Errors occur in the global capital account statistics shown in Part 2 of the Yearbook because certain countries do not report to the Fund and others report with excessive delay. In some instances, missing data are estimated by the Fund while, in others, data are completely missing from the global accounts. This chapter discusses improvements in global recording that would occur if more complete and timely reporting were achieved; it suggests ways in which the Fund estimates in Part 2 of the Yearbook could be improved; and it shows that inclusion of new information for the U.S.S.R. slightly affects the global discrepancy.


Several topics related to the quality and completeness of global capital account statistics are not specifically covered in earlier chapters. This chapter deals with those topics. It begins by describing the basic data for this study of the measurement of international capital flows: the figures published by the Fund’s Statistics Department in the Balance of Payments Statistics Yearbook, Part 2.

The data in Part 2 reflect what has been both measured and reported by national statisticians and estimated by Fund staff. The following section focuses primarily on how the data are assembled and adjusted, and why and how Fund estimates are made. And another section discusses the impact of estimates on the global aggregates shown in Part 2.154

Some data are also completely missing from Part 2. For lack of information, the Fund has been unable to make any capital flow estimates for certain countries. Chapter 9 has already considered the matter of missing data for offshore financial centers. A final section of this chapter describes how the Working Party used new information on the U.S.S.R. to round out the global data.

Management of the Fund’s Balance of Payments Data Base

Publication of the Data

Most members of the IMF, together with a few non-members, submit a variety of domestic and international economic data to the Fund’s Statistics Department for publication. Balance of payments data are published, around November of each year, by the Statistics Department in two volumes of the Yearbook.155 Both volumes (Part 1 and Part 2) cover balance of payments statistics through the end of the preceding year.

Part 1 contains balance of payments statements for individual countries; each statement is ordered in the standard categories described in the Balance of Payments Manual and expressed in U.S. dollars. Data for 137 countries are included in the 1990 issue. Annual data included in this issue cover the eight years 1982 through 1989. Data are arranged in both an aggregated presentation that focuses on certain “analytic” balances and a detailed presentation that shows data for all of the standard current and capital account categories.

Part 2 aggregates the data given in Part 1 by balance of payments category. Each category displays data for countries, country groups, international organizations, and “total.” Part 2 consists of Part 1 data augmented by estimates for nonreporters, estimated updates for late reporters, figures for reporters that do not report in sufficient detail to be included in Part 1, and transactions of international organizations. Part 2, in effect, expands Part 1 data to an almost global scale and ensures that all the years (the seven years of 1983 through 1989 in the 1990 issue) are consistently covered.156 It therefore contains the global balances that concerned the Working Party.

Part 2 begins with three summary tables. Table 1 provides a country-group summary of selected balances on international transactions; Table 2 shows global balances for each of the main current account components; and Table 3 shows global current account balances adjusted for selected statistical discrepancies, in line with the methodology adopted in the Report on the World Current Account Discrepancy.

There follows a set of “A” tables that organize the data into a number of analytic balances which show national, regional, and global contributions to each type of balance. A set of “B” tables follows; these give global and country-group balance of payments statements structured to summarize the contribution of major types of transactions. Finally, there is a set of “C” tables that give country, country-group, and global data for each of the main individual categories of current and capital transactions in the Fund’s standard presentation. The “C” tables contain country data for individual types of transactions. These data were the Working Party’s primary area of concentration and the main focus of its analysis. The “C” tables were the basis for Table 3 in Chapter 1 and other tables in this report.157

Collection and Compilation of Capital Account Data

Designated “balance of payments correspondents” in reporting countries regularly submit time series of their national balance of payments statistics to the Fund’s Statistics Department. The forms in which countries present their capital account data in national publications differ greatly, so a standardized presentation such as that produced by the Statistics Department is required for global compilations. The form used by the Fund follows the classification scheme set forth in the Manual.158

The Statistics Department makes available a standardized report form to assist in the provision of data to the Fund. Not all countries return figures on this standard form. In cases of divergence, the Statistics Department (often in consultation with the countries concerned) rearranges data from the national submissions to conform to the standard presentation. Also, data on Fund transactions—for example, transactions in special drawing rights and in the Fund’s General Resources Account—are obtained from Fund sources rather than from national submissions.

The Statistics Department checks and processes data received from countries. Checking and editing of national submissions are regular procedures followed by the department. When necessary, the Fund confers with national balance of payments correspondents to seek clarification. In a few cases, reported data are amended or supplemented by data from other sources, and figures are entered into the Fund’s computer system. Around September of each year, country-reported data are supplemented by Fund estimates and other figures, and the tables in both Parts 1 and 2 are “frozen” for publication.

Relationship Between Part 1 and Part 2 Data

The country-reported data from Part 1 are carried directly into Part 2. Thus, data for individual countries in Part 2 are identical to figures in Part 1.159 The 1990 issue of Part 2 shows data for 142 countries: the 137 reporting countries represented in Part 1 and 5 more countries (Mozambique, Romania, Kiribati, Guyana, and the Netherlands Antilles) that appeared in earlier issues of Part 1 but that had not reported in recent years. These five countries are not the only “late reporters”; many others send figures but do not provide up-to-date data for inclusion in the Yearbook. Of the 142 countries listed in the 1990 issue, 38—or 27 percent, all in the “developing countries” group—did not provide data for 1989, and some of these also did not provide data for earlier years. When data reports are delayed, estimates made by Fund staff are inserted in Part 2. These estimates are not shown separately; the existence of an estimate is indicated by the entry “…” in the relevant cell of a table. Estimates are added into regional and global totals.

There are other significant instances in which capital flow data that appear in Part 2 do not appear in Part 1. First, the Fund adds entries for international organizations. These data are derived by the Statistics Department primarily from survey results.160 Second, the 1990 issue of Part 2 includes data for Bulgaria, Czechoslovakia, Hungary, Hong Kong, and Taiwan Province of China that are not found in Part 1. Data for the first three are included indistinguishably under the heading “Europe not specified.” Figures for the latter two are included under the heading “Asia not specified.” Third, Part 2 contains estimates made by the Fund for certain countries, some of them Fund members, that do not report data to the Fund. These countries, described here as “nonreporters,” are Cape Verde, Djibouti, Guinea, Cambodia, Viet Nam, Iraq, Lebanon, Qatar, and the United Arab Emirates.161 Fund estimates for these nonreporters are included indistinguishably in the lines for “Africa not specified,” “Asia not specified,” and “Middle East not specified,” as applicable, in the Part 2 tables.

