THE RELIABILITY of balance of payments statements is a subject of concern to both the compilers and the users of these statistics. Although the figures could be expected to be about as accurate as the general run of economic data,1 the topic frequently comes under discussion for at least three main reasons:


THE RELIABILITY of balance of payments statements is a subject of concern to both the compilers and the users of these statistics. Although the figures could be expected to be about as accurate as the general run of economic data,1 the topic frequently comes under discussion for at least three main reasons:

THE RELIABILITY of balance of payments statements is a subject of concern to both the compilers and the users of these statistics. Although the figures could be expected to be about as accurate as the general run of economic data,1 the topic frequently comes under discussion for at least three main reasons:

(1) Some of the most critical relationships to be observed from a balance of payments statement depend on net changes that are very small in relation to the gross magnitudes of the stocks or flows involved. For example, a U.S. deficit, say of $1 billion in a year, could be considered quite significant, yet it would represent the net change in a stock of reserve assets and liabilities each amounting to some $15 billion. Furthermore, it would be related to visible and invisible trade flows amounting to some $40 billion in each direction and to changes in stocks of foreign assets totaling roughly $90 billion and of liabilities (excluding reserves) totaling $45 billion.

(2) A balance of payments statement is presented as a double-entry accounting statement, in which credits and debits are made to balance out with mathematical precision by incorporating a residual item representing net errors and omissions. This item may sometimes be substantial in relation to the other net balances (e.g., monetary movements) given in the statement. The interpretation of developments may, in fact, depend to some extent on an assumption about which items give rise to the errors. The form of presentation thus tends to focus the user’s attention on the existence of discrepancies in a way that is not usual with other statistical series, most of which are not shown in the form of a self-contained, balanced statement.

(3) Almost every transaction entered in one country’s statement has, in principle, its counterpart in the figures of other countries. Data covering the same transactions are thus available from two independent sources, which may be compared, and the question as to which set of figures is the more reliable frequently arises.2

The fact that a balance of payments statement must either be internally consistent or reveal its inconsistency by an explicit entry for net errors and omissions probably provides reasonable assurance for most countries that their individual statements are adequate guides for policy formation in the compiling countries. This assurance is particularly valid because the “over-all” balance of payments surplus or deficit, which is the single figure most likely to concern policymakers,3 can be measured indirectly (from “below the line”) by the official financing supplied (i.e., changes in official reserves and similar transactions), for which the data are in all likelihood quite reliable. All other transactions in the balance of payments then simply explain how the surplus or deficit arose, and gaps or inaccuracies in this knowledge, while naturally undesirable, are not likely to lead to wrong policy decisions.

Flaws in the statistics become a matter of greater concern in connection with some other types of analysis for which the figures are used. Studies of international capital flows, in particular, usually depend on the balance of payments as their main data source, and the question of the quality of the statistics arises immediately. Omissions and inconsistencies in the figures, together with differences between the information on the same transactions reported by partner countries, are all too readily apparent to analysts attempting to work with the data. To resolve these difficulties, users often assume that (1) the main source of errors and omissions in the balance of payments statements is likely to be unrecorded capital flows, and (2) the balance of payments figures reported by developed countries are more accurate than those reported by the developing countries. The first of these assumptions means, in effect, that the recorded figures for goods, services, and transfers are thought to be fairly accurate, at least in comparison with the available information on capital flows, and may thus be used as an indirect measure of capital movements. The second assumption implies that the figures reported by developed countries may be used to bridge gaps in the statements of the less developed areas. From time to time, figures are quoted that have been based on both assumptions; that is, figures purporting to represent capital inflows into less developed countries have been constructed from the developed countries’ records of current transactions.

The examination of the reported balance of payments statistics undertaken in this paper does not uncover any strong evidence that would seem to lend support to either of these assumptions. In fact, quite a convincing showing can be made that both the developed and less developed countries’ statistics contain major inconsistencies; it can also be demonstrated that the current as well as the capital account is subject to discrepancies of substantial magnitude. On the other hand, the approach followed cannot be expected to yield any positive evidence of the superiority of one series over another. At most, it can disclose where some of the main pitfalls seem to lie.

