standard error of the estimate in Tables 2 to 16. The estimation period, determined by the availability of data for each country and category, is shown in column 2 of each table. A seasonal dummy taking the value of one for the first half year and zero for the second was added to the equation where there was an indication of significant seasonal variations. Other dummy variables were included to represent the effects of special factors or events and are documented in the footnotes to the tables.

The model was estimated using the equations as specified in Section I in log-linear form. To check whether the equation should be linear or log-linear, the equations in Section I were also specified in linear form. To choose between the two specifications, the variance of the linear specification (ût) was compared with the antilog of the variance of the log-linear specification (ût). The ratio of these two variances (ûtt’) was greater than one for most invisibles, so that the log-linear specification was preferred, since it predicts more efficiently over the estimation period. Thus, we assume that the elasticities remain constant over the estimation period. 25 The linear specification was used for those invisibles that are estimated as net flows.

It was felt that the neglect of long-run influences on general competitive and specific one-price invisibles would have two effects, (i) A misleading interpretation of the demand parameters—in particular, income and trade elasticities—that would pick up both cyclical and long-run effects. For example, income elasticities for travel and passenger transportation would include the effects of changes in tastes as well as income effects, (ii) The misspecification of the equation because of omitted variables. 26 The addition of a trend term to represent these omitted long-run factors has, however, led to estimation problems; sometimes the short-run variations in the variables are not large enough, and strong interrelationships among the independent variables, particularly the demand and trend variables, make it difficult to disentangle their separate effects on the dependent variable. In several cases, these interrelations have caused severe multicol-linearity problems, and coefficients with “wrong signs” or with magnitudes that are meaningless have resulted. In these cases, two possible solutions can be tried. First, we can look at the linear combination of the coefficients together; unfortunately, this solution does not enable us to separate the long-run and short-run effects. Second, we can use extraneous estimates of one coefficient to correct for variation in one of the variables, and then estimate the parameters on the other variables. That is the approach followed in this study; in the event of severe multicollinearity, the parameter on the trend term in the equation is constrained to equal the average value of the estimated coefficients of the trend term for the 14 industrial countries. The underlying assumption is that long-run factors that influence the dependent variable are the same for all countries but that it is difficult to measure their impact for individual countries with any degree of precision in the presence of multicollinearity. In the tables, the parameters whose values were restricted a priori to assumed values are not given a standard error.

For some countries and categories of invisibles, the number of observations was quite small. For example, there were only 14 observations for several of the industrial countries for freight transportation and other private service flows. Obviously, the point estimates of the coefficients are less likely to be reliable with a small number of observations than if the number were large, and in these cases some caution should be used in attributing accuracy to findings.

A detailed discussion of the empirical results for each of the six groups of invisibles follows.

services

Freight transportation services

The estimates of the parameters for equations (10) and (11) for each of the 14 countries are presented in Tables 2 to 4.

Table 2.Fourteen Industrial Countries: Regression Coefficients for Imports of Freight Transportation Services1
CountryEstimation

Period2
ConstantMXVTiMPFRiTrend
Dummies
R2D-WSEE
Austria1970–76–3.94*0.141.04*0.072*0.9881.580.061
(0.33)(0.21)(0.26)(0.013)
Belgium1965–76–2.07*0.53*1.54*0.0150.25*30.9911.470.064
(0.25)(0.21)(0.21)(0.013)(0.06)
Canada1964–76–1.64*0.61*1.26*0.0050.9871.910.050
(0.35)(0.19)(0.18)(0.009)
Denmark1970–76–1.56*0.321.47*0.0170.9791.680.062
(0.48)(0.45)(0.21)(0.012)
France1968–76–1.250.430.44*0.056*–0.31*40.9851.490.094
(0.98)(0.36)(0.12)(0.020)(0.10)
Germany, Fed. Rep.1961–76–1.98*0.82*0.62*0.0030.17*50.17*60.9941.520.051
(0.46)(0.20)(0.10)(0.010)(0.04)(0.06)
Italy1968–760.32–0.40*0.60*0.054*0.9772.660.054
(0.35)(0.18)(0.09)(0.007)
Japan1964–76–0.15–0.48*0.58*0.103*0.16*70.9981.370.030
(0.11)(0.07)(0.05)(0.005)(0.03)
Netherlands1970–76–1.170.181.06*0.030*0.9661.360.073
(1.12)(0.48)(0.25)(0.012)
Norway1970–76–0.16–1.64*1.25*0.074*–0.17*80.9892.190.044
(0.33)(0.36)(0.20)(0.010)(0.05)
Sweden1970–76–2.38*0.730.490.043*0.9572.190.083
(1.00)(0.60)(0.28)(0.013)
Switzerland1970–76–3.49*0.57*1.00*0.0190.9721.810.054
(0.51)(0.19)(0.26)(0.012)
United Kingdom1961–76–0.940.68*0.60*0.0010.36*90.16*100.9931.390.041
(0.59)(0.24)(0.07)(0.005)(0.03)(0.05)
United States111961–76–1.98*0.60*0.27*0.037*0.12*12–0.07*130.8272.050.036
(0.63)(0.17)(0.09)(0.010)(0.03)(0.03)

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the first half of 1976.

Increase in payments on bunkering in Belgium, first half and second half 1974.

Shift dummy, first half 1968 to second half 1970—these observations are not reconcilable with series after first half 1971.

Shift dummy, second half 1971 to first half 1976.

Temporary increase in imports of raw materials, first half 1975.

Oil crisis, first half and second half 1974.

End of period of long-term contracts, second half 1975.

Shift dummy, second half 1970 to first half 1976.

Middle East crisis, second half 1967.

Corrected for first-order autocorrelation. The first-order autoregressive coefficient was 0.9028 (0.0379).

Oil crisis, second half 1974.

Dock strike in the United States, first half 1969.

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the first half of 1976.

Increase in payments on bunkering in Belgium, first half and second half 1974.

Shift dummy, first half 1968 to second half 1970—these observations are not reconcilable with series after first half 1971.

Shift dummy, second half 1971 to first half 1976.

Temporary increase in imports of raw materials, first half 1975.

Oil crisis, first half and second half 1974.

End of period of long-term contracts, second half 1975.

Shift dummy, second half 1970 to first half 1976.

Middle East crisis, second half 1967.

Corrected for first-order autocorrelation. The first-order autoregressive coefficient was 0.9028 (0.0379).

Oil crisis, second half 1974.

Dock strike in the United States, first half 1969.

(1) Imports. Table 2 shows that 13 of the 14 countries have more than 95 per cent of the variance of their imports explained as measured by the R2 statistic. The Durbin-Watson test rejected the hypothesis of serial independence in the residuals for the U. S. equation, and this equation was re-estimated using the Cochrane-Orcutt correction method.

The results for this category of invisibles are rather disappointing for some countries because it was not possible to identify both short-term and long-term influences with as much precision as one would like. As a result of interrelationships between the trade variable (MXVT) and the time trend, the estimates of the coefficient for the trade variable are poorly determined and are significant at the 95 per cent level for only 9 of the 14 countries; of these 9 countries, only 6 (Belgium, Canada, the Federal Republic of Germany, Switzerland, the United Kingdom, and the United States) have the predicted positive sign. All the coefficients are less than one; use of fleets at less than capacity during a depression and economies of scale during a boom could account for these coefficients being less than one. The values and signs of the coefficients on the trade variable for Austria, France, Italy, Japan, the Netherlands, and Norway, and the fact that a low or negative estimated coefficient for the trade variable is accompanied by a high coefficient on the trend term, lead one to conclude that the multicollinearity between the trade variable and the trend term may be the reason for the poor results for these countries. 27 The cycle in the trade variable and the dependent variable is not measured with sufficient accuracy by the data, and these variables are strongly trended. The large positive coefficient on the trend term for the United States could be due to the effects on freight imports of a long-run decline in the size of the shipping fleet.

The estimates of the effects of changes in tramp, liner, tanker, and surface rates (MPFR) are rather more satisfactory, as they are of the expected sign, and are significant at the 95 per cent level for 13 of the 14 countries. The estimated price coefficients are close to unity only for Austria, the Netherlands, and Switzerland; most of the other countries have coefficients that are significantly less than one. This would indicate that imports of freight transportation services are price sensitive. 28 The results may be biased, however, because of the likely existence of large errors of measurement in the price series used, and because of the effects of nonprice factors, such as maintenance time, loading and unloading time, and safety factors, which are omitted from the equation. The low price coefficient for the United States could be the result of measurement errors in the form of omitted labor costs. The coefficients on the price term for Belgium and Denmark appear larger than one; the size of these coefficients may also reflect errors in data measurement.

To increase the precision of the trade variable for six countries for which multicollinearity was a problem (Austria, France, Italy, Japan, the Netherlands, and Norway), the coefficient on the trend term was constrained to be 0.013. This value is the average value of the estimated coefficients for six countries (Belgium, Canada, the Federal Republic of Germany, Switzerland, the United Kingdom, and the United States) for which multicollinearity problems did not seem to be too severe. 29 The new results are presented in Table 3. The estimates of the trade coefficients have gained in precision and are now significant at the 95 per cent level for Austria, France, Italy, and Japan; the imposed restriction has led to little difference in the R2 statistic except for Italy, where it has fallen from 0.977 to 0.880. The freight price coefficient for Norway, however, remained much larger than one, while the coefficient on the trade variable remained negative. The results were more reasonable when the freight price coefficient was constrained to equal 1.0. The equation with the added constraint gave a coefficient on the trade variable that is reasonable in size and sign; unfortunately, this was at the cost of a drastically reduced R2 statistic.

Dummy variables were included in 7 of the 14 country equations and are described in the footnotes to Table 3. A shift dummy variable was included in the equations for France, the Federal Republic of Germany, and the United Kingdom because the data definition changed during the period of estimation. The dock strike in the United States in the first half of 1969 was captured with a dummy variable; the full impact of the strike was not reflected in imports and exports because the strike led to substantial rerouting of goods through Canada. The Middle East crisis and the closing of the Suez Canal in 1967 had an unfavorable influence on imports by the United Kingdom. The equations for several countries included a dummy variable to account for the impact of the oil crisis on freight-transportation imports. The oil embargo affected countries’ imports at different periods, depending on the length of the lag involved.

Table 3.Fourteen Industrial Countries: Regression Coefficients for Imports of Freight Transportation with Trend Coefficient Constrained to 0.013 for Six Countries1
CountryEstimation

Period2
ConstantMXVTiMPFRiTrend
Dummies
R2D-WSEE
Austria1970–76–3.08*0.53*1.96*0.0130.9550.940.100
(0.42)(0.13)(0.24)
Belgium1965–76–2.07*0.53*1.54*0.0150.25*30.9911.470.064
(0.25)(0.21)(0.21)(0.013)(0.06)
Canada1964–76–1.64*0.61*1.26*0.0050.9871.910.050
(0.35)(0.19)(0.18)(0.009)
Denmark1970–76–1.56*0.321.47*0.0170.9791.680.063
(0.48)(0.45)(0.21)(0.012)
France1968–76–1.430.95*0.59*0.013–0.40*40.9751.540.107
(1.11)(0.32)(0.11)(0.10)
Germany, Fed. Rep.1961–76–1.98*0.82*0.62*0.0030.17*50.17*60.9941.520.051
(0.46)(0.20)(0.10)(0.010)(0.04)(0.06)
Italy1968–76–0.890.44*0.73*0.0130.8801.310.096
(0.50)(0.19)(0.17)
Japan1964–76–1.66*0.89*1.03*0.013–0.0170.9570.680.132
(0.29)(0.10)(0.22)(0.12)
Netherlands1970–76–1.620.521.19*0.0130.9471.370.076
(1.12)(0.44)(0.23)
Norway1970–76–0.84*0.421.000.0130.1480.4111.490.094
(0.26)(0.23)(0.10)
Sweden1970–76–2.38*0.730.490.043*0.9572.190.083
(1.00)(0.60)(0.28)(0.013)
Switzerland1970–76–3.49*0.57*1.00*0.0190.9721.810.054
(0.51)(0.19)(0.26)(0.012)
United Kingdom1961–76–0.940.68*0.60*0.0010.36*90.16*100.9931.390.041
(0.59)(0.24)(0.07)(0.005)(0.03)(0.05)
United States111961–76–1.98*0.60*0.27*0.037*0.12*12–0.07*130.8272.050.036
(0.63)(0.17)(0.09)(0.010)(0.03)(0.03)

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the first half of 1976.

Increase in payments on bunkering in Belgium, first half and second half 1974.

Shift dummy, first half 1968 to second half 1970—these observations are not reconcilable with series after first half 1971.

Shift dummy, second half 1971 to first half 1976.

Temporary increase in imports of raw materials, first half 1975.

Oil crisis, first half and second half 1974.

End of period of long-term contracts, second half 1975.

Shift dummy, second half 1970 to first half 1976.

Middle East crisis, second half 1967.

Corrected for first-order autocorrelation. The first-order autoregressive coefficient was 0.983 (0.008).

Oil crisis, second half 1974.

Dock strike in the United States, first half 1969.

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the first half of 1976.

Increase in payments on bunkering in Belgium, first half and second half 1974.

Shift dummy, first half 1968 to second half 1970—these observations are not reconcilable with series after first half 1971.

Shift dummy, second half 1971 to first half 1976.

Temporary increase in imports of raw materials, first half 1975.

Oil crisis, first half and second half 1974.

End of period of long-term contracts, second half 1975.

Shift dummy, second half 1970 to first half 1976.

Middle East crisis, second half 1967.

Corrected for first-order autocorrelation. The first-order autoregressive coefficient was 0.983 (0.008).

Oil crisis, second half 1974.

Dock strike in the United States, first half 1969.

(2) Exports. The parameter estimates for exports of freight transportation derived from equation (11) are presented in Table 4, and these results in general support the hypothesized relationships that freight transportation exports depend on exports of goods and port disbursements and on freight rates. The coefficients on the long-run trend term are also significant in 7 of the 14 equations; for the United Kingdom, the coefficient is negative and could represent supply factors that have a depressing influence on exports, such as the growth in the number of ships operating under flags of convenience to take advantage of lower operating costs. The large positive coefficient on the trend term for Japan can be explained by the considerable expansion of the Japanese fleet, particularly in the early 1970s. The trend term for Japan is also thought to reflect the increase in receipts from leasing domestically owned ships to foreign operators. The increase in transshipment from Rotterdam to the Rhine and other river ports and increased road haulage activity may explain part of the large positive coefficient on the time trend for the Netherlands equation.