Therefore, the only individual country data shown explicitly in Part 2 are those obtained directly from Part 1. However, because estimates for late reporters are indicated by the entry “…”, the regional and global totals often differ from the sum of the country figures printed in the “C” tables.

Because both reported data and estimates are lacking, for some countries there are no capital account data at all in the Part 2 tabulations. Among the countries without capital account data are Albania, Angola, Bhutan, Brunei, Cuba, the Democratic People’s Republic of Korea, Mongolia, Namibia, the U.S.S.R., and certain offshore financial centers such as Bermuda and the Cayman Islands.

As should be evident from the foregoing, nonreporting and late reporting of data by some countries can contribute to discrepancies in global capital flow totals, quite apart from errors that may reside in the figures submitted by reporting countries.162

Other Fund Data Bases

The principal sources of estimates needed to move from Part 1 to Part 2 are country specialists in area departments of the Fund.163 These specialists, in turn, rely on a variety of sources to maintain their country data bases. Typical sources are (a) national statistical agencies, (b) data obtained during official Fund visits to countries, (c) estimates made on the basis of partial information from countries, and (d) other sources such as the Statistics Department and international organizations. Area department data bases are maintained and processed independently by country specialists. These data bases may be kept in personal computer files, and they are not readily accessible in bulk.

Area department data bases are used in reports on “Recent Economic Developments” (REDs) of individual countries. The specific forms of data found in REDs vary by country, but they frequently conform with “national presentations” or contain only limited detail. REDs were not designed to be integrated with the statistical operations of the Statistics Department. They rarely show balance of payments detail in the standard classifications recommended by the Fund’s Manual. The result is that RED statistics can diverge noticeably from those maintained by the Statistics Department.164

Area department documents and studies are designed to aid policy, especially for countries having or seeking Fund programs. They constitute, in addition, basic raw material needed for publication of the Fund’s semiannual World Economic Outlook (WEO). To produce this global synthesis, national balance of payments and other statistics are transmitted from country experts to WEO staff, who centralize and standardize them in the WEO data base. This data base contains both the historical and projected estimates necessary for a broad assessment of the global outlook. The standardized WEO format contains only a limited number of capital flow categories.165 Because WEO data are in centralized computer files, they are readily accessible to the Statistics Department for estimation of missing Part 2 balance of payments components.

Estimation Procedures

Six categories of capital flows are maintained in the WEO data base, and these are the principal sources of Yearbook estimates when moving from Part 1 to Part 2. When there is no balance of payments report for a particular country covered in Part 2, the Statistics Department maps the six WEO accounts into six Yearbook accounts that are part of its standard capital account format. Table 65 shows the correspondence between WEO and Yearbook entries.

Table 65.

Correspondence between WEO and Statistics Department Categories of Capital Flows

article image
n.i.e. = not including exceptional financing and LCFAR.

Except for short-term assets, Table 65 shows that in each case the net amount carried in WEO files is assigned to the liabilities side of a country’s accounts in the Yearbook. Except for direct investment, exceptional financing, and LCFAR, the sectorization is to “other sectors.” This extraction of WEO figures is supplemented by the mapping of IFS data on “international reserves” (after the deduction of Fund transactions included in reserves) for the standard Yearbook capital account component “foreign exchange reserves.”166

Role of Fund Estimates in the Global Totals

The Statistics Department does not make any independent balance of payments estimates for countries (or years) if estimates are required to complete the “world picture.” The translation from WEO and IFS into Yearbook format is purely mechanical. Three observations are in order about this procedure:

The procedures followed by the Fund for inserting estimates into global capital flows may be somewhat problematic, but it is difficult to envisage a better general approach, given present data availability and the need to make estimates within a very short period of time. Inserting WEO estimates may not impart any bias to the global discrepancy if country specialists from area departments can make relatively good estimates about total capital flows and reserves. The paucity of detail, however, will result in great uncertainty about the “true” contents of individual line items.

The contribution, by value, of Fund estimates to global capital flow balances is relatively small, but not insignificant. This can be seen in Table 66, which presents, for years 1986 through 1989, data that are consistent with those in the 1990 Yearbook. For example, the global capital account balance of $66.4 billion in 1989 includes a net estimated amount of $2.3 billion to account for the missing data of late reporters and –$1.6 billion to account for the missing data of nonreporters. The estimates for 1989 contribute –$10.4 billion to global capital account assets and $11.0 billion to global capital account liabilities (about 1 percent of each total).

Table 66.

Contribution of Country Reports and Fund Estimates to Global Balances, 1986–891

(In billions of U.S. dollars; outflows ( – ))

article image

Data derived from the “C” tables in the Balance of Payments Statistics Yearbook, Part 2, 1990.

Also includes international organizations and actual data for Bulgaria, Czechoslovakia, Hungary, China.

Does not include exceptional financing and LCFAR.

Despite the relatively small net sizes of Fund estimates, revisions made (mainly as actual data become available) to these estimates contribute quite significantly to revisions to global aggregates. This is illustrated in Table 67, which shows the contribution of various types of revisions between the first published and latest published figures for the three reference years 1986 through 1988. The table shows, for example, that the first-published figure (in the 1987 issue of Part 2) for the 1986 global capital account balance was $15.3 billion; that the revised balance (in the 1990 issue) was $18.8 billion; that the revision can be explained by various positive and negative revisions to underlying data; and that the net impact of replacing Fund estimates with reported data or revised Fund estimates is –$6.1 billion. The contributions of revisions to Fund estimates to first-published data for 1987 and 1988 are of a similar magnitude, about $6 billion each.

Table 67.

Total Capital Account Flows and Discrepancies: Effects of Revisions, 1986–88

(In billions of U.S. dollars; outflows ( – ))

article image

Data derived from Table C-13, “Capital Account: Assets and Liabilities” in the 1987–89 issues of the Balance of Payments Statistics Yearbook, Part 2. They include small amounts of exceptional financing in the current account.

As defined in Appendix VIII.

Data are derived from Table C-13 in the Yearbook, Part 2, 1990.