The Balance of Payments of the World

How the ideal statement should look

If a comprehensive and accurate statistical statement of the international transactions of the world as a whole could be compiled in accordance with the concepts and definitions of the Fund’s Manual,4 only two broad categories of transaction—nonmonetary gold and monetary gold—would fail to show equal credit and debit entries. Other main categories (i.e., goods, services, transfer payments, and capital) should ideally cancel out, leaving a net balance of zero in each category; for example, a merchandise export recorded by one country should be exactly offset by a merchandise import recorded by another country.

Gold transactions, however, are classified in accordance with a convention5 that treats all such transactions carried out by the domestic nonmonetary sectors (whether the partner is the domestic monetary sector or a foreigner) as nonmonetary gold, and all transactions of the domestic monetary sectors (regardless of the partner) as monetary gold. Consequently, transactions between the nonmonetary and monetary sectors do not cancel out when the entries for nonmonetary and monetary gold are totaled separately for the world as a whole. Instead, these two accounts should show equal entries with opposite signs; the entry for monetary gold represents the net change in the world’s official gold reserves, while the entry for nonmonetary gold shows the source of this change (i.e., production less consumption, including hoarding or dishoarding and use or return from industry and the arts).

Because the balance of payments of the world is thus in theory a known quantity, it should be possible to gain some insight into the reliability of the reported figures by adding them up to a world statement and seeing how that statement compares with the theoretical ideal. In practice, this method has several serious shortcomings that make the approach less promising than might be anticipated. Nevertheless, the results of an exercise of this sort, carried out with reported balance of payments data for 1961-64, are of considerable interest.

Gaps in the reporting of statistics

The Fund’s Balance of Payments Yearbook for recent years, the source of the reported balance of payments figures used in this paper, covers statements for about 80 countries and country groups. These countries account for approximately 80 per cent of the world’s merchandise exports,6 and probably no less than this proportion of most other types of international transaction. Some 60 per cent or more of the exports of countries not reporting balance of payments data is attributable to the Soviet countries and Mainland China, and over two thirds of the exports of these countries go to countries in the same group. The implications of this seem to be that close to 90 per cent of the international transactions of the rest of the world is reported by both partners, and much of the remainder is reported by one of the partners. The only transactions of the rest of the world that are entirely unreported are those among non-reporters and between them and the Soviet countries and Mainland China, which must be quite minor.

One further qualification about the coverage of the reported statistics is that international organizations are considered by definition as residents of an international area outside national boundaries. Some international companies whose country of residence is difficult to determine are treated, as a practical matter, the same way (i.e., no reporting country regards them as residents). A prominent example is the merchant shipping fleets flying flags of convenience (including those of Honduras, Liberia, and Panama), which are considered as nonresidents even by those countries whose flags are used. A world summary compiled from country reports will thus cover only one side of the transactions between countries and international organizations or companies.

The statistical gap caused by the failure of one partner to report may or may not manifest itself when a consolidated statement of international transactions for the world is prepared. If the two flows making up a transaction7 are classified in different balance of payments accounts or items (e.g., the sale of a commodity for cash, which is recorded as a merchandise credit and a capital account debit), the two affected accounts will be out of balance by the same amount, one showing an excess of credits and the other an excess of debits. On the other hand, when both flows are classified in the same account (the sale of a financial item for cash, for instance), the figures for the account as a whole will show no evidence that one partner’s transactions have not been reported. In practice, the exchange of merchandise for merchandise (barter) sometimes occurs but does not represent a significant component of total trade; therefore, trade transactions reported by only one partner are likely to show up as an asymmetry in the world statement. The exchange of services for services also seems most unlikely, and transactions where both flows would be classified as transfer payments are precluded by definition. However, the interchange of financial items is common, and missing partner data on such transactions would not appear as an asymmetry in a world balance of payments statement.

If the capital account were to be subdivided into categories covering individual types of assets and liabilities, and if these categories were defined symmetrically from the viewpoint of both the creditor and the debtor, a clearer indication of missing partner data might emerge.8 The classification scheme of the Fund’s Manual, however, is based on an entirely different principle, which is intended to be more suitable for the analysis of the balance of payments per se. The primary breakdown of the Manual’s capital account emphasizes the sector of the domestic creditor or debtor. Since virtually any asset or liability of a given domestic sector may be held abroad by foreigners in any sector, item-by-item matching, say of claims on foreigners with liabilities reported by foreigners, becomes difficult, if not impossible. A further complication arises because transfers of foreign assets (or even liabilities) between different domestic sectors, such as sales of foreign exchange held by the commercial banks to the central monetary authorities, are in principle recorded in the balance of payments. These transactions have no counterpart in the statement of any foreign partner, and thus they do not cancel out in a global summary, unless the capital account is taken in its entirety.