Table 4.Fourteen Industrial Countries: Regression Coefficients for Exports of Freight Transportation Services1
CountryEstimation

Period2
ConstantXMVTiXPFRiTrendDummiesR2D-WSEE
Seasonal
Other
Austria1970–76–3.50*0.681.43*0.019–0.100.9431.290.103
(0.61)(0.54)(0.37)(0.07)
Belgium1965–76–2.01*0.60*1.66*0.0050.21*30.9951.890.045
(0.18)(0.17)(0.18)(0.012)(0.04)
Canada1964–76–1.67*0.45*0.43*0.026*–0.09*0.9931.720.035
(0.25)(0.12)(0.07)(0.004)(0.02)
Denmark1970–76–3.310.441.20*0.027–0.06*0.9851.470.060
(2.74)(0.60)(0.24)(0.162)(0.05)
France1968–76–1.38*0.49*1.16*0.019–0.36*40.9892.150.061
(0.60)(0.20)(0.16)(0.06)
Germany,

Fed. Rep.
1961–76–1.43*0.65*1.04*–0.0040.21*50.06*60.9941.880.042
(0.40)(0.17)(0.11)(0.009)(0.05)(0.03)
Italy1968–76–1.06*0.340.98*0.012–0.16*70.9882.230.042
(0.36)(0.22)(0.14)(0.011)(0.05)
Japan1964–76–3.84*0.84*1.52*0.0290.9952.140.059
(1.09)(0.33)(0.17)(0.017)
Netherlands1970–76–3.68*0.73 [-1]0.46*0.053*0.0970.9901.520.049
(1.80)(0.51)(0.19)(0.013)(0.06)
Norway1970–76–5.151.220.48*–0.0030.9281.410.084
(2.88)(0.66)(0.22)(0.019)
Sweden1970–76–1.56*0.660.54*0.015–0.28*0.9681.800.063
(0.67)(0.35)(0.18)(0.012)(0.03)
Switzerland1970–76–2.86*0.58*1.21*–0.0170.9841.600.035
(0.33)(0.22)(0.19)(0.011)
United Kingdom1964–76–3.74*1.14*0.85*–0.029*0.19*80.25*90.9901.580.047
(1.10)(0.29)(0.12)(0.013)(0.05)(0.04)
United States1961–76–0.130.29*1.08*0.010*–0.11*100.9961.480.028
(0.36)(0.12)(0.07)(0.004)(0.03)

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the first half of 1976.

Increase in receipts on bunkering in Belgium, 1974.

Shift dummy, first half 1968 to second half 1970—these observations are not reconcilable with series after first half 1971.

Increased exports of investment goods to members of the Organization of Petroleum Exporting Countries, first half 1975.

Shift dummy, second half 1971 to first half 1976.

Oil crisis, second half 1974.

Devaluation of sterling, second half 1967.

Shift dummy, second half 1970 to first half 1976.

Dock strike in the United States, first half 1969.

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the first half of 1976.

Increase in receipts on bunkering in Belgium, 1974.

Shift dummy, first half 1968 to second half 1970—these observations are not reconcilable with series after first half 1971.

Increased exports of investment goods to members of the Organization of Petroleum Exporting Countries, first half 1975.

Shift dummy, second half 1971 to first half 1976.

Oil crisis, second half 1974.

Devaluation of sterling, second half 1967.

Shift dummy, second half 1970 to first half 1976.

Dock strike in the United States, first half 1969.

For some countries, the trade variable did not fluctuate widely enough over the cycle, with the result that the trade variable and the trend term are highly collinear. This lowered the precision of the estimates, particularly for Austria and France, and for these 2 countries the trend term was constrained to be equal to the average value of the coefficients of the trend term (0.019) for 12 of the 14 countries; Japan and the United Kingdom were excluded because of their large, changing shares in active world tonnage. The unconstrained results for Austria and France are recorded in footnote 30; 30 these can be compared with the more precise constrained results in Table 4. Of the 14 countries, 8 show estimated coefficients for the trade variable (XMVT) that are significant at the 95 per cent level; these coefficients range in size between 0.29 and 1.14. In general, one would expect the elasticity of freight transportation exports with respect to the volume of trade to be unity; however, the average elasticity for all 14 countries here is 0.65. More efficient use of the fleet during a boom and inefficient use during a depression could account for this average elasticity measurement.

The results show that for all countries freight rates are significant at the 95 per cent level and that for three countries (the Federal Republic of Germany, Italy, and the United States) the coefficients on the price term are close to unity. Results for the other countries are widely distributed on both sides of one, so that no clear conclusion can be reached as to the price sensitivity of freight transportation exports with respect to freight rates.

The seasonal dummy variables indicate a strong negative seasonal pattern for the first half year for Austria, Canada, and Sweden. Dummies for other disturbances had an important impact for Belgium, the Federal Republic of Germany, Italy, the Netherlands, the United Kingdom, and the United States. The sign of the coefficient on the dummy for the oil embargo is not uniform; the coefficient is positive for the Netherlands and negative for Italy. The difference in signs is thought to be due to the different impact that oil prices had on different sectors of the shipping market; transportation credits rose as a result of the impact of the oil crisis on the tramp and liner market, but fell as a result of the impact on the tanker market. The positive coefficient on the oil crisis dummy for the Netherlands, for example, is due largely to higher values of bunker deliveries. The dummy for the devaluation of sterling was included in the equation for the United Kingdom, because freight rates on third-country trade were set in terms of the sterling value of foreign goods.

Travel and passenger transportation services

Equations (12) and (13) were fitted to data on travel and passenger transportation flows for 14 industrial countries. Direct estimation of the distributed lags on the relative price variable is not possible because of the collinearity of the lagged values, and an indirect method using a polynomial distributed lag was used. A polynomial of degree two gave the best results in terms of minimum standard error criteria. The length of the lag was also determined by minimum standard error criteria, and in some instances a zero constraint was imposed on the value of the end-period coefficient. 31

(1) Imports. The estimates of the parameters for equation (12) are presented in Table 5. For all of the countries, more than 96 per cent of their variance is explained by the chosen variables as can be seen from the R2 statistic, and there was no evidence of serial autocorrelation in the residuals.

Table 5.Fourteen Industrial Countries: Regression Coefficients for Imports of Travel and Passenger Transportation Services1
CountryEstimation

Period2
ConstantYPiRMPTViTTrendDummiesR2D-WSEE
Long-run elasticity3Mean lag4Seasonal
Other
S1S2
Austria51967–763.24*

(1.17)
–2.56*

(0.59)
–0.83 [4]

(0.97)
0.70

(3.18)
0.13

(0.18)
0.112*

(0.017)
–0.33*

(0.07)
0.9882.510.035
Belgium1964–76–6.69*

(1.60)
2.06*

(0.63)
–4.85 [4]

(2.75)
2.19*

(1.06)
–0.12*

(0.06)
–0.020

(0.019)
–0.22*

(0.01)
0.14*6

(0.04)
0.10*7

(0.04)
0.9942.210.034
Canada1960–76–13.35*

(4.92)
3.50*

(1.31)
–1.35* [4]

(0.41)
2.08

(1.09)
–0.51*

(0.14)
–0.077

(0.037)
–0.05*

(0.02)
0.10*8

(0.04)
0.35*9

(0.09)
–0.13*10

(0.05)
0.9731.440.061
Denmark1961–76–5.65*

(1.62)
1.79*

(0.69)
–2.76* [6]

(0.41)
2.10*

(0.73)
–0.03

(0.14)
–0.023

(0.015)
–0.28*

(0.02)
0.9871.420.046
France1960–76–14.17*

(4.04)
3.13*

(0.92)
–2.93* [4]

(0.52)
1.01*

(0.37)
–0.14

(0.22)
–0.017

(0.025)
–0.37*

(0.02)
–0.29*8

(0.04)
–0.71*11

(0.08)
0.33*12

(0.06)
0.19*13

(0.06)
0.9891.480.063
Germany, Fed. Rep.1960–76–4.34*

(1.80)
1.04*

(0.38)
–2.78* [4]

(0.42)
0.25

(0.47)
–0.22

(0.13)
0.013

(0.009)
–0.33

(0.18)
0.13*14

(0.05)
0.9931.460.052
Italy1961–76–8.09*

(3.06)
1.85*

(0.76)
0.71 [4]

(0.47)
2.63

(1.33)
–1.96*

(0.23)
–0.014

(0.016)
–0.28*

(0.02)
0.16*15

(0.06)
–0.14*16

(0.07)
–0.21*17

(0.06)
0.9931.500.058
Japan1961–76–14.04*

(2.68)
2.67*

(0.61)
–0.71 [4]

(0.78)
3.81

(3.73)
0.17

(0.27)
–0.016

(0.026)
–0.09*

(0.03)
–0.19*18

(0.04)
0.9961.320.071
Netherlands1960–76–4.25

(2.40)
0.65

(0.85)
4.47* [4]

(1.30)
1.47*

(0.59)
0.27

(0.21)
0.078*

(0.027)
–0.22*

(0.02)
–0.16*19

(0.06)
0.9921.580.054
Norway1964–76–7.08

(6.22)
2.78

(3.44)
–0.35 [4]

(1.84)
0.54

(2.70)
0.40*

(0.10)
–0.039

(0.089)
–0.21*

(0.02)
0.9691.830.052
Sweden1961–76–5.34*

(1.73)
1.25*

(0.57)
–1.39 [4]

(0.87)
0.34

(1.54)
–0.43*

(0.17)
0.018

(0.011)
–0.16*

(0.01)
0.11*20

(0.06)
0.9952.110.040
Switzerland1960–76–2.56*

(0.57)
0.55*

(0.25)
–1.88* [6]

(0.42)
1.86*

(0.86)
–0.63*

(0.13)
0.007

(0.009)
–0.26*

(0.08)
0.09*21

(0.02)
0.9971.660.023
United Kingdom1961–76–4.50

(3.60)
1.03

(0.80)
–0.66* [4]

(0.26)
0.51

(1.31)
–0.34*

(0.07)
0.012

(0.012)
–0.43*

(0.13)
–0.28*8

(0.02)
–0.12*22

(0.02)
–0.10*23

(0.03)
0.9942.320.031
United States1960–76–3.74

(2.24)
0.72*

(0.34)
–1.26* [3]

(0.26)
2.20*

(0.60)
00.021*

(0.008)
–0.17*

(0.02)
0.07*10

(0.03)
0.9782.550.049

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the second half of 1976.

The length of the lag is indicated in square brackets.

The mean lag is in half-year units.

Corrected for first-order autocorrelation. The first-order autoregressive coefficient was -0.768 (0.166).

Speculative flows in Belgium, first half and second half 1973.

Speculative flows in Belgium, first half and second half 1969.

Shift in seasonal pattern as follows—Canada: S1, first half 1960 to second half 1973; S2, first half 1974 to second half 1976; France and the United Kingdom: S1, first half 1960 to second half 1971; S2, first half 1972 to second half 1976.

Bicentennial year in the United States, first half and second half 1976.

Expo in Canada, first half and second half 1967.

Shift dummy, first half 1960 to second half 1967.

Political disturbances and strikes in France, first half and second half 1968.

Foreign exchange restrictions in France, first half 1968 to second half 1972.

Austrian Olympics and backlog of delayed travel plans, second half 1975 to first half 1976.

Speculative flows in Italy, first half and second half 1971.

Speculative flows in Italy, first half and second half 1972.

Speculative flows in Italy, first half and second half 1974.

Foreign exchange restrictions in Japan, first half 1967 to second half 1970.

Temporary increase in direct and indirect taxation in the Netherlands, first half and second half 1971.

Rapid expansion of organized flights abroad from Sweden, first half and second half 1970.

Shift dummy, first half 1960 to second half 1966.

Foreign exchange restrictions in the United Kingdom, first half 1967 to second half 1969.

Political disturbances in Europe, first half and second half 1968.

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the second half of 1976.

The length of the lag is indicated in square brackets.

The mean lag is in half-year units.

Corrected for first-order autocorrelation. The first-order autoregressive coefficient was -0.768 (0.166).

Speculative flows in Belgium, first half and second half 1973.

Speculative flows in Belgium, first half and second half 1969.

Shift in seasonal pattern as follows—Canada: S1, first half 1960 to second half 1973; S2, first half 1974 to second half 1976; France and the United Kingdom: S1, first half 1960 to second half 1971; S2, first half 1972 to second half 1976.

Bicentennial year in the United States, first half and second half 1976.

Expo in Canada, first half and second half 1967.

Shift dummy, first half 1960 to second half 1967.

Political disturbances and strikes in France, first half and second half 1968.

Foreign exchange restrictions in France, first half 1968 to second half 1972.

Austrian Olympics and backlog of delayed travel plans, second half 1975 to first half 1976.

Speculative flows in Italy, first half and second half 1971.

Speculative flows in Italy, first half and second half 1972.

Speculative flows in Italy, first half and second half 1974.

Foreign exchange restrictions in Japan, first half 1967 to second half 1970.

Temporary increase in direct and indirect taxation in the Netherlands, first half and second half 1971.

Rapid expansion of organized flights abroad from Sweden, first half and second half 1970.

Shift dummy, first half 1960 to second half 1966.

Foreign exchange restrictions in the United Kingdom, first half 1967 to second half 1969.

Political disturbances in Europe, first half and second half 1968.

Because none of the variables—prices, permanent incomes, or travel flows—fluctuates widely around the long-term trend, it is quite difficult to evaluate the influences of one on the other. This is particularly true when the data contain measurement errors so that cyclical effects cannot be measured precisely. It was felt, however, that the neglect of the effect of long-run influences on travel would result in a misleading interpretation of the permanent income elasticity, and might also lead to mis-specification of the equation. Long-run influences include changes in preferences for foreign travel, cultural and psychological influences, and the effect on imports of spending on advertising and promotion by the travel trade; the time trend was included in the equation to make some allowance for these omitted variables. 32

The dominant factor determining travel and passenger transportation imports is the permanent income variable. The permanent income index was derived using a variety of values for B in the first equation in Section I, Travel and passenger transportation services. The final choice was made on the basis of the minimum standard error criteria, and a value of B= 0.9 was used. The estimated coefficient for this variable, which measures the short-run effect of percentage changes in income over the past three years, is positive and significantly different from zero for 10 of the 14 countries. The sizes of the coefficients on the permanent income variable range from 0.55 to 3.50. The coefficients for Belgium, Canada, Denmark, France, Italy, Japan, and Norway show travel and passenger transportation imports to be very elastic with respect to permanent income; these flows for Switzerland and the United States, however, are quite inelastic. Travel and passenger transportation flows for the Federal Republic of Germany and the United Kingdom vary proportionately with permanent income. A comparison of these estimated income elasticities with those of other studies is not possible, because previously little effort was made to separate income effects from effects of other long-run factors.

Relative prices, which include the effects of exchange rate changes, were found to be of considerable importance in determining travel and passenger fare imports. The sizes of the coefficients indicate a high degree of price elasticity with respect to changes in foreign or domestic prices, or foreign or domestic exchange rates. Only five countries had an estimated price elasticity lower than one. For the majority of countries, the lag distribution implied by the roots of the lag polynomial is plausible; the length of the lag (indicated in square brackets) varies from three to six half years. The average mean lag is 1.8 for the six countries for which the mean lag is significant. The lack of significance of relative prices for Norway and Sweden is probably due to the fact that movements in relative prices were very small over the period of estimation. The significant positive coefficient on the price variable for the Netherlands is thought to reflect speculative currency flows that are recorded as travel expenditure. Seven countries have significant coefficients on the price of transportation variable, and all but one (Italy) show travel and passenger transportation services to be rather inelastic with respect to changes in passenger fares. 33 The price of transportation was dropped from the equation for the United States because the multicollinearity between this variable and the other prices and income variables led to implausible results.