Clearly, problems arise for global capital flow measurement from the lack of timely reporting and the need for the Fund to estimate missing components of these flows. For this reason the Working Party urges countries to compile and submit to the Fund capital account data, and subsequent revisions to these data, on a timely basis, to reduce the need for estimation by the Fund. The most significant nonreporters for which estimates are inserted by the Fund are Hong Kong and certain Middle Eastern countries. Also, a few countries with significant capital flows have been late in their reporting in recent years: Algeria, Cameroon, India, Libya, Oman, Saudi Arabia, and Brazil. The Fund should give high priority to obtaining adequate and timely reports from these countries.

The Hong Kong authorities do not compile a balance of payments statement (see Chapter 9), although data are published for certain international transactions. The Statistics Department used these data to make Part 2 entries through 1986, and for this reason Hong Kong is not classified as a nonreporter in Table 66. (However, since the department did not update its data beyond 1986, WEO-based estimates for Hong Kong are included in the “late reports” columns of Table 66 for 1987 through 1989.) The Hong Kong data make a significant contribution to “Asia not specified” in the Part 2 tables.

Investigations by the Working Party indicate that there is further scope to harmonize the WEO, area department, and Statistics Department data bases and to make some technical improvements to the process by which Fund estimates are incorporated into Part 2. Improvements in Part 2 data can be achieved through greater collaboration among staff in different departments of the Fund and by better articulation of area department and WEO data bases on international transactions. The Statistics Department and other departments of the Fund should aim to achieve a greater harmonization of available balance of payments statistics and should seek to improve estimates for countries that do not report or report late.

Effects of the U.S.S.R.’s Balance of Payments Data on the Global Capital Account Discrepancy

Balance of payments statistics for the former U.S.S.R. have not been reported to the Statistics Department and therefore are not included in the Yearbook, but balance of payments statistics for some U.S.S.R. partner countries are included. This difference in coverage contributes to the asymmetries in the global capital account. During the course of the Working Party’s research, the Fund and other international organizations obtained data on the U.S.S.R.’s external transactions in nonconvertible currencies with socialist countries and with nonsocialist countries as groups. Some of the socialist countries are covered in the Yearbook; among these are Bulgaria, the People’s Republic of China, Czechoslovakia, Hungary, Poland, Romania, and several others. Others are not included—notably Albania, Cuba, the former German Democratic Republic, and North Korea—and therefore they do not contribute to the asymmetry. In addition, data were obtained on the U.S.S.R.’s total external transactions in convertible currencies, which were assumed to be entirely with nonsocialist countries. Using a combination of these data, the Working Party was able to make adjustments to reduce the U.S.S.R.-related asymmetry in existing capital flow figures.

Adjustments to global capital account balances to remove existing asymmetries can be made in one of either two ways: (1) by entering into the data the U.S.S.R.’s transactions with countries for which balance of payments data already are included in the Yearbook; or (2) by excluding from Yearbook data the capital account transactions with the U.S.S.R. of those same countries.

Available data permit a mixture of these adjustments to fill this gap in the global statistics. The Working Party has added to Yearbook data the U.S.S.R.’s transactions in nonconvertible currencies with nonsocialist countries and those in convertible currencies with the rest of the world, and it has excluded from Yearbook data the separately available data on the transactions of Bulgaria, Czechoslovakia, Hungary, Poland, and Romania with the U.S.S.R.167

These capital account adjustments are, naturally, approximate and do not eliminate all sources of asymmetry. They succeed in building a capital flow statement for the U.S.S.R. vis-à-vis all nonsocialist countries and a number of socialist countries, but the coverage remains imbalanced as regards four countries for which data are included in the Yearbook: the People’s Republic of China, the Lao People’s Republic, Viet Nam, and Yugoslavia.

The capital account balance included in the Yearbook for countries covered by U.S.S.R.-related adjustments, the adjustments, and the adjusted capital account total are shown in Table 68. Over the four years the net adjustments average –$2.4 billion and thus somewhat reduce the global capital account imbalance in the Yearbook statistics. Inclusion of balance of payments data for the U.S.S.R. and successor states plainly is a matter of importance to the quality of the global accounts, and the Working Party recommends that these statistics be incorporated into the Yearbook as soon as possible.

Table 68.

Capital Account Balances and Adjustments for U.S.S.R. and Selected Socialist Countries, 1986–89

(In billions of U.S. dollars; outflows ( – ))

article image

Balance of Payments Statistics Yearbook, 1990.

Bulgaria, Czechoslovakia, Hungary, Poland, and Romania.

Conclusions and Recommendations

Errors and misclassifications occur in global capital account recording because certain countries do not report to the Fund and others report with excessive delay. In some instances, estimates of missing data are made by Fund staff while, in others, data are completely absent from the global accounts. Improvements to Fund estimates included in Part 2 of the Yearbook could be achieved by more collaboration between staff in different departments and by better articulation of area department and WEO data bases. Estimates also can be made for economies, such as offshore financial centers and the U.S.S.R., that are completely missing from Part 2. Chapter 9 has illustrated how these estimates can be made for offshore financial centers. This chapter shows how new data for the U.S.S.R. helps to round out the global picture.

The Working Party recommends that:

Statistical Appendices

Appendix I Respondents to Special Questionnaire on International Capital Flows

article image

Appendix II Draft Fifth Edition of the Balance of Payments Manual168

The scope and orientation of the draft fifth edition of the Manual differ from those of the fourth edition in a number of respects. First, the conceptual framework has been expanded, in that the balance of payments accounts, the international investment position, and the external sector accounts of the United Nations’ System of National Accounts are viewed as an integrated set of accounts. The linkages among them have been strengthened and harmonized to the maximum extent practicable. Second, the proposed sectorization of both the investment income and portfolio and other capital accounts should better integrate the Manual with other statistical systems, like money and banking, government finance, international banking, and external debt.

Within the enlarged framework, the capital account section of the Manual will be expanded, and a number of significant changes are proposed in the coverage and classification of capital account components, in the related income components of the current account, and in the international investment position. The following are some of the salient changes proposed in the fifth edition:

Appendix III Survey of National Direct Investment Practices


Chapter 3 of this report makes use of the results of a supplementary survey carried out by the Working Party in early 1991 of a cross section of Fund members and Switzerland. The survey forms contained a list of questions about the concepts, definitions, sources, and methods that countries use for the compilation of the direct investment statistics. The survey’s purpose was to identify differences among countries that contribute to the imbalances in the recording of international capital flows at the world level.