The main exception to the asymmetrical definition of capital items is direct investment capital; for transactions in this category, the Fund’s Manual definition should in principle lead to consistent reporting by both partners. In addition, for the domestic monetary sectors, the Manual provides for the identification of the foreign sector involved in some important types of transaction. Some matching of subcategories within the capital account could thus be attempted, but this paper goes no further than a division between monetary gold and other capital items.

Errors in reporting

Errors in the statistics reported by countries affect a consolidated world payments statement in the same ways as the omissions caused by nonreporting on the part of one or both partners. Each individual country statement balances out to zero through the inclusion of a balancing item—“net errors and omissions.” If a transaction is omitted (or overvalued or undervalued by the same amount) by both partners, no evidence of this error will be found in the world statement. Both the category concerned and the balancing item will be equally affected but in opposite directions in the statements of the two partners, and adding the two or comparing them will be no more revealing than was the balancing item in each of the original statements. On the other hand, it is perhaps rather unlikely that two reporters will make the identical mistake, and an error appearing in only one statement will show up as net balances in two categories (one of them being the balancing item) of a world summary. An inconsistency in the classification of any transaction in the statements of two partners should also show up in two categories, although the balancing item would not be one of the categories affected.

A world balance of payments summary thus cannot reliably provide clues to all the possible statistical weaknesses in the reported statements of individual countries. The aggregates may conceal offsetting errors, as well as transactions omitted by both partners. Nevertheless, such a tabulation should provide some evidence of systematic tendencies to misreporting and gross errors of types that are not offsetting.

Global summary of reported statistics, 1961-64

Reported balance of payments statistics, compiled from Volumes 16 and 17 of the Fund’s Balance of Payments Yearbook, are summarized in Table 1. Countries are classified into two groups, developed and less developed, according to the scheme used in the world trade table in the Fund’s International Financial Statistics. Individual categories for each country group separately are not expected in principle to show net balances of zero. It may nonetheless be of some interest to observe whether a global imbalance, say an excess of credits, is composed of excess credits recorded by both groups (which might possibly be evidence of some sort of systematic error in the statistical reporting), or whether an excess of credits reported by one group is only partly offset by an excess of debits reported by the other, a situation that would presumably be difficult to evaluate. The grouping chosen was considered relevant because it might be expected, a priori, that the developed countries would produce more accurate statistics.9 There are many known exceptions to such a generalization, however, and some less developed countries undoubtedly compile statistics of high quality. In any case, data are available for all the countries classified as developed; all the nonreporting countries excluding the Soviet countries and Mainland China are among the less developed countries.

Table 1.

Summary of Reported Balance of Payments Statements, 1961-641

(In billions of U.S. dollars)

article image

The figures cover the transactions of some 80 countries and country groups for which statements are shown in the Fund’s Balance of Payments Yearbook, Vols. 16 and 17. Merchandise exports of countries not reporting are estimated roughly to have been $35 billion for 1964, of which about two thirds represented exports from the Soviet countries and Mainland China. Transactions of international institutions and certain international companies are also omitted.

The division between developed and less developed countries is that given in the table on world trade in the Fund’s International Financial Statistics. Balance of payments statements are available for all those countries classified as developed.

For similar reasons, credits and debits as well as net balances are shown. “Merchandise f.o.b.” and “services” are on a truly gross basis. Most other items, however, are reported net by each individual country. Therefore, the credit (debit) entries for those items represent the sum of net credit (debit) entries for those countries reporting net credits (debits), rather than gross flows in each direction.

The categories of transaction shown in Table 1 are confined to those that should in principle be reported symmetrically by both partners (“nonmonetary gold (net)” and “monetary gold” may be considered as together forming a symmetrical gold account). Statistics of individual types of service transaction, which are also symmetrical, are discussed in a later section of this paper (see p. 224).