The trend term coefficients are significant and quite large for the Netherlands and Austria. For the Netherlands, the large positive trend term is associated with a low coefficient on the income variable, and for Austria, with a negative income coefficient. A similar problem exists with the estimates for Canada; a large negative trend term is associated with a large positive income coefficient.

To improve the precision of estimation for those countries that had problems associated with multicollinearity, equation (12) was rerun for Austria, Canada, France, the Netherlands, Norway, Switzerland, and the United Kingdom, with the coefficient on the trend term constrained to be equal to 0.0039. This is the average calculated value of the estimated coefficients on the trend term obtained from the results in Table 5. The results for these seven countries using this prior information are presented in Table 6. The precision of the estimated coefficients for permanent income has now improved for all seven countries; this coefficient for Austria now has the correct sign, and the sizes of the coefficients have been reduced for Canada, France, and Norway, and increased for the Netherlands, Switzerland, and the United Kingdom.

Table 6.Fourteen Industrial Countries: Regression Coefficients for Imports of Travel and Passenger Transportation Services with Trend Coefficient Constrained to 0.0039 for Austria, Canada, France, the Netherlands, Norway, Switzerland, and the United Kingdom1
CountryEstimation

Period2
ConstantYPiRMPTViTTrendDummiesR2D-WSEE
Long-run elasticity3Mean lag4Seasonal
Other
S1S2
Austria1967–76–4.12*1.15*–4.02* [4]0.40–0.290.0039–0.35*0.9822.270.062
(0.71)(0.27)(1.84)(0.62)(0.39)(0.03)
Belgium1964–76–6.69*2.06*–4.85 [4]2.19*–0.12*–0.020–0.22*0.14*50.10*60.9942.210.034
(1.60)(0.63)(2.75)(1.06)(0.06)(0.019)(0.01)(0.04)(0.04)
Canada1960–76–2.81*0.69*–1.83* [4]1.13–0.35*0.0039–0.05*0.10*70.24*8–0.08*90.9581.120.066
(0.36)(0.08)(0.37)(0.84)(0.12)(0.02)(0.04)(0.08)(0.04)
Denmark1961–76–5.65*1.79*–2.76* [6]2.10*–0.03–0.023–0.28*0.9871.420.046
(1.62)(0.69)(0.41)(0.73)(0.14)(0.015)(0.02)
France1960–76–10.53*2.31*–2.79* [4]0.89*–0.37*0.0039–0.37*–0.29*7–0.69*100.32*110.15*120.9871.490.065
(0.61)(0.13)(0.54)(0.43)(0.12)(0.03)(0.04)(0.07)(0.06)(0.05)
Germany, Fed. Rep.1960–76–4.34*1.04*–2.78* [4]0.25–0.220.013–0.33*0.13*130.9931.460.052
(1.80)(0.38)(0.42)(0.47)(0.13)(0.009)(0.18)(0.05)
Italy1961–76–11.01*2.56*–1.67*–0.028–0.29*0.18*14–0.1015–0.14*160.9891.810.067
(3.11)(0.78)(0.12)(0.017)(0.02)(0.07)(0.07)(0.05)
Japan1961–76–12.69*2.38*–1.03 [4]2.05–0.011–0.09*–0.19*170.9961.230.069
(1.58)(0.40)(0.58)(1.87)(0.024)(0.02)(0.04)
Netherlands1960–76–5.99*1.64*–0.100.0039–0.23*–0.10180.9841.010.067
(0.18)(0.05)(0.08)(0.02)(0.07)
Norway1964–76–2.55*0.51*–2.29* [4]1.760.0039–0.21*0.9311.010.066
(0.52)(0.21)(0.91)(1.19)(0.03)
Sweden1961–76–5.34*1.25*–1.39 [4]0.34–0.43*0.018–0.16*0.11190.9952.110.040
(1.73)(0.57)(0.87)(1.54)(0.17)(0.011)(0.01)(0.06)
Switzerland1960–76–2.75*0.64*–2.02* [6]1.94*–0.66*0.0039–0.26*0.08*200.9971.670.023
(0.27)(0.09)(0.18)(0.44)(0.11)(0.08)(0.02)
United Kingdom1961–76–6.85*1.56*–0.56* [4]0.31–0.32*0.0039–0.43*–0.28*7–0.12*21–0.10*220.9932.290.031
(0.80)(0.16)(0.21)(1.29)(0.06)(0.13)(0.02)(0.02)(0.03)
United States1961–76–3.740.72*–1.26* [3]2.20*0.021*–0.17*0.07*90.9782.550.049
(2.24)(0.34)(0.26)(0.60)(0.008)(0.02)(0.03)

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the second half of 1976.

The length of the lag is indicated in square brackets.

The mean lag is in half-year units.

Speculative flows in Belgium, first half and second half 1973.

Speculative flows in Belgium, first half and second half 1969.

Shift in seasonal pattern as follows—Canada: S1, first half 1960 to second half 1973; S2, first half 1974 to second half 1976; France and the United Kingdom: S1, first half 1960 to second half 1971; S2, first half 1972 to second half 1976.

Bicentennial year in the United States, first half and second half 1976.

Expo in Canada, first half and second half 1967.

Shift dummy, first half 1960 to second half 1967.

Political disturbances and strikes in France, first half and second half 1968.

Foreign exchange restrictions in France, first half 1968 to second half 1972.

Austrian Olympics and backlog of delayed travel plans, second half 1975 to first half 1976.

Speculative flows in Italy, first half and second half 1971.

Speculative flows in Italy, first half and second half 1972.

Speculative flows in Italy, first half and second half 1974.

Foreign exchange restrictions in Japan, first half 1967 to second half 1970.

Temporary increase in direct and indirect taxation in the Netherlands, first half and second half 1971.

Rapid expansion of organized flights abroad from Sweden, first half and second half 1970.

Shift dummy, first half 1960 to second half 1966.

Foreign exchange restrictions in the United Kingdom, first half 1967 to second half 1969.

Political disturbances in Europe, first half and second half 1968.

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the second half of 1976.

The length of the lag is indicated in square brackets.

The mean lag is in half-year units.

Speculative flows in Belgium, first half and second half 1973.

Speculative flows in Belgium, first half and second half 1969.

Shift in seasonal pattern as follows—Canada: S1, first half 1960 to second half 1973; S2, first half 1974 to second half 1976; France and the United Kingdom: S1, first half 1960 to second half 1971; S2, first half 1972 to second half 1976.

Bicentennial year in the United States, first half and second half 1976.

Expo in Canada, first half and second half 1967.

Shift dummy, first half 1960 to second half 1967.

Political disturbances and strikes in France, first half and second half 1968.

Foreign exchange restrictions in France, first half 1968 to second half 1972.

Austrian Olympics and backlog of delayed travel plans, second half 1975 to first half 1976.

Speculative flows in Italy, first half and second half 1971.

Speculative flows in Italy, first half and second half 1972.

Speculative flows in Italy, first half and second half 1974.

Foreign exchange restrictions in Japan, first half 1967 to second half 1970.

Temporary increase in direct and indirect taxation in the Netherlands, first half and second half 1971.

Rapid expansion of organized flights abroad from Sweden, first half and second half 1970.

Shift dummy, first half 1960 to second half 1966.

Foreign exchange restrictions in the United Kingdom, first half 1967 to second half 1969.

Political disturbances in Europe, first half and second half 1968.

In the initial estimation, four countries had coefficients on price variables with a sign that was not consistent with a priori theoretical considerations; these were the travel price for Italy and the Netherlands and the transportation price for Japan and Norway. In the final estimation, these variables were omitted from the equation, and the omission of the transportation price variable has led to a significant coefficient on the travel price variable for Norway. Unfortunately, autocorrelation in the residuals has resulted from omitting the price variables in the equations for Japan, the Netherlands, and Norway.

The coefficients on the dummies for seasonal effects are significant in all equations and are uniformly negative for the first half year. For three countries (Canada, France, and the United Kingdom), a change in the seasonal patterns was allowed for by including two seasonal dummies. Shift dummies were introduced for two countries (France and Switzerland) for which there was a break in the data definition over the period of observation. 34Other dummies for disturbances had an important impact on travel and passenger transportation flows. For Belgium and Italy, dummy variables are included in the equation to capture the effects of capital movements disguised as travel payments. “Leads and lags” exist because payments for travel are held back because of an expectation of an appreciation of the currency. The positive coefficient on the dummy for foreign exchange restrictions in France from 1968 to 1972 is also thought to be the result of speculative movements recorded as travel flows.

(2) Exports. The main explanatory variables for exports of travel and passenger transportation services are the foreign market share variables, FTV, as can be seen from Table 7. Here we have been reasonably successful in separating the effects of long-run tendencies from the effects of other factors. The coefficient on the foreign market share variable is statistically significant for all 14 countries, and the estimates are concentrated between 0.46 and 0.99 if the two extreme values (0.36 for Norway and 1.32 for the Netherlands) are excluded.

Table 7.Fourteen Industrial Countries: Regression Coefficients for Exports of Travel and Passenger Transportation Services1
RXPTViDummies
CountryEstimation

Period2
ConstantFTViLong-run

elasticity3
Mean

lag
TTrendSeasonalOtherR2D-WSEE
Austria1967–76–0.58*0.58*–3.17* [6]2.95*0.031*–0.14*–0.14*40.9931.400.033
(0.26)(0.15)(0.80)(0.82)(0.013)(0.04)(0.04)
Belgium1963–76–0.82*0.50*0.013*–0.02–0.15*50.9701.730.053
(0.14)(0.14)(0.005)(0.04)(0.06)
Canada1960–760.36*0.74*–1.71* [6]2.05*–0.23–0.007–0.57*0.34*60.9871.430.061
(0.14)(0.33)(0.49)(1.16)(0.16)(0.009)(0.07)(0.05)
Denmark1960–76–0.55*0.61*–1.37* [4]0.100.011–0.26*0.9891.620.039
(0.32)(0.13)(0.57)(1.00)(0.008)(0.03)
France1968–76–0.14*0.63*–0.25 [4]2.030.016*–0.040.9802.580.043
(0.07)(0.23)(0.74)(1.64)(0.008)(0.05)
Germany, Fed. Rep.1961–76–0.36*0.81*–1.79* [8]4.87*0.016*–0.040.9891.250.045
(0.07)(0.20)(0.55)(1.95)(0.008)(0.05)
Italy1960–760.62*0.58*–0.007–0.30*–0.20*70.9841.620.050
(0.05)(0.14)(0.006)(0.04)(0.03)
Japan1961–76–1.54*0.99*–2.46* [3]0.650.051*0.13*0.28*80.9722.530.077
(0.53)(0.24)(0.70)(0.56)(0.015)(0.05)(0.04)
Netherlands1960–760.94*1.32*–0.98 [3]0.82–0.027*0.15*–0.27*90.9321.320.083
(0.35)(0.19)(0.72)(2.10)(0.010)(0.06)(0.09)
Norway1964–76–1.59*0.36*–1.34 [4]0.590.022*–0.34*0.9811.320.048
(0.61)(0.20)(0.84)(2.03)(0.011)(0.05)
Sweden101960–76–1.83*0.46*–2.19 [4]1.940.048*–0.22*0.9891.750.037
(0.45)(0.10)(1.29)(1.38)(0.010)(0.05)
Switzerland1969–760.130.68*–0.97* [5]1.08–0.005–0.010.9872.240.028
(0.12)(0.12)(0.15)(1.06)(0.006)(0.03)
United Kingdom1964–76–0.18*0.66*–2.25* [5]1.92*0.022*–0.18*0.9961.630.038
(0.08)(0.14)(0.43)(0.51)(0.006)(0.03)
United States1960–760.16*0.68*–1.21* [6]2.72–0.21*0.0060.07*0.9962.120.028
(0.06)(0.09)(0.51)(1.60)(0.07)(0.005)(0.02)

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the second half of 1976 except for Japan, which began with the second half of 1961.

The length of the lag is indicated in square brackets.

Change in vacation date from June to July for travelers from the Federal Republic of Germany, first half 1974.

Speculative flows in Belgium, first half and second half 1972.

Expo in Canada, first half and second half 1967.

Shift dummy, first half 1960 to first half 1965.

Osaka Exhibition in Japan, first half and second half 1970.

Speculative flows, first half and second half 1970.

Corrected for first-order autocorrelation. The first-order autoregressive coefficient was 0.914 (0.020).

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the second half of 1976 except for Japan, which began with the second half of 1961.

The length of the lag is indicated in square brackets.

Change in vacation date from June to July for travelers from the Federal Republic of Germany, first half 1974.

Speculative flows in Belgium, first half and second half 1972.

Expo in Canada, first half and second half 1967.

Shift dummy, first half 1960 to first half 1965.

Osaka Exhibition in Japan, first half and second half 1970.

Speculative flows, first half and second half 1970.

Corrected for first-order autocorrelation. The first-order autoregressive coefficient was 0.914 (0.020).

Table 8.Canada, the United Kingdom, and the United States: Regression Coefficients for Inflows of Investment Income1
Dependent

Variable
Estimation

Period2
ConstantLagged

Dependent
DummiesR2D-WhSEE
Seasonal
S1S2Other
Inflows of Direct Investment Income
ASSRPfCYf
yidica1962–76–5.08*1.18*–0.261.50*0.06080.22*–0.16 30.59*40.9511.780.8800.130
(1.59)(0.27)(0.40)(0.62)(0.1379)(0.09)(0.09)(0.14)
yidiuk1962–76–5.26*1.19*–0.64*1.45*0.5981*–0.12*5–0.14*60.9892.050.1960.060
(1.23)(0.29)(0.20)(0.35)(0.1551)(0.05)(0.04)
yidius1961–76–5.13*0.58*0.170.76*0.4051*–0.08*–0.08*70.9962.311.1880.034
(1.38)(0.13)(0.12)(0.19)(0.1236)(0.01)(0.01)
Inflows of Financial Investment Income
rf. ASS
yifica1962–76–0.90*0.42*0.5699*–0.61*–0.27*8–0.45*9–0.38*100.9851.531.3790.081
(0.13)(0.07)(0.0655)(0.07)(0.04)(0.08)(0.08)
yifiuk1962–76–0.88*0.06*0.6013*–0.14*110.12*120.15*50.8162.130.4520.051
(0.23)(0.02)(0.1037)(0.05)(0.05)(0.06)
yifius1961–76–3.10*0.48*0.5084*0.16*50.9981.541.4680.037
(0.54)(0.08)(0.0818)(0.03)
Inflows of Other Investment Income
yioica1965–76–1.13*0.43*0.4753*0.17*50.31*130.9742.120.3520.101
(0.27)(0.10)(0.1203)(0.07)(0.11)
yioiuk1962–76–0.78*0.17*0.5007*0.30*0.36*5–0.29*140.9462.260.8000.115
(0.14)(0,04)(0.1109)(0.05)(0.08)(0.08)
yioius1961–76–4.98*0.52*0.2445–0.22*–0.31*50.9462.020.0870.094
(0.91)(0.09)(0.1343)(0.04)(0.10)

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the second half of the first year indicated to the second half of 1976 except for the United States, which began with the first half of the first year indicated.

International utility reclassified as U.S. instead of Canadian foreign company, first half and second half 1971.

Tax cut in the United States, first half and second half 1964.

Oil crisis, first half 1974 to second half 1974.