Thirty-eight completed forms were obtained: 31 that were submitted directly by national compilers and 7 others that were completed by the Working Party on the basis of discussions with, or information provided by, national compilers and international organizations. The 38 respondents consisted of 19 from each of the Fund’s “industrial countries” and “developing countries” categories.169

This appendix presents a summary of replies to the survey. It also compares country practices with the international standards applicable at the time the replies were received, that is, as set out in the fourth edition of the Fund’s Balance of Payments Manual and articulated in the OECD’s 1983 “Benchmark Definition.”170 It shows that, notwithstanding international standards, there are widespread variations in national statistical practices.

Survey Results

Percentage Threshold

The Manual defines a direct investor as an entity that intends to have an effective voice in the management of the direct investment enterprise and states that this intention is usually associated with some degree of equity ownership. The “Benchmark Definition” recommends using a minimum ownership share (or equity threshold) of 10 percent as evidence of a direct investment relationship, although it permits upward or downward variations from this threshold in particular cases.

A majority of survey respondents rely upon some minimum equity percentage to identify the direct investment relationship. However, the chosen percentages are diverse. Six respondents noted that they make exceptions to their general minimum-percentage rule when other evidence suggests that a direct investment relationship has or has not been achieved.

Among the industrial countries, eight respondents use the 10 percent threshold. These include the United States, Japan, Canada, Australia, and Sweden. Six use a 20 percent threshold—including France, Germany, and the United Kingdom—and one uses a 25 percent threshold.171 Among the developing countries, only two use the 10 percent threshold; one of them uses this for outward investment only, with a 20 percent threshold for inward investment. Another seven use higher percentage thresholds—mostly on the order of 20 to 25 percent but with one country using 50 percent. One uses a 5 percent threshold.

Two countries—one industrial and one developing—use minimum-value ownership, rather than percentage ownership, as the threshold. One of these uses its value threshold for inward investment only, but with a 20 percent threshold for outward investment.

The other respondents do not rely upon a specific percentage or value criterion; this remainder includes three industrial country respondents and nine developing country respondents. Some of these countries, including Belgium-Luxembourg and the Netherlands, prefer to treat each case on its merits, while other countries rely upon decisions of national authorities regarding such matters as investment authorizations or registrations to determine when a direct investment relationship exists.

“Indirect Connections”

According to international standards, direct investment statistics should cover all enterprises in which the investor directly or indirectly has a direct investment interest. This means that once the 10 percent “across the border” link is achieved with, say, enterprise A, certain other enterprises related “down the line” to enterprise A also should be treated as direct investment enterprises. The statistics should cover all transactions between the direct investor and enterprise A and certain defined affiliates of A (and between the affiliates if they are in separate economies). If indirectly owned enterprises are excluded from a country’s definition of a direct investment enterprise, then that country’s direct investment statistics may be incomplete.

The majority of survey respondents said that they take account of indirectly owned enterprises in their statistics; 13 of the 19 industrial country respondents and 14 of the 19 developing country respondents are in this group. Among these countries, procedures for determining the coverage of the indirectly owned enterprises vary considerably. The variation was not tested in the survey. However, it is known that some countries—including the United States, Canada, Australia, and the United Kingdom—take extensive account of indirectly owned enterprises, while others cover them less comprehensively. For example, Germany takes account only of loans between indirectly connected enterprises, and only since 1990.

Real Estate

According to international standards, land and structures (except structures owned by a foreign government) are always considered to be owned by a resident of the country in which they are located. Therefore, if a nonresident owns real estate, by convention he is regarded as owning, instead of a nonfinancial asset, a claim on an actual or notional resident entity to which the ownership of the real estate is attributed. Thus when a resident of one country directly purchases or sells real estate in another, the value of this purchase should be recorded as direct investment.172

About three-fourths of survey respondents adopt fully or partially the recommended treatment of direct purchases of real estate. However, only about half of these particular respondents adopt it for investment in holiday homes and second homes as well as for investment in other (usually more commercial) types of real estate.

Mobile Equipment

Mobile equipment (for example, ships, aircraft, and oil-drilling rigs) that operates in the territory of an economy for one year or more is considered always to be owned by a resident of that economy. When the actual legal owner of the equipment is a nonresident, the owner can only be considered to have a claim on an actual or notional resident entity that must be construed as being the owner. Thus, if a ship is chartered (under operational lease) by a resident of one country to a resident of another for 12 months or more, the actual owner should be considered to have a direct investment enterprise in the other country, and a direct investment capital flow should be recorded equal to the value of the ship. A similar transaction should be recorded if, say, an offshore oil-drilling rig owned and managed by a resident of one country operates for 12 months or more in the territorial waters of another.173

Few respondents adhere strictly to these particular rules. Although not tested by the survey, it is understood that most countries allocate ownership to the country of residence of the actual owner; however, one respondent, Norway, stated that prior to 1988 it had allocated ownership of Norwegian-owned ships to their country of registration. Only eight respondents, including three industrial countries, said that they fully or partly adopt the recommended treatment in cases when the equipment is “directly owned by a nonresident.”

Construction Activity

A similar international standard applies in certain cases of international construction activity. An enterprise can undertake the construction of plants and buildings abroad through an affiliate abroad or directly undertake the work itself. The current international standard is that construction work abroad should be regarded as being done by a direct investment enterprise resident in the economy in which the work is being carried out.174

The standard generally is followed when construction is carried out by an incorporated enterprise established abroad or by an “established branch.” However, it is common practice for countries to regard earnings from construction work carried out “directly abroad” as services and to record them in the current account; that is, in these instances the existence of a direct investment enterprise is not recognized and no direct investment transactions are recorded. Only 10 respondents—the majority are developing countries—said that they fully or partially follow the standard when the work abroad is “directly managed from the host country of the construction enterprise.” Five of these 10 countries adopt it only when the construction contract is for 12 months or more; these include the United States and the United Kingdom.

Direction of Investment

Direct investment statistics should cover inward and outward direct investment separately. All respondents measure inward direct investment. All except six developing country respondents also measure outward direct investment. However, for the six developing countries the value of outward transactions is relatively small.