Adjustments for known discrepancies

A sounder basis for appraising the statistics can be obtained if Table 1 is first adjusted for two known sources of discrepancy that can be quantified with some degree of assurance. These adjustments involve the gold accounts and transactions with international organizations, primarily IMF, IBRD (International Bank for Reconstruction and Development) and affiliates, and UN.

The items “nonmonetary gold (net)” and “monetary gold” in Table 1 appear to show a reasonable correspondence for 1961 and 1963 and a large disparity for 1962 and 1964. The figures for monetary gold are believed to be fairly accurate; at least, they agree exactly with the figures for the same countries shown in International Financial Statistics, which can be accepted as the authoritative source for data on official gold holdings. This being the case, it can be demonstrated that the figures for “nonmonetary gold (net)” must be seriously in error in all years, even in those in which they agree fairly well with the entries for “monetary gold.”

The reported figures for changes in monetary gold holdings of countries are shown in Group 1 of Table 2, and the sources of those changes are given in Group 2. Purchases from the U.S.S.R.10 are estimated figures, while transactions with international organizations—IMF, BIS (Bank for International Settlements), and EF (European Fund)—of course, are known quantities. The only other source of changes in official monetary gold holdings, transactions with the nonmonetary sectors (i.e., nonmonetary gold), can be calculated as a residual.11 It may be seen that the difference between this calculated entry and the sum of the figures actually reported for nonmonetary gold (Group 3) is very large; it appears in the reported balance of payments statements as a debit entry in net errors and omissions.

Table 2.

World Gold Accounts, 1961-641

(In millions of U.S. dollars)

article image

Excluding Soviet countries and Mainland China.

The figures cover a few countries for which full balance of payments statements are not available; they are thus not actually identical with the monetary gold entries as reported in Table 1.

The entries represent gold purchases by the monetary sectors from the non-monetary sections.

These calculations provide the information needed to adjust the figures in Table 1 to show the theoretically correct entries for world gold transactions that would appear in the ideal balance of payments summary for the world. The missing debit entry (or entries) that should be inserted in Table 1 to match the decrease (credit) in Soviet gold holdings cannot be determined, but it will be assumed that it is appropriate to the merchandise account, as discussed below. The entries offsetting changes in the international institutions’ holdings can be ascertained and will be described later. Finally, the changes in holdings for non-Soviet countries not reporting balance of payments data are small (cf. the monetary gold item in Table 1 with Group 1 in Table 2), and their offset will be allowed to fall into net errors and omissions.

For nonmonetary gold, the adjustment needed is a very large debit (see Table 2, Group 4), which conceptually represents unrecorded net consumption, including private hoarding. It seems highly likely that such transactions do in fact occur but escape recording. The necessary adjustment thus amounts to transferring this debit from net errors and omissions to the nonmonetary gold item.

Transactions of IMF and IBRD with their member countries consist mostly of the exchange of one financial item for another. The omission of balance of payments statements for these agencies thus has little net effect on a world summary, for the reasons discussed above. The only transactions outside the capital account are receipts and payments of interest and charges, and administrative receipts and expenditures. These require an entry in the services account, offset in the capital account by an entry reflecting the resulting change in the agencies’ net claims on member countries. The amounts of these transactions are shown in Table 3. The Fund also engages in gold transactions; since monetary gold is separated from other capital items in the table presented here, an offset to the gold transactions must be made in the capital account.12

Transactions of BIS would appear to be similar, and since its net income (which is matched by an increase in its gold holdings) is not large, the omission of BIS transactions from the world summary would not have any appreciable effect. However, the figure labeled “BIS” that enters the calculation of world monetary gold holdings is an adjustment that must be taken into account in the world summary, although it can hardly be construed as part of the statement for BIS itself. The need for this adjustment arises because countries, in their reserves statistics, count gold deposits with BIS as gold rather than foreign exchange. The gold deposited is almost invariably sold by BIS to some other country, so that the same gold appears in the figures of two countries. Since the monetary gold item is intended to refer to physical gold only, changes in gold deposits with BIS must be deducted from the world total of monetary gold transactions derived from country statements and reallocated to foreign exchange (part of “other” capital transactions).

Table 3.

Transactions of International Financial Agencies with Countries, 1961-641

(In millions of U.S. dollars)

article image

Positive figures are credits; negative figures are debits.