Policies designed to restrict outflows of U.K. direct investment, first half 1965 to second half 1966.

Capital restrictions on U.S. foreign direct investment, first half 1965 to first half 1972.

Shift in seasonal pattern: SI, second half 1962 to first half 1966; S2, second half 1966 to second half 1976.

Foreign exchange crisis in Canada, first half 1968.

Large outflow of banking funds and increase in redemption of Canadian securities, first half 1967.

Increase in portfolio investment in second half 1967, associated with mining boom in Australia.

Exchange control measures in the United Kingdom, first half 1965.

Abolition of target level for Canadian exchange reserves under the arrangement with the United States in December 1968 had impact in first half 1969.

Suez Canal closure in 1967 affected costs of oil companies, first half and second half 1968.

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the second half of the first year indicated to the second half of 1976 except for the United States, which began with the first half of the first year indicated.

International utility reclassified as U.S. instead of Canadian foreign company, first half and second half 1971.

Tax cut in the United States, first half and second half 1964.

Oil crisis, first half 1974 to second half 1974.

Policies designed to restrict outflows of U.K. direct investment, first half 1965 to second half 1966.

Capital restrictions on U.S. foreign direct investment, first half 1965 to first half 1972.

Shift in seasonal pattern: SI, second half 1962 to first half 1966; S2, second half 1966 to second half 1976.

Foreign exchange crisis in Canada, first half 1968.

Large outflow of banking funds and increase in redemption of Canadian securities, first half 1967.

Increase in portfolio investment in second half 1967, associated with mining boom in Australia.

Exchange control measures in the United Kingdom, first half 1965.

Abolition of target level for Canadian exchange reserves under the arrangement with the United States in December 1968 had impact in first half 1969.

Suez Canal closure in 1967 affected costs of oil companies, first half and second half 1968.

The average value of the foreign demand parameters for the 14 industrial countries is 0.68. Since the average value of the coefficients on foreign market variables for the world as a whole must equal one, an average coefficient of 0.68 for the 14 industrial countries tends to indicate that exports of travel and passenger transportation flows for the rest of the world are rather more sensitive to short-run changes than are exports for the 14 industrial countries. The possibility of statistical bias in this number, however, should be taken into account.

The estimated long-run travel price elasticity, which measures the change in a country’s travel price relative to that of its competitors, is significant at the 95 per cent level for eight countries. The sizes of these coefficients indicate that the country in question exports predominantly to markets that are highly price sensitive to changes in its export prices relative to those of its competitors. In the initial estimation run, the coefficient on the long-run travel price variable was positive for Belgium (0.004 (0.800)) and Italy (0.77(0.31)); the estimate for Italy is probably the result of speculative flows that are recorded as travel flows. In the final estimation, the travel price variable for these two countries was omitted from the equation. Constraining these parameters to zero had very little effect on the coefficients for the other exogenous variables. The average of the export competitors’ price elasticities, calculated for the remaining 12 countries, is -1.64. The import and export price elasticities taken together with the ratio of imports to exports show that a fall in the home travel price (or depreciation of the domestic exchange rate) relative to the foreign travel price (or foreign exchange rate) will improve the travel and passenger transportation account of the balance of payments of all countries but Italy and the Netherlands.

The price of transportation variable, included for all countries in the initial estimation, produced coefficients that were not significantly different from zero for all countries except Canada and the United States. Therefore, except for those two countries, this variable was omitted in the final estimation. The transportation price variable is the same for each country, and if this price is the same proportion of the total travel price for each country, it is obvious that a change in the transportation price will have no effect on market shares. A fall in the price of transportation for Canada and the United States, however, may lead to a significant fall in the total price of travel and transportation. This is so because the air fare constitutes a larger proportion of the total price for these countries than for many other countries, since the average distances involved are quite large. The impact of the price of transportation for individual countries may be distorted by changes in the pattern of routes flown by passengers or by the effects of reductions in charter rates for individual countries; these may be important reasons for the lack of significance of the price of transportation variable.

Dummy variables to allow for the effects of international fairs and unusual events were included in various equations. The change in vacation dates from June 1974 to July 1974 for a substantial number of Germans, because of a shift in school holidays for that year only, led to a significant fall in travel flows to Austria for the first half of 1974. For Belgium and the Netherlands, capital flows disguised as travel flows led to receipts being held back because of an expected currency depreciation. The effects of disturbance factors were also quite large for Canada (Expo 1967) and Japan (Osaka Exhibition 1970).

Flows of investment income

(I) Gross flows approach. The estimated coefficients of equations (18), (21), (22), and (23) for inflows and outflows of direct, financial, and other investment income are presented in Tables 8 (inflows) and 10 (outflows) for the three countries (Canada, the United Kingdom, and the United States) for which data on gross flows are available. The empirical results are discussed in two sections: (a) the inflow equations, and (b) the outflow equations.

(a) Inflow equations. Although the overall explained variance exceeds 0.94 in all equations except that for financial investment income for the United Kingdom, in some cases the standard errors of estimate are rather unsatisfactory. The worst equations, in general, are those that refer to direct investment and other investment, and the use of proxy rates of return in these equations is thought to account in part for these large standard errors. The h-statistic 35 shows that for all equations autocorrelation in the residuals is not a serious problem. Table 9 contains the impact (or short-run) and steady-state (or final adjustment) coefficients on the asset and liability levels and on the rate of return variables. For equations for inflows of direct investment income, b1, b2, and b3 are the impact coefficients on the asset (ASS), long-run rate of return (RPf), and cyclical rate of return (CYf), respectively. The steady-state coefficients are calculated as θ1 = b1(1–b5), θ2 = b2/(1–b5), and θ3 = b3/(1–b5), where b5 is the coefficient of the lagged dependent variable. For equations for inflows of financial and other investment income, b1 is the impact coefficient on the product of the rates of return and the asset level, and the steady-state coefficients are calculated as θ1 = b1/(1–b2).

Table 9.Canada, the United Kingdom, and the United States: Impact and Steady-State Coefficients for Flows of Investment Income1
Investment

Income Group and

Country
InflowsOutflows
ImpactSteady-StateImpactSteady-State
b1b2b3θ1θ2θ3a1a2a3Ω1Ω2Ω3
(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)
dica1.18*–0.261.50*1.26*–0.271.59*0.77*–0.270.821.27*–0.441.35
diuk1.19*–0.64*1.45*2.96–1.59*3.610.72*–0.34–0.551.34–0.63*–1.02
dius0.58*0.170.76*0.98*0.29*1.28*0.50*0.28*0.92*0.71*0.40*1.30*
fica0.42*0.98*0.32*1.08*
fiuk0.06*0.15*0.10*0.57
fius0.48*0.98*0.80*1.23*
oica0.43*0.82*0.41*1.06*
oiuk0.17*0.34*0.56*0.45*
oius0.52*0.52*

An asterisk indicates that the estimate is significantly different from zero at the 95 per cent confidence level. The standard errors on the long-run coefficients were computed using the formula derived in Mood, Graybill, and Boes (1974), pp. 180–81.

An asterisk indicates that the estimate is significantly different from zero at the 95 per cent confidence level. The standard errors on the long-run coefficients were computed using the formula derived in Mood, Graybill, and Boes (1974), pp. 180–81.

In the equations for inflows of direct investment income, the estimated coefficients on the asset level b1 in equation (22) are positive and significant at the 95 per cent level for all three countries. The b1 coefficient shows the effect of an increase in asset levels on direct investment income inflows in the short run; an increase in asset levels abroad will be met by a slightly greater proportionate rise in income inflows in the short run for Canada and the United Kingdom but by a less than proportionate rise for the United States. The steady-state coefficient is close to one for Canada and the United States, but implausibly high (2.96) for the United Kingdom.

The coefficient on the long-run rate of return (a weighted average of the government bond yield in foreign countries) is poorly determined for all countries. The estimates of the cyclical rate of return variables are highly significant for all three countries. An increase of 1 per cent in the ratio of actual to potential output in manufacturing in foreign countries in which the assets are held will, on impact, cause direct investment income inflows to rise by 1.50 per cent, 1.45 per cent, and 0.76 per cent for Canada, the United Kingdom, and the United States, respectively. The steady-state coefficients (θ3) range from 1.28 to 3.61.

The variable included in equation (22) to represent the relative profitability (PRi) of domestic and foreign investment was insignificant for Canada and the United States and was therefore dropped from the equation. We can conclude that either relative profitability is not important in the determination of the proportion of earnings repatriated or the variable selected to represent relative profitability was a poor choice.

In the equations for inflows of financial and other investment income, all the b1 coefficients on the rate of return variable (rf. ASS) have the predicted positive sign and are significant at the 95 per cent level. Both the impact and the steady-state coefficients appear to be rather small for the U. K. financial and other investment income equations. 36

The lagged dependent variable is significant in seven of the nine inflow equations. In the direct investment equation, this variable captures the lagged effect of income inflows to higher or lower levels of earnings; in the financial and other investment income equations, it captures the lagged effect of past interest rates on current inflows, since, although current interest rates are paid on new issues, past issues are generally contracted at the interest rate prevalent on the day of the issue. 37 The partial adjustment coefficients that are significantly different from zero and from one range from 0.39 to 0.59. These speed of adjustment coefficients reveal that direct investment income inflows are fairly inelastic with respect to changes in earnings, and financial and other investment income inflows are fairly inelastic with respect to changes in foreign interest rates. Since noneconomic factors must be expected to dominate the determination of government inflows, the absence of lagged adjustment in the U. S. “Other investment” equation is not surprising.

The degree of seasonality is substantial for some of the equations, but it is not uniform. For inflows of Canadian financial investment income, seasonality was found to be significantly greater in the first half of the estimation period than in the second half. Other dummy variables were included in the equations to allow for the effects of special events that may affect the proportion of earnings repatriated, 38 the oil crisis, restrictions, and other special disturbances. The oil crisis was found to have a positive effect on the U. K. and the U. S. financial investment income inflows and on Canadian and U. K. other investment income inflows, and a negative effect on U. K. direct and U. S. other income inflows. 39 Such results must of course be regarded with caution, since the dummy variables may pick up the effects of omitted variables and other measurement errors over that unstable period. The foreign exchange crisis in the first half of 1968 in Canada had negative effects on Canadian financial investment income inflows. These inflows were postponed because of the expectation of devaluation of the currency. The large outflows of banking funds and the increased redemption of Canadian securities as part of the government agreement with the United States to reduce reserves was responsible for the reduction in financial inflows to Canada in the first half of 1967. The mining boom, which greatly increased portfolio investment in Australia in the second half of 1967, was responsible for the fall in financial investment income inflows to the United Kingdom for that period. The exchange control measures in the first half of 1965, ending the ruling whereby U. K. residents with holdings of foreign securities were allowed to use all foreign currency received as returns on securities or from their sale for reinvestment in foreign securities, led to an increase in financial investment income inflows.

(b) Outflow equations. The parameter estimates for equations (18) and (21) for the investment income outflow equations are presented in Table 10, and the impact and steady-state coefficients are itemized in columns 7 to 12 of Table 9. Column 7 shows that the coefficients on direct investment liability levels are measured quite precisely and have the expected positive signs. In the short run, the liability level is inelastic with respect to income outflows for the United States. The long-run elasticities, Ω1, however, are not significantly different from one for all three countries.

Table 10.Canada, the United Kingdom, and the United States: Regression Coefficients for Outflows of Investment Income1
Dummies
Dependent

Variable
Estimation

Period
ConstantLagged

Dependent
Seasonal
Other
R2D-WhSEE
S1S2
Outflows of Direct Investment Income
UARPhCYh
yodica1963–762–3.71*0.77*–0.270.820.3927*–0.22*–0.42*3–0.22*40.2050.9592.361.6170.095
(1.41)(0.20)(0.28)(0.51)(0.1487)(0.07)(0.06)(0.10)(0.11)
yodiuk1962–766–3.18*0.72*–0.34–0.550.4622*0.29*7–0.19*80.9522.170.6480.118
(1.30)(0.30)(0.26)(0.80)(0.1903)(0.14)(0.07)
yodius1961–762–4.71*0.50*0.28*0.92*0.2928*0.39*90.19*10–0.11*110.9952.481.2520.044
(1.10)(0.08)(0.14)(0.16)(0.0687)(0.03)(0.04)(0.04)
Outflows of Financial Investment Income
LIA
yofica1963–762–0.60*0.32*0.7045*0.06*0.18*120.9971.890.2940.030
(0.16)(0.07)(0.0742)(0.01)(0.04)
yofiu1962–766–0.37*0.10*0.8254*0.12*0.19130.9602.361.1450.097
(0.18)(0.05)(0.0904)(0.04)(0.11)
yofius1961–762–1.16*0.80*0.3493*–0.20*140.9921.561.4980.078
(0.18)(0.11)(0.0931)(0.06)
Outflows of Other Investment Income
yooica1965–766–0.65*0.41*0.6118*0.030.12*90.9922.150.3820.048
(0.15)(0.08)(0.0794)(0.02)(0.05)
yooiuk1962–766–2.42*0.56*–0.2298–0.12*0.43*130.9581.621.6090.109
(0.35)(0.08)(0.1634)(0.05)(0.12)
yooius1961–7620.20–0.00041.00*0.40*150.9941.481.5100.077
(0.12)(0.0004)(0.04)(0.08)

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the second half of 1976.

Shift in seasonal pattern: SI, first half 1963 to second half 1965; S2, first half 1966 to second half 1976.

Exchange rate uncertainty in Canada, first half 1972.

Superior cyclically related profit performance of foreign firms in Canada, second half 1975.

The period covers the second half 1962 to the second half 1976.

Coal and dock strikes, first half 1973.

Weakness of sterling in anticipation of possible devaluation, first half 1966 to second half 1967.

Oil crisis, second half 1973 to second half 1974.

Dummy shift, first half 1971 to second half 1976.

Dummy shift, first half 1961 to first half 1966.

Pressure on Canadian dollar, second half 1976.

Reopening of Suez Canal, first half 1975.

Foreign exchange crisis, second half 1970 to first half 1971.

Oil crisis, second half 1973.

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the second half of 1976.

Shift in seasonal pattern: SI, first half 1963 to second half 1965; S2, first half 1966 to second half 1976.

Exchange rate uncertainty in Canada, first half 1972.

Superior cyclically related profit performance of foreign firms in Canada, second half 1975.

The period covers the second half 1962 to the second half 1976.

Coal and dock strikes, first half 1973.

Weakness of sterling in anticipation of possible devaluation, first half 1966 to second half 1967.

Oil crisis, second half 1973 to second half 1974.

Dummy shift, first half 1971 to second half 1976.

Dummy shift, first half 1961 to first half 1966.

Pressure on Canadian dollar, second half 1976.

Reopening of Suez Canal, first half 1975.

Foreign exchange crisis, second half 1970 to first half 1971.

Oil crisis, second half 1973.