Sectoral Coverage

Direct investment statistics should cover all industrial sectors. However, some countries exclude certain sectors from the coverage of their statistics. About three-fourths of respondents stated that their direct investment statistics, both inward and outward, cover all sectors. Among industrial countries, very few reported that there are any exclusions. The Netherlands and Belgium-Luxembourg exclude special financial institutions and holding companies respectively. These are, in both cases, domestic subsidiaries of foreign companies set up to channel capital flows via these countries. Canada excludes the construction sector from its outward statistics. Among developing countries, eight stated that there are sectoral exclusions from the coverage of either their inward or their outward statistics or both: the most commonly excluded sectors reported by seven compilers were banking or finance.

Coverage of Forms of Capital Flows

According to international standards, direct investment covers all forms of capital transactions between enterprises in a direct investment relationship, with the exception of short-term flows between depository institutions and their affiliates.175 These forms include equity investment, reinvestment of earnings, and other transactions, both long-term and short-term. They include all direct (cross-border) purchases and sales of real estate, whether for “commercial” or “noncommercial” purposes. In addition to transactions associated with a cash payment, they include changes in intercompany accounts that represent the accounting offset to goods, services, income, and capital transactions between related enterprises not matched by a cash payment.

All respondents include at least some forms of equity investment in their direct investment statistics. More than four-fifths said that they include noncash acquisitions of equity (such as through the provision of capital equipment and manufacturing rights). Among the six respondents that exclude noncash acquisitions are three industrial countries: Japan, Belgium-Luxembourg, and Italy. About three-fourths of the respondents said that they include “direct purchases of land and real estate for commercial purposes,” and about half of these respondents also include “direct purchases of land and real estate for noncommercial purposes.” Among the respondents that exclude both commercial and noncommercial real estate transactions are Belgium-Luxembourg, Italy, and Spain. Among those respondents that exclude only noncommercial real estate transactions are the United States, Japan, Denmark, and Germany; however, a number of these countries publish separate series on the omitted real estate transactions.

Twenty-three respondents (about 60 percent) include reinvestment of earnings in their direct investment data. Two countries (Sweden and Finland) added these data to their accounts during 1991. Among the 11 industrial economies that do not include reinvestment in their direct investment statistics are Japan, Canada, Belgium-Luxembourg, Denmark, France, Italy, and the Netherlands. However, Denmark, France (outward only), and the Netherlands compile separate data, and Belgium-Luxembourg is developing a survey to collect the statistics.

About three-fourths of respondents include some form of nonequity transactions in their statistics. The only industrial country respondent that does not is Denmark, while nine of the developing country respondents do not. All of the respondents that include any nonequity transactions—with one developing country exception—include long-term loans. However, at least three countries—France, Spain, and Portugal—only include such loans when they have an original contractual maturity of more than five years. Many others, in accordance with international conventions, consider long-term loans to be those that have a maturity of more than one year. Only 60 percent of the respondents that include long-term loans said that their loan data include long-term trade credit.

Fourteen respondents reported that they include short-term loans (other than interbank loans) in their direct investment definition. Therefore, 24 respondents (about two-thirds) said that they exclude these short-term loans.176 Among the 10 industrial country respondents that do not include any short-term loans are Canada, Japan, France, and Germany, although several of these countries provide separate data on certain of these transactions. Among the 14 respondents that do include short-term loans in their direct investment definition is Finland, which expanded its definition to include them during 1991. In this group, however, only one-third said that short-term trade credit is included in the loan data.

Notwithstanding international standards, of the 15 respondents that include any short-term transactions in their direct investment definition, 9 include short-term transactions between banks and their affiliates. Four industrial country respondents are in this group: Australia, New Zealand, Norway, and Switzerland.

Some of the countries that exclude short-term interbank transactions exclude certain other banking transactions as well. For example, the United States excludes all transactions that do not represent “permanent investment” by parent banks in their affiliates, while the Netherlands and the United Kingdom exclude all nonequity long-term interbank transactions. The United Kingdom recently commenced excluding short-term transactions between security dealers and their affiliates.

Realized and Unrealized Capital Gains and Losses

International standards state that both realized and unrealized capital gains and losses should be excluded from the calculation of reinvested earnings data. Of the 23 respondents that include reinvested earnings, only 4 said that their data exclude realized gains and losses, and 13 said that their data exclude unrealized gains and losses. After allowing for a number of respondents that did not know whether their data exclude either realized or unrealized gains and losses or both, about 14 of the 23 definitely said that their data include realized capital gains and losses, but only 3 definitely said that their data include unrealized gains and losses. The United States is among the countries that include capital gains and losses in their data, but it alone publishes separate data on these gains and losses.

“Reverse Transactions”

According to international standards, capital transactions between parties in a direct investment relationship are to be recorded on a net basis; that is, flows from a direct investment enterprise (an affiliate) to a direct investor (a parent)—“reverse transactions”—are to be treated as disinvestment under the direct investment heading.177 This treatment should even apply when affiliates raise funds for lending to their parents.

In the case of loans, there are various departures from this principle. First, the Netherlands treats loans from an affiliate to its parent as investment by the affiliate rather than disinvestment by the parent. Thus, Dutch compilers record loans from a resident enterprise to its direct investor as outward investment rather than as negative inward investment. Second, almost half the respondents do not record loans from an affiliate to its parent as direct investment at all. Allowing for the fact that about one-fourth of respondents do not include any loans in direct investment, this means that another quarter regard loans from a parent to its affiliate, but not “reverse loans,” as direct investment. Six industrial countries, among which are Japan and Germany, follow this treatment. Third, notwithstanding international guidelines, a few countries “look through” the activities of finance affiliates that have been set up purely to raise or channel funds to their parents, so that any such reverse loans received by the parents are likely to appear in the country’s accounts as inward portfolio or other kinds of investment. Four countries—Canada, Australia, and two developing countries—that normally record reverse loans under direct investment stated that they do not do so in the case of loans extended by an affiliate that has been “set up purely to raise or channel funds to its parent.”

In the case of equity investment, a variety of treatments also occurs. This was not tested in the survey, but it is known that in the Australian and French statistics equity investments by an affiliate in its parent are uniformly treated as disinvestment, in accordance with the international standards. In the Dutch statistics, if a resident affiliate acquires shares in its parent company, this is treated as outward direct investment rather than as inward disinvestment. In the German and U.S. statistics, such reverse investments by a resident affiliate are recorded as portfolio capital, until the equity threshold—20 percent in Germany and 10 percent in the United States—is reached, after which these investments are treated as outward direct investment. In the Australian, French, Dutch, German, and U.S. statistics, a symmetrical treatment applies for investments by a foreign affiliate in a resident parent.