Including transactions in which BIS acts as agent for EF.

The omission of UN as a partner has virtually no effect on a world summary, since that agency acts as an intermediary for most transactions. For example, country A donates goods to UN, which that agency distributes to country B; the UN balance of payments would show a merchandise debit and a transfer payment credit with country A, and a merchandise credit and transfer debit with country B. The net entries in each account of the UN statement are therefore nil. The administrative expenditures of UN may well be embodied in the value placed on the assistance it furnishes, the organization makes no profit, and changes in its cash holdings are not very significant. Thus there is no practical need to incorporate a statement for UN in a world summary.

The net asymmetries remaining in the balance of payments statement for the world, after the adjustments for gold transactions and transactions of international organizations described above, are shown in Table 4. The offsetting balances in the gold accounts are those that are in principle to be expected in a world summary. The balances in the other items reflect net errors and omissions. These net figures in all probability comprise gross credits and debits that are larger but partly offsetting. Their magnitude cannot be estimated, but it seems unlikely that these gross errors could be very significant in relation to reported gross flows for the categories, “merchandise f.o.b.” and “services.” For “transfer payments” and “other capital,” on the other hand, even the net asymmetry is large enough in comparison with the recorded figures to be disturbing for certain types of analysis. Some of the known and putative sources of errors and asymmetries in the individual accounts are discussed below.

Table 4.

Net Asymmetries in Adjusted World Balance of Payments Statistics, 1961-641

(In billions of U.S. dollars)

article image

Derived from Table 1, adjusted for gold transactions and transactions of international agencies (Tables 2 and 3), as described in the text. Positive figures show an excess of reported credits; negative figures, an excess of reported debits.

Errors and Asymmetries in Individual Categories


The merchandise trade of countries not reporting balance of payments statistics is quite significant, exports and imports each rising from over $25 billion in 1961 to over $35 billion in 1964. Much of this trade was among Soviet countries and Mainland China and would not give rise to asymmetries in the reported figures presented in this paper. No close estimate can be given of the trade balance of reporting with nonreporting countries, since the balance of payments figures for merchandise are substantially different from the customs data on which such global compilations as the Direction of Trade are based. The balance of payments figures, for example, are sometimes derived from the exchange record rather than the trade returns and have been adjusted for valuation, coverage, and timing; in particular, estimates (which are at times very rough) have been made to adjust imports to a uniform f.o.b. valuation basis. As far as can be judged, the reported figures include an export surplus with nonreporting countries that may be in the order of $0.5-1.5 billion annually on an f.o.b. basis. Some allowance has been made for this factor in Table 4, by assuming that Soviet gold sales to non-Soviet countries excluding Mainland China were in settlement of the trade balance. That is, the credit entry for Soviet gold sales that has been added to monetary gold (see Table 2) has been given its counterpart in a debit adjustment to merchandise.

To adjust those merchandise imports that were reported c.i.f. to an f.o.b. basis for purposes of this paper, the arbitrary assumption (which has long been the rule-of-thumb adopted by balance of payments compilers) has generally been made that freight amounted to 9 per cent of the c.i.f. value. This percentage is, if anything, somewhat high. The reported c.i.f. value of imports to which the adjustment was applied totaled some $30 billion in 1964; imports f.o.b. would thus be understated by $300 million for each percentage point by which the 9 per cent adjustment factor was an overestimate.

The growth of world trade during 1961-64 would also have created a systematic tendency for recorded exports to exceed imports. Merchandise trade should in principle be entered in the balance of payments as of the time that the change in ownership takes place, which—being a single moment in time—results in the simultaneous recording of exports and imports. The trade returns, on the other hand, usually show trade at the time the goods clear customs, so that an import is often recorded perhaps one or even two months later than the corresponding export. Few countries, however, have the necessary information on which to base a timing adjustment for balance of payments purposes. As a result, exports recorded in the balance of payments may be expected to exceed imports during periods of increasing trade.13 To take an example that would approximate the actual situation in 1961 and 1962, assume a steady increase in trade of 0.5 per cent a month (somewhat over 6 per cent annually) and exports of about $90 billion annually; the excess of exports over imports would be about 0.5 per cent of the annual total ($450 million) if there were a one-month lag in recording imports, and 1.0 per cent ($900 million) if the lag were two months. By 1964, the rate of increase had doubled and the export total was $115 billion, so that the reported export excess might have reached $1 billion (or even $2 billion, assuming a two-month lag).