Again, as for income inflows, the coefficients on the permanent rate of return are measured rather imprecisely, except for the coefficient on U. S. outflows of direct investment income; this coefficient implies that an increase of 1 per cent in the U. S. Government bond yield will lead to an increase of 0.28 per cent in the U. S. income outflow. In general, it is felt that the long-run bond rate is a poor approximation to the actual long-run rate of return on direct investment. The collinearity between the long-run rate of return and the liability stock variable may be responsible for the wrong sign on the coefficients for Canada and the United Kingdom. The measures of the cyclical rate of return on direct investment liabilities were significant in the equation for the United States; the estimated elasticity implies that an increase of 1 per cent in the ratio of actual to potential output in manufacturing in the United States will cause direct investment income outflows to rise by 0.92 per cent.

The measures of the rate of return variables, a1 in equation (21) are significant in five of the six financial and other investment income equations. This measure includes three elements: the liability level, the long-run interest rate (government bond yield), and the short-run interest rate (discount rate). The values of the coefficients imply low-income outflow elasticities with respect to earnings. The steady-state coefficients are larger than the impact coefficients and are quite close to unity for three of the six equations.

The partial adjustment coefficients, significant at the 95 per cent level in seven of the nine equations, are 0.6073, 0.5378, 0.7072, 0.2955, 0.1746, 0.6507, and 0.3882 in equations (1), (2), (3), (4), (5), (6), and (7), respectively. Outflows of U. K. financial investment income and of Canadian other investment income are rather inelastic in the short run with respect to changes in interest rates. On the other hand, the partial adjustment coefficients for U. S. direct and financial investment income equations show these income outflows to be more elastic with respect to earnings and interest rates. The partial adjustment coefficient for U. S. other (government) income outflows is equal to zero. This implies that last period’s situation prevails this period, and since non-economic factors dominate in the determination of government outflows, this result is not surprising. When the lagged dependent variable is excluded from the equation, the rate of return variable becomes significant with the expected positive sign. The coefficient on the lagged dependent variable for the U. K. other investment equation is negative but not significant at the 95 per cent level. 40

Dummy variables were included to allow for the effects of the oil crisis, exchange rate uncertainty, and other special factors. The expectation of an appreciation in the exchange rate following exchange rate uncertainty in Canada in the first half of 1972 means that income outflows for Canadian direct investment were postponed because of the expectation that the transfer would require less currency at a later date. On the other hand, coal and dock strikes in the United Kingdom in the first half of 1973 led to an increase in direct investment outflows because of uncertainty about the future of the economy, and fear of further pressure on the Canadian dollar in the second half of 1976 caused outflows of Canadian financial investment to be brought forward. The dummy shift variable for outflows of U. S. direct investment income for the first half of 1961 to the first half of 1966 represents the different definition of data following the 1966 benchmark survey by the U.S. Office of Business Economics and, for the first half of 1971 to the second half of 1976, revised data on earnings and therefore on income outflows, which include investments not previously recorded.

(2) Net flows approach. Equation (27) was estimated for 11 industrial countries for which data on stocks of foreign assets and liabilities are not generally available.

Table 11 shows the calculated coefficients of net earnings with respect to net flows of investment income; this coefficient had the correct sign and is statistically significant at the 95 per cent level for all 11 countries. For the majority of countries, the U. S. Government bond yield (GBYUS) gave the best results in terms of R2 and standard errors; however, for Austria and the Netherlands, the U.S. discount rate (DRUS) was superior, and for Switzerland, the U. S. commercial bank rate (CBRUS) gave the best results. In all cases excluding Switzerland, however, the size of the coefficient is considerably less than one. There could be several reasons for these low coefficients. 41 First, if data on net flows of investment income exclude reinvested earnings, as they do for the Federal Republic of Germany, and the country is a net recipient, downward bias in the coefficient will occur. Second, if the country’s assets are held in gold rather than in the form of assets abroad, the country’s net asset position on which a return is earned will be overestimated by the current account balance, leading to considerable downward bias in the c1 coefficient. This may provide an explanation for the low c1 coefficient for France. Third, if the data for the rate of return and the net asset level contain errors of measurement, this will lead to bias of a downward nature. Switzerland is a net recipient of investment income flows, and for this country the c1 coefficient of 1.4 is rather high. Swiss banks are involved in a large amount of financial intermediation, and the assumption that the foreign rate of return is equal to the domestic rate of return may have biased the c1 coefficient upward in the equation for Switzerland. Since the constant term is significant in the equations for 8 of the 11 countries, this lends further support to the hypothesis that our proxy variables are not particularly good measurements for the earnings variables.

The seasonal dummy variable coefficients are significant in 5 of the 11 equations and are uniformly negative for the first half year; the seasonal effect is particularly large for the Netherlands. The significance of this seasonal dummy variable could be accounted for by year-end accounting practices that lead to higher inflows or outflows in the second half of the year. An oil crisis dummy was found to be significant in most equations and has a negative coefficient in all but three equations—namely, Austria, France, and Switzerland. Its effects were particularly large for the Federal Republic of Germany (–0.39) and Switzerland (0.43). For France, the effects of the oil crisis appear to be of a more permanent nature than for the other ten countries. The dummies for other disturbances, such as exchange rate speculation and restrictions on capital inflows and outflows, are documented in the appropriate footnotes to Table 11.

Table 11.Eleven Industrial Countries: Regression Coefficients for Net Flows of Investment Income1
CountryEstimation Period2Interest RateConstantRA * NASSDummiesR2D-WSEE
SeasonalOil3Other
S1S2
Austria1970–76drus–0.0010.53*0.03*0.8011.600.014
(0.009)(0.08)(0.01)
Belgium1962–76gbyus–0.03*0.58*–0.05*–0.07*0.9351.210.026
(0.01)(0.03)(0.01)(0.02)
Denmark1970–76gbyus0.03*0.52*0.02–0.06*0.03*40.9631.690.014
(0.01)(0.04)(0.01)(0.01)(0.01)
France1970–76gbyus0.21*0.24*–0.030.26*0.09*50.8862.280.043
(0.02)(0.10)(0.02)(0.04)(0.04)
Germany,1966–76gbyus–0.23*0.26*0.04–0.39*0.7301.500.132
Fed. Rep.(0.07)(0.04)(0.05)(0.07)
Italy1970–76gbyus–0.73*0.65*–0.11*–0.18*0.22*60.9502.340.079
(0.07)(0.09)(0.04)(0.06)(0.04)
Japan71961–76gbyus–0.23*0.46*–0.31*–0.31*80.9732.410.049
(0.03)(0.04)(0.05)(0.02)
Netherlands1970–76drus0.090.53*–0.14*–0.35*–0.2190.8351.700.098
(0.06)(0.19)(0.05)(0.09)(0.12)
Norway1970–76gbyus0.010.46*–0.02*–0.04*0.9571.980.018
(0.01)(0.03)(0.01)(0.01)
Sweden1970–76gbyus0.02*0.31*–0.03*–0.09*100.9172.040.014
(0.01)(0.06)(0.01)(0.01)
Switzerland1970–76cbrus0.30*1.40*0.43*0.8761.560.112
(0.06)(0.19)(0.07)

Standard errors are shown in parentheses. All variables are in levels. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the second half 1976.

Periods covering the oil crisis were Austria, second half 1973 to first half 1974 (–1, 1); Belgium, second half 1973 to first half 1974; Denmark, second half 1973; France, second half 1973 to second half 1974; the Federal Republic of Germany, second half 1973 to first half 1975 (–1,1,1,1); Italy, second half 1974 to first half 1975; Japan, second half 1974; the Netherlands, second half 1974 to first half 1976; Norway, second half 1973 to second half 1974; and Switzerland, second half 1973 to second half 1975.

Restriction on capital inflows to ensure inflows used to finance real investment, second half 1973.

Leads and lags favorable to franc, owing to U.S. dollar crisis, first half 1973.

Statistics distorted by measures taken to reduce the exports of Italian banknotes, first half 1970 to first half 1972.

Corrected for first-order autocorrelation. The first-order autoregressive coefficient was –0.4852 (0.2561).

Reversal of policy measures adopted in 1972 to encourage capital outflows, first half 1974 to second half 1976.

Leads and lags unfavorable to guilder, first half 1975.

The second seasonal dummy for Sweden is for the first half 1974 to the second half 1976.

Standard errors are shown in parentheses. All variables are in levels. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the second half 1976.

Periods covering the oil crisis were Austria, second half 1973 to first half 1974 (–1, 1); Belgium, second half 1973 to first half 1974; Denmark, second half 1973; France, second half 1973 to second half 1974; the Federal Republic of Germany, second half 1973 to first half 1975 (–1,1,1,1); Italy, second half 1974 to first half 1975; Japan, second half 1974; the Netherlands, second half 1974 to first half 1976; Norway, second half 1973 to second half 1974; and Switzerland, second half 1973 to second half 1975.

Restriction on capital inflows to ensure inflows used to finance real investment, second half 1973.

Leads and lags favorable to franc, owing to U.S. dollar crisis, first half 1973.

Statistics distorted by measures taken to reduce the exports of Italian banknotes, first half 1970 to first half 1972.

Corrected for first-order autocorrelation. The first-order autoregressive coefficient was –0.4852 (0.2561).

Reversal of policy measures adopted in 1972 to encourage capital outflows, first half 1974 to second half 1976.

Leads and lags unfavorable to guilder, first half 1975.

The second seasonal dummy for Sweden is for the first half 1974 to the second half 1976.

Other private services

The heterogeneous nature of the “Other private services” group of invisibles discourages simple analysis. In view of the poor quality of the data of this group, however, it was decided to keep the analysis at a reasonably simple level and to assume that other private services are a function of GNP and relative prices; to the extent that items included in other private services (such as cargo insurance) respond to factors affecting merchandise trade, no allowance has been made for these effects.

Estimates of the distributed lags on the relative price variable were achieved by use of a polynomial distributed lag. A polynomial of degree two was used, and for some countries a zero constraint was imposed on the value of the end-period coefficient. 42

(I) Imports. The import volume equations for other private services, presented in Table 12, are characterized by rather large standard errors of estimate, and are not very satisfactory. In many cases, large standard errors of the estimates are due to the use of a simple model to represent a highly heterogeneous group of items. For example, the omission of trade flows as an explanatory variable is a shortcoming for countries for which merchandise insurance and commissions are a significant item in the other private services account.

Table 12.Fourteen Industrial Countries: Regression Coefficients for Imports of Other Private Services1
RMPOPiDummies
CountryEstimation

Period
ConstantGNPiLong-run elasticity3Mean lag4TrendSeasonal
Other
R2D-WSEE
Austria1965–76–5.83*

(0.85)
1.66*

(0.40)
0.007

(0.009)
–0.08*

(0.02)
0.9761.860.057
Belgium51965–76–5.52*

(1.78)
1.72*

(0.67)
–1.51 [4]

(0.86)
3.01*

(1.03)
–0.007

(0.018)
0.02

(0.02)
0.9671.530.057
Canada51961–76–3.06*

(0.49)
0.70*

(0.11)
–0.87* [4]

(0.30)
2.19

(1.24)
0.00980.03

(0.02)
–0.18*6(0.04)0.9651.300.060
Denmark71965–76–6.25*

(2.22)
1.78

(0.95)
–1.93 [3]

(1.10)
0.93

(1.43)
–0.016

(0.024)
–0.76*8(0.04)0.9881.690.073
France1968–76–4.02*

(1.10)
0.85*

(0.22)
–0.10 [4]

(0.19)
2.35

(4.10)
0.00980.14*9 0.16*10

(0.04)

(0.05)
0.9641.980.042
Germany, Fed. Rep.1963–76–1.85*

(0.71)
0.45*

(0.15)
0.032*

(0.004)
–0.11*11(0.03)0.9881.700.035
Italy1968–76–6.40*

(2.70)
1.44*

(0.66)
–2.63* [6]

(0.98)
3.78

(1.73)
0.009

(0.012)
–0.04*

(0.02)
0.9612.880.049
Japan1961–76–3.63*

(0.68)
0.65*

(0.17)
–1.32* [4]

(0.34)
1.87*

(0.81)
0.025*

(0.009)
0.02*

(0.01)
0.9961.520.041
Netherlands1965–76–3.39

(3.09)
0.91

(1.27)
–1.74 [6]

(2.85)
5.71

(6.72)
0.006

(0.060)
0.9752.510.061
Norway1967–76–5.79*

(1.60)
1.63*

(0.66)
–4.50* [5]

(0.92)
1.91*

(0.75)
0.00980.49*12

(0.08)
0.9782.390.096
Sweden1967–76–17.17*

(1.41)
4.62*

(0.40)
–12.22* [6]

(5.57)
1.37

(1.31)
0.0098–0.43*13

(0.08)
0.48*14

(0.09)
0.9621.770.127
Switzerland151965–76–3.51*

(0.38)
0.54*

(0.13)
–1.37* [4]

(0.11)
1.64*

(0.51)
0.00980.012*

(0.006)
0.9732.210.029
United Kingdom1963–76–1.45*

(0.36)
0.30*

(0.07)
0.00980.15*16

(0.02)
–0.09*17

(0.02)
0.8221.580.034
United States1966–76–9.64*

(3.62)
1.30*

(0.54)
–1.79* [4]

(0.22)
0.48

(0.50)
0.023*

(0.008)
0.8812.230.044

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the second half of 1976.

The length of the lag is indicated in square brackets.

The mean lag is in half-year units.

For Belgium and Canada, “Total Other Services” were estimated.

Prior to increase in Canadian foreign aid in services, first half 1961 to second half 1966.

Corrected for first-order autocorrelation. First-order autoregressive coefficient was –0.750 (0.156).

Shift dummy, first half 1968 to second half 1970, prior to change in statistical recording after first half 1971.

Shift dummy, first half 1973 to second half 1976, owing to change in recording methods.

Heavy flow of earnings from foreign engineering consulting services, second half 1976.

Shift dummy, first half 1974 to second half 1976, because of break in data series definition.

Unfavorable conditions in shipping sector, first half and second half 1969, led to fall in earnings of foreign crews in Norway.

Temporary rise in savings ratio, first half 1972 to second half 1973.

Discrepancy in statistics, first half 1975 to second half 1976.

Corrected for second-order autocorrelation +1.00 (0.173) first, –0.647 (0.173) second.

Rise in imports of other services associated with imports of North Sea investment, second half 1974 to second half 1975.

Coal and dock strikes and labor disputes, first half and second half 1973.

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the second half of 1976.

The length of the lag is indicated in square brackets.

The mean lag is in half-year units.

For Belgium and Canada, “Total Other Services” were estimated.

Prior to increase in Canadian foreign aid in services, first half 1961 to second half 1966.

Corrected for first-order autocorrelation. First-order autoregressive coefficient was –0.750 (0.156).

Shift dummy, first half 1968 to second half 1970, prior to change in statistical recording after first half 1971.

Shift dummy, first half 1973 to second half 1976, owing to change in recording methods.

Heavy flow of earnings from foreign engineering consulting services, second half 1976.

Shift dummy, first half 1974 to second half 1976, because of break in data series definition.

Unfavorable conditions in shipping sector, first half and second half 1969, led to fall in earnings of foreign crews in Norway.

Temporary rise in savings ratio, first half 1972 to second half 1973.

Discrepancy in statistics, first half 1975 to second half 1976.

Corrected for second-order autocorrelation +1.00 (0.173) first, –0.647 (0.173) second.

Rise in imports of other services associated with imports of North Sea investment, second half 1974 to second half 1975.

Coal and dock strikes and labor disputes, first half and second half 1973.