Sources and Methods

Fifteen respondents—including six from industrial countries—reported that their direct investment statistics are based on company surveys. Generally, these respondents reported that their data are obtained from resident investors and affiliates; in a few instances the data are supplemented by information obtained from the authorities responsible for investment approval or exchange control or from other sources such as customs records. Twelve of these respondents said that reporting in their surveys is compulsory, and three said that it is voluntary.

The remaining 23 respondents—including 13 from industrial economies—said that their collections are solely or partly based on the reporting of each transaction. In a number of cases, these transaction reports are supplemented by company surveys and other data sources. The industrial country reporters said that their data are most frequently submitted by resident banks or other resident enterprises and investors. In a few cases, data are received or supplemented by information from the authorities responsible for exchange control or investment approval. Among developing country reporters, most said that they make sole or partial use of data obtained from the authorities responsible for investment approval or exchange control. A minority reported using bank data, usually in combination with the other sources, and several said that they also use data collected directly from enterprises. Twenty-two of these 23 countries said that their collections are based solely or largely on compulsory reporting.

Twenty-four respondents—including 12 industrial countries—said that the collected raw data relate to the whole population of transactors. Of the other 14, six make adjustments to allow for transactors not covered directly. These include several industrial countries that conduct periodic comprehensive benchmark surveys but adjust the data collected using less comprehensive quarterly inquiries. The remaining eight make no adjustments. They include a number of industrial countries in which reporting thresholds have been introduced to reduce reporting and processing costs, but in which the excluded values are small. They also include several developing countries in which coverage is known to be deficient.

Appendix IV Detailed Direct Investment Tables

The four tables in this appendix present the more detailed data that underlie the summary tables appearing in Chapter 3.

Tables 69 and 70 relate to the reinvestment of earnings category. Table 69 gives a more detailed breakdown of the Working Party’s adjustments to the global discrepancy than is shown in Table 14 of the report. Table 70 shows the individual bilateral comparisons on which Table 15 is based.

Table 69.

Global Reinvestment of Earnings Discrepancy and Adjustments, 1986–89

(In billions of U.S. dollars; outflows ( – ))

article image

Balance of Payments Statistics Yearbook, 1990.

Data provided to the Working Party but not reported in the Yearbook. Includes country estimates provided for the Working Party’s purposes and new official series compiled by several countries.

Amounts reported as reinvested by the United States, the United Kingdom, Germany, and the Netherlands in countries not reporting inward reinvestment of earnings.

Amounts reported as reinvested in their economies by the countries mentioned in footnote 3 against countries not reporting outward reinvestment of earnings.

Takes account of the reinvestment of earnings in six offshore centers (The Bahamas, Bahrain, the Cayman Islands, the Netherlands Antilles, Panama, and Singapore) missing from (or not fully covered in) the Yearbook, and also not completely covered in adjustments 2a(2) and 2a(3).

Records German earnings when earned.

To exclude U.S. capital gains and losses.

Table 70.

Bilateral Comparisons of Reinvestment of Earnings, 1986–88

(In billions of U.S. dollars; outflows ( – ))

article image
Source: National tabulations.

Note: Data reported by the United States exclude capital gains and losses. Data reported by Germany have been moved back one year.

Tables 71 and 72 relate in a similar way to the other direct investment capital flows category. Table 71 provides more detail on adjustments than is shown in Table 16, and Table 72 gives the bilateral comparisons that are summarized in Table 17.

Table 71.

Discrepancy on Global Other Direct Investment Capital Flows and Adjustments, 1986–89

(In billions of U.S. dollars; outflows ( – ))

article image

Balance of Payments Statistics Yearbook, 1990.

Takes account of the direct investment transactions not included in the Yearbook.

Entries in the “abroad” column show amounts reported by the United States, Australia, France, Germany, the Netherlands, and the United Kingdom as inward investment from seven offshore financial centers that do not report (or report incomplete data) to the Fund: The Bahamas, Bahrain, Bermuda, the Cayman Islands, Hong Kong, the Netherlands Antilles, and Panama. Entries in the “in reporting economy” column show amounts reported by the same countries and Japan as outward direct investment to the seven offshore centers. Small reported or estimated amounts for The Bahamas, Hong Kong, the Netherlands Antilles, and Panama have been deducted.

Shifts data on certain short-term transactions between affiliated enterprises, recorded as “other capital” in German statistics, to direct investment.

Shifts data on real estate transactions, recorded in other items in the statistics of Japan, Denmark, Germany, Italy, and Spain, to direct investment.

Shifts data on transactions with international organizations, recorded as direct investment by Belgium-Luxembourg and Italy, to “other capital.”

Entries in the “abroad” column are amounts reported by the United States as inward direct investment from Saudi Arabia, which does not report data on outward direct investment. Amounts in the “in reporting economy” column are made to shift “oil sector and other investment” in Saudi Arabia from direct investment to “other capital.”

Takes account of capital flows of the U.S.S.R., which are not included in the Yearbook.

Additions or corrections to the data of various countries based on responses to the Working Party’s Special Questionnaire and other inquiries.

Table 72.

Bilateral Comparisons of Other Direct Investment Capital Flows, 1986–88

(In billions of U.S. dollars; outflows ( – ))

article image
article image
Source: National tabulations.Note: Totals exclude country pairs where data from only one side are available, or where data are not available for all three years.

Appendix V The BIS International Bond Data Base

In response to the rising volume of securities issued in international markets, the Bank for International Settlements developed in 1987 a new data base on international bonds. This data base, which covers both Eurobonds and foreign issues in national markets, provides from December 1985 comprehensive information on issues announced, issues completed (both gross and net of redemptions), and the amount of bonded debt outstanding. These data focus on liabilities arising from the issuance of international bonds. No data are available on the holders of these securities. Bonds issued by developing countries as a result of debt-rescheduling agreements are at present not covered in the BIS statistics.

The BIS data base includes certain historical data provided by the OECD for issues prior to 1980, and it is updated quarterly with information from two sources. The Bank of England provides details on all individual issues announced during the quarter under review on the basis of market information.178 The Association of International Bond Dealers (AIBD), using data obtained from CEDEL and Euroclear, gives the outstanding value of individual issues still in the market at the end of each quarter. The AIBD data take account of new issues launched during the quarter and of scheduled and early bond redemptions.