Furthermore, identifying short-term and long-term influences was difficult for some countries. This was due in part to the high degree of measurement errors contained in the dependent variable. An additional handicap encountered with the “Other private services” category of invisibles is that changes in statistical recording or discrepancies in the statistics occurred over the estimation period. Where these changes are noted, they are allowed for in the regression by the inclusion of dummy variables; nevertheless, it is felt that these changes also hampered our efforts to evaluate the influences of prices, incomes, and trend on imports of other private services. The lack of wide fluctuations in income around the long-run trend also hampered these efforts. The coefficients on the trend term were constrained to equal 0.0098 for six countries where collinearity problems were particularly severe; this is the average value of the estimated coefficients on the time trend for the remaining eight countries.

The main explanatory variable for other private service imports is GNP; seven countries have short-term elasticities of less than one, and six have short-term elasticities between one and two. The coefficient for Sweden seems excessively large. The precision of estimation for two countries (Denmark and the Netherlands) was so poor that the income variable was insignificant at the 95 per cent level.

The relative price variable that measures the long-run elasticity on the foreign/domestic price variable (RMPOPi) is significant for only 7 of the 14 industrial countries. For 6 of the 7 countries, the sizes of the coefficients indicate that other private service imports are quite sensitive to changes in foreign or domestic prices and foreign or domestic exchange rates, although the coefficients for Norway and Sweden exceed four and seem to be excessively large. Only Canada has an elasticity of less than one. These estimates of price elasticities should be viewed with care, however, because of possible measurement errors in the price variable. 43

Shift dummies were included for three countries (Denmark, France, and the Federal Republic of Germany) for which changes in the data definition occurred over the estimation period. Other dummies are documented in the footnotes to Table 12.

(2) Exports. The foreign market share variable, FOP, is significant for only 3 of the 14 industrial countries, as can be seen from Table 13. Since the “Other private services” item of flows of invisibles is less sensitive to cyclical factors than are other items, collinearity between the foreign market share variable and the trend term is particularly severe, and the foreign market variable cannot be measured too precisely. For three countries (Denmark, the Federal Republic of Germany, and the United States), the market coefficient was constrained to equal one because of the dominance of the trend term. The estimated coefficient on the market variable did not differ significantly from one for eight countries; the exceptions are Canada, Switzerland, and the United Kingdom. The large estimated market coefficient for Canada is accompanied by a large negative trend term.

Table 13.Fourteen Industrial Countries: Regression Coefficients for Exports of Other Private Services1
RXPOPiDummies
CountryEstimation Period2ConstantFOPiLong-run elasticity3Mean lag*TrendSeasonal
Other
R2D-WSEE
Austria1968–76–1.81*

(0.85)
0.83

(0.59)
–4.62* [8]

(0.95)
3.05*

(1.35)
0.035

(0.025)
–0.12

(0.10)
0.9711.820.048
Belgium51968–760.09

(0.54)
0.51

(0.38)
–0.38 [2]

(0.45)
1.16

(3.86)
0.012

(0.017)
0.9371.590.048
Canada51968–760.99

(0.51)
2.52*

(0.68)
–0.054*

(0.023)
0.17*6(0.03)0.9321.400.044
Denmark1968–76–1.80*

(0.30)
1.00–3.42* [4]

(0.98)
2.49*

(0.62)
0.048*

(0.013)
–0.18*

(0.05)
–0.36*7(0.07)0.8792.290.099
France1968–760.85

(0.63)
1.37*

(0.57)
–0.011

(0.025)
–0.04

(0.03)
0.14*8(0.07)–0.22*9(0.08)0.9621.820.073
Germany, Fed. Rep.1968–760.48*

(0.16)
1.00–0.48 [6]

(0.36)
2.56

(3.60)
0.008

(0.007)
0.07*10(0.03)–0.0311(0.04)0.7552.780.031
Italy1968–760.65

(0.99)
0.82

(0.77)
–4.38* [4]

(1.58)
1.49

(0.86)
–0.012

(0.032)
–0.18*12(0.10)0.31*13(0.09)0.9492.710.097
Japan1968–76–1.37*

(0.35)
1.02

(0.59)
0.036*

(0.013)
–0.06*

(0.03)
0.9451.440.082
Netherlands1968–76–0.13

(1.17)
0.96

(0.78)
–1.12 [2]

(0.88)
1.46

(1.19)
0.018

(0.042)
–0.06*

(0.03)
0.9342.700.073
Norway1968–76–1.86

(0.98)
1.29

(0.74)
0.030

(0.03)
0.03

(0.029)
0.35*14(0.08)0.9442.190.0%
Sweden1968–760.11

(1.13)
1.18

(0.79)
–0.031

(0.031)
1.53*15(0.07)0.9821.620.078
Switzerland1968–76–0.77*

(0.26)
0.41*

(0.18)
–1.93* [8]

(0.14)
3.55*

(0.49)
0.030*

(0.008)
0.9891.810.013
United Kingdom1968–760.91*

(0.15)
0.35

(0.18)
–0.55* [6]

(0.23)
0.42

(1.83)
0.007

(0.006)
0.04*

(0.01)
0.9982.640.007
United States1968–760.77*

(0.15)
1.00–0.70 [6]

(0.59)
0.75

(0.36)
–0.016*

(0.006)
–0.13*16(0.05)0.6512.040.035

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of 1968 to the second half of 1976.

The length of the lag is indicated in square brackets.

The mean lag is in half-year units.

For Belgium and Canada, “Total Other Services” were estimated.

Dock strikes in the United States, first half and second half 1969.

Shift dummy, first half 1968 to second half 1970, prior to change in statistical recording after first half 1971.

Shift dummy, first half 1973 to second half 1976, owing to change in recording methods.

Crisis in Franco-Algerian relations, first half 1971.

Export boom, second half 1972 to first half 1973.

Shift dummy, first half 1974 to second half 1976 because of break in definition of data series.

Strikes in Italy, second half 1969.

Two-tier foreign exchange market in Italy, first half and second half 1973.

High export level of ships, first half 1968 to second half 1969, which led to an increase in merchandise insurance credit.

Discrepancy in statistics, first half and second half 1976.

Dummy, first half 1968 to first half 1971, prior to gains in exports of aircraft and other technologically sophisticated exports.

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of 1968 to the second half of 1976.

The length of the lag is indicated in square brackets.

The mean lag is in half-year units.

For Belgium and Canada, “Total Other Services” were estimated.

Dock strikes in the United States, first half and second half 1969.

Shift dummy, first half 1968 to second half 1970, prior to change in statistical recording after first half 1971.

Shift dummy, first half 1973 to second half 1976, owing to change in recording methods.

Crisis in Franco-Algerian relations, first half 1971.

Export boom, second half 1972 to first half 1973.

Shift dummy, first half 1974 to second half 1976 because of break in definition of data series.

Strikes in Italy, second half 1969.

Two-tier foreign exchange market in Italy, first half and second half 1973.

High export level of ships, first half 1968 to second half 1969, which led to an increase in merchandise insurance credit.

Discrepancy in statistics, first half and second half 1976.

Dummy, first half 1968 to first half 1971, prior to gains in exports of aircraft and other technologically sophisticated exports.

The responsiveness of other private services to changes in a country’s price relative to its competitors’ price (RXPOPi) is important for only nine countries. The estimated coefficients for the other five countries were positive and were therefore dropped from the equations. Only five countries have estimated competitors’ price elasticities that are significant, and the magnitudes of the coefficients for four countries (Austria, Denmark, Italy, and Switzerland) indicate that the country in question exports to markets that are highly price sensitive to changes in its export prices relative to those of its competitors. The U. K. economy is now fairly dependent on earnings from the City for financial and insurance services, and the size of the competitors’ price coefficient for this country (–0.55) suggests that this country exports to markets that are not particularly sensitive to changes in U. K. export prices relative to those of its competitors. This situation reflects the lack of effective competition in this area.

The dock strikes in the United States led to an increase in other service exports in Canada, mainly because of the increase in exports of merchandise insurance; the dummy variable included in the equation to capture this effect has a coefficient of 0.17. A dummy variable was also included in the equation for the Federal Republic of Germany because the export boom (1972–73) led to increased earnings in exports of merchandise insurance.

Workers’ earnings and remittances

The estimated parameters for this category of invisibles are contained in Table 14. Equation (31) for outflows for France and the Federal Republic of Germany and equation (32) for inflows for Italy were estimated. Dummy variables were included to capture the effects of immigration controls and other disturbances. The results show that the number of foreign workers, actual output relative to potential output in manufacturing, and the index of unit labor costs in manufacturing are all important variables in the determination of outflows of workers’ earnings and remittances in France and the Federal Republic of Germany. The estimated parameter on the number of foreign workers is particularly robust for the latter country, and the size of the coefficient (1.27) is not significantly different from one. The estimate is much more precise for that country than for France, where the size of the coefficient is also not significantly different from one; thus, we can hypothesize that for both countries the proportion of foreign workers’ earnings remitted does not change with a change in immigrant manpower.

Table 14.France, the Federal Republic of Germany, and Italy: Regression Coefficients for Inflows and Outflows of Workers’ EArnings and Remittances1
Dependent

Variable
Estimation

Period2
ConstantDummiesR2D-WSEE
SeasonalOther
Outflows of Workers’ Earnings and Remittances
FLiHHWiAHWi
France1968–76–9.78

(6.30)
1.35

(0.90)
3.14*

(0.80)
1.55*

(0.34)
–0.16*

(0.06)
0.9281.550.124
Germany, Fed. Rep.1968–76–9.23*1.27*–0.310.83*–0.12*0.9911.980.069
(1.14)(0.15)(0.73)(0.15)(0.03)
Inflows of Workers’ Earnings and Remittances
FFLiFHWiAFWiBMi
Italy1968–76–4.43*

(1.49)
0.56*

(0.19)
0.37

(0.65)
0.25

(0.19)
2.64*

(0.41)
–0.23*

(0.02)
–0.29*3

(0.06)
0.9531.540.048

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of 1968 to the second half of 1976.

Postponement of remittances, connected with repatriation of savings of emigrants returning permanently to Italy, associated with sharp depreciation of the lira, second half 1974.

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of 1968 to the second half of 1976.

Postponement of remittances, connected with repatriation of savings of emigrants returning permanently to Italy, associated with sharp depreciation of the lira, second half 1974.

Table 15.Fourteen Industrial Countries: Regression Coefficients for Inflows of Private Transfers1
CountryEstimation

Period2
ConstantWGNPTrendDummiesR2D-WSEE
Seasonal
Other
Austria1970–76–9.87*

(1.93)
1.45*

(0.53)
0.019

(0.033)
–0.08*

(0.03)
0.9861.780.062
Belgium1970–76–12.78*

(1.39)
2.84*

(0.38)
–0.128*

(0.024)
–0.07*

(0.02)
0.9662.620.044
Canada1970–76–11.04*

(0.59)
2.51*

(0.11)
–0.095–0.13*

(0.05)
0.9801.370.103
Denmark1970–76–24.87*

(5.16)
5.10*

(1.42)
–0.220*

(0.089)
0.28*

(0.10)
0.59*3

(0.13)
0.9241.830.167
France1970–76–24.64*

(4.25)
5.95*

(1.17)
–0.286

(0.172)
–0.13

(0.07)
0.67*4

(0.14)
0.42*5

(0.11)
0.9561.380.129
Germany, Fed. Rep.1970–76–13.19*

(1.96)
2.85*

(0.54)
–0.122*

(0.034)
–0.21*6

(0.05)
0.9572.630.062
Italy1970–76–3.21*

(1.31)
0.79*

(0.36)
–0.047*

(0.022)
0.10*

(0.02)
0.14*7

(0.03)
0.8581.440.042
Japan1966–76–6.85*

(1.69)
0.88*

(0.43)
0.008*

(0.002)
–0.10*

(0.04)
–0.148

(0.10)
0.9571.430.093
Netherlands1970–76–14.86*

(0.87)
2.92*

(0.16)
–0.0950.33*9

(0.11)
0.9681.240.144
Norway1970–76–8.47*

(2.21)
1.32*

(0.61)
–0.037

(0.038)
–0.12*

(0.04)
0.9181.270.071
Sweden1970–76–12.36*

(1.82)
2.04*

(0.35)
–0.0950.43*10

(0.19)
0.9191.840.224
Switzerland1970–76–6.26*

(1.97)
0.59

(0.54)
0.027

(0.034)
–0.07*

(0.03)
0.9601.810.064
United Kingdom1970–76–16.77*

(2.06)
4.13*

(0.57)
–0.205*

(0.035)
–0.16*11

(0.05)
0.9701.390.047
United States1970–76–9.01*

(2.47)
1.84*

(0.68)
–0.052

(0.042)
0.08*

(0.04)
0.9331.990.080

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the second half of 1976.

Leads and lags, first half and second half 1972.

Leads and lags, first half 1972.

Leads and lags, first half 1973 to second half 1974.

Fall in inflows after revaluation of the deutsche mark, second half 1972 to first half 1973.

Exchange rate depreciation of the lira in 1974 led to postponement of transfers to first half and second half 1975.

Lifting of restriction, first half 1972.

Rise in inflows prior to revaluation of the guilder, first half and second half 1971.

Statistical errors, first half 1975 to second half 1976.

Controls to restrict extent to which U.K. banks can convert foreign currency deposits into sterling, first half 1971 to first half 1972.

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the second half of 1976.

Leads and lags, first half and second half 1972.

Leads and lags, first half 1972.

Leads and lags, first half 1973 to second half 1974.

Fall in inflows after revaluation of the deutsche mark, second half 1972 to first half 1973.

Exchange rate depreciation of the lira in 1974 led to postponement of transfers to first half and second half 1975.

Lifting of restriction, first half 1972.

Rise in inflows prior to revaluation of the guilder, first half and second half 1971.

Statistical errors, first half 1975 to second half 1976.

Controls to restrict extent to which U.K. banks can convert foreign currency deposits into sterling, first half 1971 to first half 1972.

Table 16.Thirteen Industrial Countries: Regression Coefficients for Outflows of Private Transfers1
CountryEstimation

Period2
ConstantGNPiTrendDummiesR2D-WSEE
Seasonal
Other
Austria1970–76–7.01*2.17*–0.0510.9872.180.081
(0.25)(0.30)(0.028)
Belgium1970–76–4.78*0.70*0.0340.9881.850.047
(0.30)(0.21)(0.018)
Canada1960–76–2.92*0.47*–0.005–0.09*0.9561.810.044
(0.31)(0.09)(0.004)(0.01)
Denmark1970–76–8.05*1.82*–0.0390.8712.480.250
(0.67)(0.20)
France1970–76–9.29*1.67*–0.0290.77*30.72*40.9821.670.091
(1.43)(0.43)(0.033)(0.05)(0.10)
Germany, Fed. Rep.1970–76–4.73*0.98*–0.039–0.37*50.9461.420.097
(0.47)(0.08)(0.05)
Japan61966–76–6.63*0.88*–0.0040.63*70.9811.700.134
(0.87)(0.26)(0.023)(0.11)
Netherlands1970–76–5.84*1.13*0.0080.9941.410.041
(0.29)(0.18)(0.162)
Norway1970–76–7.00*1.86*–0.0510.9842.020.070
(0.22)(0.44)(0.040)
Sweden1970–76–9.63*2.19*–0.039–0.2480.9371.300.172
(0.66)(0.17)(0.13)
Switzerland1970–76–4.05*1.91*–0.127*–0.100.8632.020.109
(0.52)(0.41)(0.038)(0.06)
United Kingdom91970–76–7.17*1.73*–0.070*0.11*–0.16*100.7572.140.064
(1.87)(0.41)(0.027)(0.03)(0.06)
United States1970–76–14.11*2.42*–0.100*–0.030.20*110.8631.460.048
(6.60)(1.11)(0.050)(0.03)(0.04)

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the second half of the last year shown.

Tax arrears paid by French oil companies to the Algerian Government, first half 1971.

Leads and lags, first half 1972 to second half 1974.

Outflows before revaluation of the deutsche mark, first half and second half 1972.

Corrected for first-order autocorrelation. The first-order autoregressive coefficient was –0.836 (0.137).

Lifting of restrictions on remittances abroad, first half 1972.

Statistical errors, first half 1975 to second half 1976.

Corrected for first-order autocorrelation. The first-order autoregressive coefficient was 0.749 (0.101).

Controls to restrict extent to which U.K. banks can convert foreign currency deposits into sterling, first half 1971 to first half 1972.

Oil company transactions, first half and second half 1973.

Standard errors are shown in parentheses. All variables except the dummies are in logarithmic form. SEE is the standard error of the estimate. An asterisk indicates that an estimate is significantly different from zero at the 95 per cent confidence level.

The period covers the first half of the first year indicated to the second half of the last year shown.

Tax arrears paid by French oil companies to the Algerian Government, first half 1971.

Leads and lags, first half 1972 to second half 1974.

Outflows before revaluation of the deutsche mark, first half and second half 1972.

Corrected for first-order autocorrelation. The first-order autoregressive coefficient was –0.836 (0.137).

Lifting of restrictions on remittances abroad, first half 1972.

Statistical errors, first half 1975 to second half 1976.

Corrected for first-order autocorrelation. The first-order autoregressive coefficient was 0.749 (0.101).

Controls to restrict extent to which U.K. banks can convert foreign currency deposits into sterling, first half 1971 to first half 1972.

Oil company transactions, first half and second half 1973.

Immigrant manpower in France is concentrated in the manufacturing industries and, in particular, in automobile manufacture. On the whole, this foreign manpower is likely to be unskilled, and therefore remittances of income to the home country will be very sensitive to cyclical effects in manufacturing; this circumstance is reflected in the high coefficient (3.14) on the cyclical variable (HHWi) in the equation for France. This coefficient is negative for the Federal Republic of Germany (–0.31) and is insignificant at the 95 per cent level. The difference between the sign on the coefficient on the cyclical variable for these two countries can be explained by the different treatment of foreign workers over 1973–76, a period of slow growth. In France, the number of foreign workers continued to grow between 1973 and 1976 with a slight reduction in the second half of 1976. Many of these foreign workers were unemployed over this period, so that workers’ earnings and remittances fell. This effect has been captured by the cyclical variable. On the other hand, in the Federal Republic of Germany, the general recruitment freeze in 1973 for workers from countries that were not members of the European Economic Community and the high rate of unemployed foreign workers has led to a high proportion of immigrant workers returning home, so that the cyclical variable is insignificant. 44

The coefficient on the index of unit labor costs is larger for France than for the Federal Republic of Germany; the difference in size is thought to reflect the fact that the French authorities have allowed foreign workers to bring their immediate families with them, so that immigrants’ remittances are likely to be sent to other family members; this means that workers’ earnings repatriated are likely to be more responsive to a change in wages than they are for workers who send money home to members of their immediate families.

In addition to containing measurement errors, the equation for income inflows for Italy is subject to errors because of omitted variables, since only France and the Federal Republic of Germany could be included to calculate the number of domestic workers in the labor force of foreign countries (FFLi), the foreign actual output relative to potential output (FHWi), and the foreign index of unit labor costs (AFWi); data on the size of the foreign labor force was not available for the remainder of the countries to which Italy supplies workers. Measurement errors are therefore likely to be more severe in the inflow equation for Italy than in the outflow equations for France and the Federal Republic of Germany, and this may have led to downward bias of the coefficients in the inflow equation for Italy. Therefore, the estimated coefficient of 0.56 on the number of domestic workers in the foreign labor force is probably too low. Similarly, the coefficients on the foreign index of unit labor costs and the cyclical variable are probably too low.

During a period of expected depreciation (appreciation) of a currency, inflows will be delayed (brought forward) until it is expected that the depreciation (appreciation) has ended or will be reversed. At this point inflows will be increased (decreased). These speculative effects on inflows and outflows of worker’s earnings and remittances are important for all three countries, but they seem to be particularly important for Italy. For this reason, an additional variable—the ratio of the official lira/dollar rate and the black market lira/dollar rate (BMi)—was included in the equation. We hypothesize that the more the official rate for the lira exceeds the black market rate, the more likely it is that a depreciation is anticipated and that inflows will be delayed. The inclusion of this variable in the equation for Italy has led to an estimated coefficient of 2.64, which confirms the hypothesis put forward here.

For the majority of the 14 industrial countries, workers’ earnings and remittances are small or cannot be identified, so that they are analyzed with the other private services or the private transfers item. Income inflows for France and the Federal Republic of Germany and income outflows for Italy are also small, and these have been included with other private services credits (inflows) and debits (outflows) and private transfer credits and debits for these countries.

private transfers

The estimated values of the parameters for equations (33) and (34) are shown in Tables 15 (inflows) and 16 (outflows). Outflows of private transfers are negligible for Italy, and consequently an equation for this item was not estimated.

The inflow equations for private transfers are characterized by rather large standard errors of estimate for some countries. Since private transfers are probably the most volatile category of invisibles, both in terms of gaps in the coverage and in terms of errors in the data estimates, large standard errors of estimate are to be expected. Another explanation for these large standard errors is the simplistic nature of the model. In particular, the inadequate modeling of speculative behavior is a basic shortcoming. Measurement errors also obstructed efforts to identify short-term and long-term influences for some countries, and the coefficient on the trend term was constrained to equal –0.095 for three countries (Canada, the Netherlands, and Sweden) where severe multicollinearity problems were present. This is the average value of the coefficient on the trend term for the other 11 countries.

For the majority of countries, the dominant explanatory variable in the inflow equation is the world income variable (WGNP). The size of the estimated coefficient on this variable is quite large for most countries. For two countries, however, Denmark and France, large positive coefficients on the world income variable are accompanied by large negative coefficients on the trend term; in these cases, caution should be used when interpreting these coefficients. The trend term coefficients are significant and quite large for Belgium, Denmark, France, the Federal Republic of Germany, Italy, Japan, and the United Kingdom. All but one of these countries (Japan) have negative coefficients on the trend term.

In the equations for outflows of private transfers (Table 16), the coefficients on the income (GNP) variable are significant at the 95 per cent level for all 13 countries for which the equations are estimated. The values of the coefficients range from 0.47 (Canada) to 2.42 (the United States). The average value of the income coefficients is 1.53; on average, therefore, outflows of private transfers tend to be income elastic. Long-run factors appear to have led to a decline in outflows of private transfers for Switzerland, the United Kingdom, and the United States. These long-run factors can perhaps be explained by long-run changes in immigrant manpower in the country in question.

Speculation and restrictions are also important determinants of inflows and outflows of the private transfer accounts for many countries; where possible, these have been included as dummy variables documented in the footnotes to Tables 15 and 16.

III. Conclusion

The object of this paper has been to provide a structural model of world trade in invisibles. In constructing and estimating the model, three major difficulties were experienced. First, the complexity of the relationships that determine the many different kinds of flows of invisibles made it difficult to incorporate invisibles easily into the framework of international trade theory; one reason for this is that traditional international trade theory assumes that factors of production are relatively immobile, whereas five of the main categories of trade in invisibles—travel and passenger transportation, freight transportation, investment income, other private services, and workers’ earnings and remittances—exist only because of the international movement of factors of production. Second, the lack of sufficient statistical data on flows of invisibles inevitably involves making a trade-off between availability of data and ideal behavioral relationships. This has necessitated the use of a relatively small number of equations to explain the many kinds of flows of invisibles with their widely different characteristics. Third, the poor quality of the data on receipts from and payments for invisibles is such that the conclusions of any empirical study have to be treated with some caution. From this point of view, the present study is no exception.

The attempt to identify precisely the effects of income and price factors and to separate them from other long-run factors has not proved easy. Strong interrelationships between the demand and trend variables for some countries has made it difficult to disentangle their separate effects on the dependent variable, and in some cases, use has been made of extraneous estimates to constrain the trend variable. Despite these problems, however, the explanatory power of the model is quite high. An indication of the efficiency of the aggregate model is provided by the root-mean-square-error (RMSE) statistic, presented in Table 17, both for each category of invisibles and for aggregate invisibles for each of the 14 countries.45 The largest errors among categories of invisibles appear in freight transportation debits for France and Japan, travel and passenger transportation debits for the Federal Republic of Germany and the United States, and travel and passenger transportation credits for Sweden. The errors tend on average to be larger for those equations for which certain coefficients were constrained.46

Table 17.Fourteen Industrial Countries: Root-Mean-Square-Error Statistics for Exports, Imports, and Balances on Invisibles(In millions of U.S. dollars)
Freight TransportationTravel and Passenger

Transportation
Investment IncomeOther Private Services1Private TransfersTotal Invisibles
CountryCreditsDebitsBalanceCreditsDebitsBalanceCreditsDebitsBalanceCreditsDebitsBalanceCreditsDebitsBalanceBalance
Austria0.0870.1110.0600.4000.4180.5320.1740.0140.1910.2410.1140.3540.6121.619
Belgium0.3480.2930.2810.1170.1670.1190.6011.4931.5990.8130.8110.1710.8232.637
Canada0.4140.5830.6850.9631.5322.6950.0620.1050.2170.4221.2101.8651.5500.2044.2999.761
Denmark0.7230.3720.5510.0620.0840.1530.1280.8530.1140.6110.0190.2300.3321.775
France7.15222.48913.4512.5951.5351.6341.1965.2405.43213.0%1.7533.5696.73336.110
Germany, Fed. Rep.3.1754.1041.3862.05914.92510.16714.2441.9427.47713.7780.2425.0815.06744.642
Italy0.99111.7439.8992.1830.7163.4883.2383.7641.7373.1860.6880.68820.499
Japan12.021129.286131.7760.1110.5470.7391.7191.5290.5692.3140.0630.2390.314136.862
Netherlands3.5271.5171.9590.3350.8701.5755.9073.5631.9441.7580.4430.1260.66911.868
Norway10.4051.7927.3740.0230.2230.1910.2420.4040.2190.3990.0210.0170.0158.221
Sweden0.7873.6561.82815.3080.08715.2410.1502.6458.7263.6210.0190.7540.59921.439
Switzerland0.0040.0150.0230.2470.0410.196%9.8580.0320.0080.0500.0282.8262.90813.035
United Kingdom9.6455.5315.0702.6300.5073.6120.9210.5290.9410.0711.0391.7331.3583.1955.99817.354
United States1.4902.2982.5901.84317.47122.7626.3452.4207.4430.5670.2571.0701.5264.1734.86138.726

Workers’ earnings and remittances are included with “Other Private Services.”

Workers’ earnings and remittances are included with “Other Private Services.”

The empirical results obtained in this study have also led to a number of important findings. In the first place, price competitiveness is an important factor in the determination of various flows of invisibles. In particular, flows of international travel and passenger transportation and other private services seem to be strongly influenced by relative prices and exchange rates. The estimated average price elasticity for expenditure on foreign travel is calculated as –2.30 with a mean lag of 1.3 half years; this can be compared with the estimated average elasticity for market shares for this item of –1.64 with an average mean lag of 1.8 half years. Other private services are also, on average, price elastic; imports have an estimated average elasticity of –2.72 (mean lag 2.3), while the average market share elasticity is –1.95 (mean lag 1.9). The role played by transportation prices in foreign travel and passenger transportation imports is also pronounced for many countries; on average, these prices have a low elasticity (–0.41). The transportation price is an important determinant of the market shares of travel and passenger transportation only for Canada and the United States. Second, some flows of invisibles seem to be highly sensitive to variations in income. Flows of international travel and passenger transportation are strongly influenced by variations in permanent income; an average permanent income elasticity of 1.45 is calculated for the 14 industrial countries. The evidence also indicated that imports of “Other private services” respond to changes in current income; the average income elasticity for this group is 1.32. Third, the results also point to other important variables that influence various categories of invisibles. Flows of freight transportation are nonhomogeneous with respect to those of merchandise; average estimates on the merchandise trade variable of 0.62 with respect to imports of freight transportation and of 0.65 with respect to exports of freight transportation were calculated. For Canada, the United Kingdom, and the United States, outflows of direct investment income tend to be homogeneous with respect to liability levels in the long run (average elasticity, 1.11), and outflows of financial investment income are homogeneous with respect to earnings (average elasticity, 0.96). Most other groups of investment income inflows and outflows are nonhomogeneous with respect to asset or liability levels or earnings.

There remains a considerable amount of work to be done to derive a set of estimates that are completely satisfactory and that allow us to give precise support to the results obtained. Much progress depends on overcoming the present data limitations. A higher level of disaggregated data on receipts from and payments for invisibles would enable the complex relationships to be modeled more adequately, and a wider country coverage of the data would lead to a true worldwide model. Further work on price equations might also prove fruitful for countries whose imports or exports are a large proportion of world supply for a particular service item. For such countries, the present model may contain specification errors leading to bias in the estimated results. At present, lack of sufficient price data for services presents severe limitations to estimating price equations. Improved price data for services will also reduce problems associated with dividing value data by a price index subject to measurement error, which produces a quantity variable also subject to measurement error.

APPENDIX: The Data—General Information

This Appendix presents the general information on data compilation, use of proxy variables, and the arrangement of individual country data into the standard framework of the model as shown in Section I.

freight transportation services

Passenger transportation was separated from the other categories of transportation for all countries, excluding Austria and the Netherlands for which sufficient data were not available. Passenger transportation was included in the equations for international travel for all countries but Austria and the Netherlands, for which it was included in the equations for freight transportation.

The import price of freight transportation (MPFRi) was derived by taking a weighted average of surface freight rates and shipping freight rates.47 Surface freight rates were assumed to be related to the consumer price index and were calculated as a weighted average of the consumer price index of countries through which country i’s freight travels; shipping freight rates were derived by taking a weighted average of the freight rate index for the three basic components of the shipping market (liners, tankers, and tramps) where the freight rate for each market is set by world markets. The weights for each country were determined by the share of commodity group k in country i’s imports where k refers to three basic commodity groups—(i) raw materials and primary products, (ii) finished goods, and (iii) petroleum. These three groups are transported by tramps, liners, and tankers, respectively.48 The export price of freight transportation (XPFRi) was derived in a way that is similar to the import price except that the weights were determined by the share of commodity group k in country i’s exports. For countries where third-country trade is an important component of shipping receipts, the weights included third-country exports.

travel and passenger transportation services

Passenger transportation data cannot be identified for Austria and the Netherlands, and this component was omitted from the equations for these two countries.