At the end of 1990, the BIS data base contained information on 21,400 bond issues, of which 13,500 were still outstanding. The AIBD data base identified 12,100 issues outstanding at the end of 1990, of which about 10,100 were explicitly identified as common to the BIS data base.179 Only matched issues are updated with AIBD stock and repayment information, although the other AIBD issues are being researched by the BIS.

Using the BIS data base, it is possible to analyze activities in the international bond market according to various characteristics of the issues, including the following:

  • —type of paper (straight fixed, floating, zero coupon, equity related);

  • —currency of issue (which in the case of foreign issues also means by market);

  • —type of borrower (governments, international institutions, commercial banks, other financial institutions, or other borrowers); and

  • —nationality of borrower or borrower’s parent.

Table 73 provides details on international bonds issued by type of instrument, issuer, and currency denomination. These breakdowns are available in all combinations.

Table 73.

International Bonds, by Type and by Currency of Issue, 1986–89

(In billions of U.S. dollars)

article image
Source: BIS

New issues completed, less scheduled and early redemptions.

In order to estimate flows in dollar terms, the BIS converts amounts of new completions of nondollar bonds at exchange rates prevailing at the announcement dates, while bond redemptions are converted at beginning-of-quarter exchange rates.

The BIS flow statistics allow for scheduled redemptions and early partial redemptions by the issuer, which are identified by the AIBD in the majority of cases. Parts of an issue repurchased by the issuer and held in treasury are not registered as redemptions, because the issuer can resell these securities again on the market before maturity. Large amounts of international bonds issued by some borrowing countries are known to be purchased by residents of the same borrowing country. Such purchases are not taken account of by the BIS unless the entire issue is redeemed by the issuer. Also, there is at present no allowance for new bonds issued as a result of the exercise of warrants, and very little information is available about the exercise of conversion rights and warrants attached to equity-related bonds. Zero-coupon bonds are entered in the statistics at their discounted issue values, with subsequent accretions counted as new borrowing.

As regards position estimates, the BIS figures show the nominal value of outstandings, with nondollar issues converted at end-of-quarter exchange rates. At the end of 1989, the total stock of outstanding issues totaled about $1,250 billion.

Although the BIS data are not a perfect fit with balance of payments norms, they provide a useful source of information to compilers and were an important element in the Working Party’s effort to develop a matrix of global flows of portfolio capital. Along with the Special Questionnaire and other sources, information from the BIS data base also enabled the Working Party to construct estimates of the global stock of cross-border bond liabilities for the year 1988. These estimates are discussed in Appendix VII.

Appendix VI Criteria for the Geographic Allocation of Securities Transactions

For analytic purposes, many countries compile geographic detail on cross-border securities transactions. Such data may also assist in identifying statistical errors and gaps in national data, although the discussion in Chapter 4 on bilateral comparisons shows that this is difficult to achieve. In its research, the Working Party used bilateral information on portfolio flows to derive some adjustments for missing data in world capital flow statistics. In the draft fifth edition of the Balance of Payments Manual, aspects of regional allocation, covered in an appendix in the fourth edition, are included as a chapter, reflecting growing international interest in this area.

Compilation systems usually address the issue of geographic allocation of capital flows by following the transactor principle or the debtor-creditor principle. Both principles refer essentially to the kind of information to be collected on the foreign counterparty to the transaction. Under the transactor principle, respondents are to allocate transactions to the residence of the foreign party to the transaction. Under the debtor-creditor principle, changes in claims and liabilities are to be allocated to the residence of the foreign debtor and creditor, respectively. In the latter case, the results will be consistent with a set of international investment position statistics that would be obtained by a benchmark survey of cross-border assets and liabilities in portfolio securities.

It is more difficult to obtain accurate geographic detail on securities transactions than on other types of capital flows. Portfolio transactions are distinguished from other kinds of capital flows by the existence of broad secondary markets in existing assets and by the widespread participation of many types of financial intermediaries in cross-border transactions. For this reason, when the transactor principle is applied, compilers are likely to obtain information on the residence of a foreign intermediary acting on behalf of a transactor, rather than information on the residence of the foreign transactor. For example, domestic securities sold by a U.S. resident to a securities dealer in the United Kingdom would be recorded in the U.S. accounts as a transaction with the United Kingdom. If the U.K. securities dealer were an agent for a German transactor, that attribution would not show any transaction with Germany. The same weakness applies to geographic attribution of transactions in foreign securities. In concept the transactor principle should reveal the exact direction of a capital flow, but the main weakness of a transactor-based compilation system is that it cannot see through the residence of apparent transactors to the beneficial purchasers or sellers.

When the debtor-creditor principle is applied to the sale of domestic securities, geographic attributions have the same weakness as that described above for the transactor principle. That is, the residence of the foreign creditor usually cannot be ascertained, and the capital flow can only be attributed to the immediate foreign party, which may be an intermediary. Thus, the debtor-creditor approach may give the same potentially misleading picture about the geographic distribution of a country’s international liabilities. Transactors in foreign rather than domestic securities, however, know the issuer (debtor) of the securities, so that it is possible to implement the debtor-creditor principle for transactions in foreign securities.

Recording practices among countries vary considerably with respect to these two compilation systems. Countries adhering to the transactor principle include the United States, Canada, and the Netherlands. Germany, France, Italy, and Australia employ the debtor-creditor principle to determine the geographic allocation of securities transactions. However, because of the problems noted earlier, it is acknowledged that, in practice, liabilities may be allocated to the residence of the intermediary in the transaction. In the case of the Japanese compilation system, transactions in U.S. government securities are attributed to the United States (debtor-creditor principle), while transactions in other foreign securities generally follow the transactor principle. However, some transactions carried out by Japanese investors in foreign markets are being attributed to the country where the security is listed (for example, to Luxembourg in the case of many Eurobonds). Clearly, this wide range of reporting practices among countries gives rise to asymmetries in inter-country data comparisons. The United Kingdom and Switzerland, both important participants in the international securities markets, do not compile geographic detail on cross-border securities transactions of their residents.

The Fund’s Manual endorses the debtor-creditor principle because it reflects the balance of payments criterion of recording changes in ownership of foreign claims on, and in liabilities to, the rest of the world. However, it is recognized that the debtor-creditor principle can be employed rigidly only in allocating transactions in foreign securities.