Data on travel and passenger transportation receipts and payments were deflated by the travel price of the home country and the passenger transportation price (receipts) or a weighted average of travel price for the 14 industrial countries and the passenger transportation price (payments). The consumer price index, adjusted for exchange rate effects, was used as a proxy for the travel price.49 The average cost of trans-Atlantic transportation in U.S. dollars was used to approximate the world index of average price of transportation services. A weighted average of travel prices for the 14 industrial countries was used for the foreign part of the relative price term.

flows of investment income

Only 3 of the 14 industrial countries (Canada, the United Kingdom, and the United States) provided adequate data on their international asset and liability positions, and for these 3 countries the gross flows approach was used. For the remaining 11 countries, net flows of investment income were analyzed using a net flows approach.

Gross flows approach

The “Other investment” category varies slightly from the general definition given earlier for Canada and the United Kingdom. For Canada, “Other investment” includes miscellaneous investment, while for the United Kingdom it includes oil credits and debits, local authority and public corporations’ borrowing and lending, and earnings and payments of banks. Direct investment income data for Canada and the United States exclude reinvested earnings,50 but for the United Kingdom they include reinvested earnings. Therefore, equations (18) and (22) were estimated for the United Kingdom without the variable that determines the proportion of earnings reinvested.

The long-run (RPDI) and cyclical (CYDI) rates of return on direct investment assets and liabilities were proxied by the (long-run) government bond yield (GBY) and the (cyclical) index of the ratio of actual output to potential output in manufacturing (QM/QMT). 51 The relative profitability of domestic and foreign investment (PRi) is the ratio of output in manufacturing to potential at home (QMi/QMTi) relative to this ratio abroad. For the long-run and short-run rates of return on financial and other investment assets and liabilities, the (long-run) government bond yield and the (short-run) discount (DR) were used. The countries used to define the weighted foreign rates of return were the same 14 industrial countries that appear in this study. Where part of the foreign asset type k was denominated in the own country’s currency, own country rates of return were used for the proportion denominated in own currency.

Semiannual data on stocks of assets and liabilities for Canada, the United Kingdom, and the United States do not exist. Stock figures were calculated by adding two quarterly investment flow figures to a base-year stock figure and using information from revaluation and depreciation figures.

Net flows approach

The lack of data on asset and liability levels for the remaining 11 countries means that we have no data for the first observation of NASSit; this figure was calculated by accumulating the current account balance from the year 195052to the beginning of the estimation period that is documented in Table 11. The remainder of the series, NASSit, was calculated by adding the semiannual current account balance to the figure for the previous period.

The question of the appropriate rate of return variables is a difficult one. The rate of return corresponding to the currency denomination of each asset or liability type would be the correct rate of return variable, but these variables were unobtainable for two reasons, (i) Information on the proportion of direct, financial, and other assets and liabilities was not available, so that the appropriate rates of return and their weights could not be determined, (ii) Information on the size of the assets or liabilities denominated in the various currencies was unavailable, so that only one rate of return variable could be used. Since more assets and liabilities are denominated in U.S. dollars than in any other currency, the U.S. interest rate (denominated in dollars) was chosen to represent the rate of return variable. Two long-run rates—government bond yield (GBY) and commercial bank rate (CBR)—and two short-run rates—discount rate (DR) and the Treasury bill rate (TBR)—were used in the empirical estimation; the results (in terms of R2 and standard errors) were allowed to determine the appropriate rate.

other private services

For particular countries, “Other private services” contain items not included in the general definition in Section I, Other services. Norway includes entries for earnings by foreign crews on Norwegian ships under “Other private services,” and Denmark includes some entries for merchandise transactions abroad under this item. Workers’ earnings are also included with “Other private services” for all countries, excluding outflows for France and the Federal Republic of Germany and inflows for Italy. For Belgium and Canada, “Other private services” could not be separated from “Miscellaneous government transactions,” so that “Total other services” were estimated.

Data on receipts from and payments for other private services were deflated by the other private service price of the domestic country (receipts) or a weighted average of the other private service price for the 14 industrial countries (payments). The GNP deflator, adjusted for exchange rate effects, was used as a proxy for the other private service price; a weighted average of that price for the 14 industrial countries was used for the foreign part of the relative price term.

workers’ earnings and remittances

Equations for workers’ earnings and remittances were estimated for countries for which flows of workers’ earnings and remittances are important and for which balance of payments data are available; these are France and the Federal Republic of Germany (outflows) and Italy (inflows).

The average hourly wage rate in country i (AHWi) was proxied by an index of actual unit labor costs in manufacturing in country i, and the number of hours worked in country i (HHWi) was proxied by the actual output relative to potential output in manufacturing of country i (QMi/QMTi).

private transfers

Bilateral data on private transfers for most of the 14 countries were difficult to obtain, so that it was not possible to compute the weights for the foreign GNP variable (FGNPi) in the inflow equation. Instead, a world income variable (WGNP), which is the average GNP for all 14 countries in the study, was used. Outflows of private transfers were negligible for Italy, so that no equation was estimated.

BIBLIOGRAPHY

Ms. Bond, economist in the External Adjustment Division of the Research Department, is a graduate of the University of Essex and of the London School of Economics and Political Science. She was formerly a member of the faculty of the University of Reading, England.

The author is indebted to Lawrence R. Klein, H. Peter Gray, George Yan-nopoulos, and colleagues in the Fund for valuable advice and comments. Any shortcomings remain the responsibility of the author.

In 1971, world receipts from invisibles amounted to $135 billion, compared with world merchandise receipts of $307 billion.

A study of aggregate flows of invisibles was made by Driehuis (1969). Studies of disaggregate flows have been made by Amano (1973), Duffy and Renton (1970), Kwack (1974), Phillips (1974), Prachowny (1969), and Rhomberg and Boissonneault (1964). Individual studies of particular groups of invisibles include those by Artus (1972), Bond (1978), Gray (1966), and Kwack (1972 b) for travel; Bond (1977), Hemphill (1977), and Kwack (1972 a) for investment income; and Kwack (1971) for freight transportation. A good survey article of country-based models of invisibles can be found in Sawyer (1973).

See Section I, Workers’ earnings and remittances, for an explanation of the difference between earnings and remittances.

For a justification of the disaggregation of total investment income flows into their component parts, see Bond (1977).

Availability of an appropriate price deflator determined whether the service item was estimated in volume or value terms.

Export supply functions for services could have been estimated in the form of export price equations where the export price of services was a function of prices charged in the service industry plus the volume of exports of services relative to domestic potential output of the service. The lack of relevant data, however, renders such an approach impractical at this stage.

In many cases, total expenditure (i.e., expenditure on both the domestic and the foreign service item) will not be available, so that a proxy variable will be used instead.

A more detailed account of the derivation of these two approaches for travel and passenger transportation is given in Bond (1978).

The exact-adding-up properties, that total imports equal total exports, are still preserved by adding a trend term to the market shares equation. (See Hickman and Lau (1973).) Strictly speaking, however, the parameters on the trend term should be constrained, since they are not independent because changes in tastes (or other long-run factors) in favor of one country’s service exports should be offset by a change in tastes against another country’s service exports.

The unrestricted bilateral import function is specified at the first stage of the maximization procedure, where imports of the service item of country j (MVSj) are endogenous. This differs from the bilateral import function in the market shares approach, which is specified at the second stage, where imports of the service item of country j (MVSj) are treated as exogenous.

Workers’ remittances have been analyzed together with workers’ earnings under services. Although on a balance of payments basis this item should be included with private transfers, on an analytical basis it belongs with workers’ earnings.

The freight transportation account excludes purely domestic transactions, that is, freight payments from a country’s importers to its own suppliers of freight transportation services. Nevertheless, these domestic transactions, as well as third-country trade, lead to port service receipts and payments.

Some countries, such as the United States, are exceptions, and insurance is not included with the transportation account because data are not available separately from other international insurance transactions; some other countries include part of this insurance item in the transportation account and part in the other private service account.

It has been suggested by Gray (1971) that port expenditure be netted against transportation receipts, and port receipts against transportation expenditure; however, this procedure requires more disaggregation of data than is generally available for most of the countries in this study.

Strictly speaking, for these countries supply equations should also be specified.

A more detailed treatment of travel and passenger transportation flows can be found in Bond (1978). There (i) two alternative approaches to the specification of export equations are considered, (ii) bilateral travel flow equations are estimated for the United States and compared with an aggregate model of travel flows for the United States, and (iii) the price term is defined precisely.

No attempt has been made to develop a “choice of journey destination model” prior to explaining the factors determining travel receipts and expenditure. We are therefore assuming that for the 14 major countries destinational characteristics are similar.

The price of transportation within country i is included with PTVi, and similarly the price of transportation within country j is included with PTVj.

The price of transportation variable in the market share export equation will be important only to the extent that a change in the price of transportation affects a country’s market shares.

For a more detailed account of the determinants of investment income flows, see Bond (1977).

Averages are calculated by use of the formula 0.5 * (Xt + Xt–1), where X refers to semiannual frequencies.

Some financial and other investment may be denominated in the currency of the country that holds the liability (for example, some Canadian liabilities are denominated in U. S. dollars) rather than in the domestic currency. In this case, the rate of return of the country holding the liability was used for the proportion of financial and other investment liabilities denominated in the currency of that country.

This approach is an adaptation of the approach originated by Hemphill (1977).

This is considered to be an adequate proxy for total expenditure on domestic and foreign other private services.

For the market-shares export equation, the linear form would be preferred, since the identity imports equal exports is exactly satisfied. Since the fits of the relationships are good, however, the lack of consistency was not considered to be serious.

See Hickman and Lau (1973) and Khan and Ross (1975) for possible consequences of the neglect of cyclical and trend factors in the determination of imports and exports.

When the equations for these 6 countries are estimated without the trend term, larger positive coefficients on the trade variable are obtained.

Since freight transportation imports are measured in values, a coefficient of unity on the freight transportation import price is indicative of a correctly measured deflator but no price sensitivity. A coefficient on the price term of less than one will reflect some degree of price sensitivity. Price sensitivity depends upon several factors, including the value per ton of the commodity being transported and price cutting by countries other than the 14 industrial countries. For example, low value/high weight products, such as timber, might not be imported if freight rates became too high, since the freight rate represents a high proportion of the import price.

The criteria chosen to indicate a lack of multicollinearity are low covariance between the estimates and the plausibility of the results.

Austria

XFR =4.21(0.33)0.21(0.32)lnXMVT + 0.57(0.25)lnXPFR + 0.064(0.015)t0.07(0.03)H1R2 = 0.991DW = 2.07SEE = 0.050

France

XFR =0.96(0.65)0.19(0.32)lnXMVT + 0.98(0.26)lnXPFR + 0.040(0.020)t0.36(0.03)SHR2 = 0.992DW = 1.75SEE = 0.062

H1: seasonal dummy

SH: Shift dummy, second half 1971 to first half 1976

The zero end-period constraint was imposed on the equations for Denmark (imports) and for Austria and the Netherlands (exports).

The population variable was included in the empirical estimation, but this made no noticeable difference to the estimated coefficients and was therefore deleted from the equation.

This result conflicts with several previous studies that suggest that the demand for passenger transportation and travel services is fairly elastic with respect to the price of air transportation. See studies by Wheatcroft (1956) and Straszheim (1969).

The break in the data definition for France in 1968 exists because post-1968 the data refer to France and the rest of the world (including the overseas franc area) and pre-1968 they refer to France (including the overseas franc area) and the rest of the world.

The h-statistic is used to test for serial correlation in equations where lagged dependent variables are present; see Durbin (1970). If h exceeds 1.645, the hypothesis of zero autocorrelation in the residuals must be rejected at the 5 per cent level. The test is appropriate only for large samples.

It was thought that these small coefficients might be a result of deflating asset levels by the official exchange rate rather than by the more appropriate investment dollar rate. When asset levels were deflated by the investment dollar rate, however, there was no noticeable improvement in the results.

Lagged interest rates affect current income inflows because the published asset figures do not necessarily reflect changes in asset valuation caused by changes in interest rates.

This will not be applicable for the U. K. direct investment equations that are estimated as earnings, or the U. S. financial and government flows, because the data are computed by the data collecting agencies from the various asset (and liability) components and their assumed rates of return.

There are two possible effects on investment income flows owing to the oil embargo: (i) company profits from current operations might fall or rise, with the effects in the long run not necessarily the same as in the short run; and (ii) capital gains might take place on company-owned stocks of petroleum and petroleum products. Thus, the sign of the dummy variable could be positive or negative.

Where the estimated coefficient for the lagged dependent variable lies between 0 and –1, this implies that adjustment toward the steady-state income outflow tends to overshoot its target in the short run but is stable in the long run.

For the net flows approach, the variables are measured in levels, so that the interpretation of these coefficients is rather different from that for the other coefficients presented in the model.

The zero end-period constraint was imposed on the equations for Norway (imports) and the Federal Republic of Germany (exports).

Kemp (1962) has shown that if the demand function is exact and if the quantity dependent variable in the import demand equation is constructed by dividing value data by a price index subject to measurement error, then the least-squares elasticity estimate will be biased toward minus one.

Until 1973, the unemployment rate was considerably lower for foreigners than for the labor force as a whole, but since 1974 it has risen sharply for foreign labor.

For two categories—travel and passenger transportation, and other private services—constant (1970) U. S. dollar data were converted into current U. S. dollar data by use of the appropriate deflator.

The mean percentage error statistic (MPES) was also calculated for each category of invisible debits and credits. This statistic gives an indication of percentage errors involved and gives less weight to extreme errors than does the RMSE. For the majority of categories of invisibles and the majority of countries, the MPESes range from 1 per cent to 4 per cent. These prediction errors were analyzed for systematic bias by calculating the mean-square error (MSE) and decomposing it into the bias component (the square of the average prediction component) and the variance of the error around the average. For the six categories and 14 countries, the average errors are quite small and are not statistically significant at the 5 per cent level. There is, therefore, virtually no bias in the sample period predictions. (The formulas for decomposing the MSE into its bias and variance components can be found in Maddala (1977), p. 344.)

These freight rates are already a weighted average of the price of cargo transportation and port services.

To a large extent, the type of product to be transported determines the type of transportation used. For oceangoing vessels, grain, coal, iron ore, and special cargo are usually transported in bulk carriers or tramp steamers, finished goods in liners, and petroleum in tankers. For surface haulage, coal, grains, and ores are usually transported by railways, and petroleum and manufactured goods by either railways or road haulage vehicles.

The use of this proxy for a deflator has obvious limitations, as no account is taken of the reduction in costs resulting from the growth of package tours. The 14 countries used to define the weighted foreign price variable are the same 14 countries for which the equations were estimated.

The United States is currently revising the definition of flows of direct investment income to include reinvested earnings.

At the micro level, the assumption is that when the firm is at its long-run equilibrium position, both output and profits are higher than when it is at short-run equilibrium.

The year 1950 was chosen because it eliminates data from World War II and prewar years. Although this may not seem long enough to accumulate asset and liability figures for direct investment, for many countries those assets and liabilities were destroyed during that War.

Other Resources Citing This Publication