Countries can do little but allocate transactions in domestic securities to the immediate foreign party to the transaction, which may be an intermediary. For such transactions, neither system can be used successfully to solve the problem of asymmetries in bilateral portfolio flows.

Appendix VII Estimation of Global Stock of Cross-Border Bonds

As part of its examination of portfolio capital flows, the Working Party constructed estimates of the global stock of cross-border bonds, in part to cross check the cumulative effect of the large imbalances in the recorded transactions data. The stock figures reported in the Working Party’s Special Questionnaire were incomplete in many respects for both industrial and developing countries, and therefore some degree of estimation was required to fill gaps and make the data uniform.

The derived estimates, which are discussed in Chapter 4, are summarized in Table 74. The global stock of cross-border bond liabilities is estimated to have been $1,600 billion at the end of 1988. It is estimated that $200 billion of these securities were held by foreign monetary authorities as part of their foreign exchange reserves, with the balance of $1,400 billion belonging to portfolio investors. As measured from the creditor side, portfolio holdings of cross-border bonds totaled only $1,000 billion, leaving an imbalance of $400 billion. While this imbalance partly reflects differences in country methodologies and valuation procedures, the overriding problem is likely to be the underrecording of asset accumulations. There also exists the possibility that reported liabilities may be overstated.

Table 74.

Global Stock of Cross-Border Bond Assets and Liabilities, Year-end 1988

(In billions of U.S. dollars)

article image

Assets exclude bonds held as part of reserves.

Geographic detail on these bond positions is given in Table 22 in the body of the report.

Liabilities constituting foreign authorities’ reserves.

The data sources and methods used in constructing these stock estimates are discussed below. Sources included:

  • —stock figures reported in the Special Questionnaire and other information provided by national authorities;

  • —stock figures from the Bank for International Settlements data base on international bonds and the World Bank’s Debtor Reporting System;

  • —balance sheets of international organizations; and

  • —published capital flow figures.

The estimates take account of the major adjustments discussed in Chapter 4 and of data gaps where countries report transactions but no stock data. The following provides a brief overview of the methods used in constructing the liability and asset positions.


For industrial countries, about 90 percent of the estimated stock of cross-border bond liabilities was based on figures reported in the Special Questionnaire. Several adjustments were made to these data, using information provided by compilers and other sources. The Netherlands’ submission was raised by $47 billion to account for bonds issued by special financial institutions that are netted out in that country’s external accounts. Japanese data were lowered by $44 billion to take account of resident purchases of international bonds issued by residents. The German data were raised by $29 billion to include “Schuldscheine” (notes with limited marketability) of the official sector. For countries that classified international bonds as part of loans rather than as portfolio positions (or that reported no stocks), estimates of outstanding international bonds were constructed using the BIS data base.180 Italy, Denmark, New Zealand, and Austria accounted for most of the $100 billion added to reported totals. To account for domestically issued bonds sold to nonresidents, estimates were derived by cumulating capital flows. Bond liabilities of international organizations were derived from their balance sheets and from information in the BIS data base.

For developing countries, only about a third of the estimated stock could be based on data reported in the Special Questionnaire. The balance was derived using data from the World Bank’s Debtor Reporting System and the BIS data base, supplemented with flow accumulations for positions not covered by these data bases.

In order to compare bond liabilities with the stock of portfolio bond assets, an adjustment was made to liabilities to deduct securities held as part of other countries’ reserves. This amount was estimated to be $200 billion at the end of 1988. Some three-fourths of this total was based on information reported to the Working Party by large reserve holders in respect of their securities holdings. Holdings by other countries were derived from a combination of reported and estimated reserve flows.


The stock of portfolio bonds held by industrial countries was based largely on information reported in the Special Questionnaire. Estimates were constructed for France (based on a 1989 French inquiry on resident holdings of foreign securities) and for Japan, whose stock figures did not distinguish between bonds and equities. Canadian data were adjusted to include securities held by banks.

For developing countries, bond holdings were derived mainly by cumulating published capital flows, which for some countries likely included equities. Also, the flow data were available only from the early 1970s, because of different classifications used in earlier periods. The bulk of the overall estimate represents holdings by Saudi Arabia and was based on the large portfolio outflows recorded during the 1970s and early 1980s. The recorded portfolio flows for other oil-producing countries were relatively small. This may in part reflect different institutional arrangements, as it is known that some countries carried out investment operations through agencies in other countries. Obviously, several parts of the asset calculation reflect measurement weaknesses.

Some of the characteristics of stock compilation methodologies, including valuation methods, are shown for selected industrial countries in Table 75. Several countries construct stock estimates by cumulating capital flows. In such cases the stock estimates cannot serve as a reliable check on reported capital flows.

Table 75.

Methods Used by Selected Countries to Estimate Bond Positions

article image

Benchmark survey conducted at five-year intervals.

Bonds and equities not distinguished separately.

Assets of financial institutions are at market values; estimates for nonfinancial companies are at book values.

Surveys conducted in respect of 1989 positions.

Benchmark survey of U.S. corporate securities conducted at five-year intervals.

Data for international bonds are obtained from surveys; for domestic bonds estimates are derived from flow data.

Appendix VIII Do Revisions Reduce the Global Capital Account Discrepancy?

The first-published figures for national capital accounts are usually preliminary, and global totals naturally are subject to revision in subsequent years. Do such revisions tend to reduce imbalances in the global accounts?

Tables 76 to 79 show a simple “revisions analysis” for the global capital account discrepancy, using figures published in successive issues of the Balance of Payments Statistics Yearbook, Part 2, from Volume 34 (1983) through Volume 41 (1990). Stubs in the tables show publication dates, and column headings are years to which the data refer. Reading down the columns of Table 76 illustrates the evolution of each annual figure through time. For example, the first-published world discrepancy for the year 1983 was $73.2 billion. This figure was revised in each succeeding issue of the Yearbook and, as published in Volume 41, was $67.6 billion.181

Table 76.

Revisions Analysis of Capital Account Discrepancy: Discrepancy Amounts, 1982–89

(In billions of U.S. dollars; outflows ( – ))

article image
Table 77.

Revisions Analysis of Capital Account Discrepancy: Revision Amounts, 1982–88

(In billions of U.S. dollars; outflows ( – ))

article image
Table 78.

Revisions Analysis of Capital Account Discrepancy: Revisions due to Selected Coverage Changes, 1982–88

(In billions of U.S. dollars; outflows ( – ))

article image