Details of the gross credit and debit entries for individual categories of services, broken down between developed and less developed countries, are given in Table 5. The discrepancies are small in the aggregate, amounting to 5 per cent or less of the gross one-way flow, but they are considerably more important for some individual items. The pattern of the asymmetries strongly suggests that the underlying causes are systematic rather than random. For some items (“other transportation,” “government, n.i.e.,” and “other services”) the amounts remain constant in spite of a substantial year-to-year growth in the gross flows; for others (“freight” and “direct investment income”) they appear to maintain a fairly steady ratio to the increasing gross flows. Either type of behavior could be consistent with the hypothesis that the estimates contain mainly systematic errors.

In the “freight” account, as constructed in Table 5, the credit entries conceptually represent the earnings of the national fleet of each reporting country on international shipments, and the debit entries show the total import freight bill for each country. For some countries compiling their balance of payments statistics on an f.o.b. basis, the earnings of the national fleets on imports have had to be estimated; since these estimates have been added to both the credit and debit sides of the account, errors in the estimates would not introduce any asymmetry into the table. When countries report on a c.i.f. basis, the import freight bill has been estimated. An error in this estimate would result in an asymmetry in the merchandise account offset by a corresponding asymmetry in the “freight” account. If such an error has occurred, it seems likely to have been in a direction that would overstate the freight debits, as discussed above in connection with merchandise.

Another type of asymmetrical reporting arises when imports of nonreporting countries are transported by carriers of reporting countries. In this case the earnings (credit) would show in the table, but the freight bill (debit) would not appear. The converse could also occur but seems less likely in practice to affect the figures, since probably few international carriers are operated by those countries that do not report. (Imports of the Soviet countries and Mainland China on carriers operated by that area would not give rise to an asymmetry, both sides of the transaction being unreported.)

Table 5.

Reported International Transactions in Services, by Category, 1961-641

(In billions of U.S. dollars)

article image

See footnotes to Table 1 for description of coverage and area classification.

Credit entries represent total earnings of carriers that are residents of the reporting countries. Debit entries represent the total import freight bill of the reporting countries. The net entries are thus consistent in concept with an f.o.b. valuation of merchandise.

A further source of asymmetry in the “freight” account—and probably the most important numerically—comes about through the operations of fleets flying flags of convenience, which are not claimed as residents by any of the reporting countries. Merchant vessels registered in Honduras, Liberia, and Panama accounted for 11 per cent in 1961 and 13 per cent in 1964 of the world’s gross registered tonnage. If the earnings of these fleets were to bear those same proportions to the total freight bill, the excess of debits over credits in the freight account would be entirely explained. Although such a close correspondence is likely to be at least partly fortuitous, there can be little doubt that this factor contributes in an important degree to the discrepancy in the freight account.

The excess of debits in the “other transportation” account is more difficult to rationalize. In fact, an excess of credits in this item could have been expected. The missing credit entries in the “freight” account discussed in the previous paragraph must have their counterpart in missing debits in other items, and it could be anticipated that the largest part of them, representing the operating expenses of the fleets flying flags of convenience, would affect the “other transportation” account.

One reason for an excess of transportation debits may be noted. The carriers of reporting countries visit and make expenditures in nonreporting countries, while the converse is probably much less usual. In particular, nonreporting countries include a number of oil producing and refining countries where the reporting countries’ fleets may bunker. However, the strongest probability is that the statistics for this account are unreliable. The Bernstein Report, for example, states: “Some evidence suggests that international transactions in transportation are less reconcilable than most other categories of international transactions among balance of payments statistics of different countries.”14 The Fund staff’s own experience with the figures would lead it to share this view.

In the “direct investment income” item, the treatment of reinvested earnings is known to be a source of excess credits. The figures in Table 5 for this item include net reinvested earning credits reported by the United States amounting to $0.8 billion for 1961, $1.0 billion for 1962, $1.2 billion for 1963, and $1.1 billion for 1964; for all years except 1964, these amounts are larger than the recorded excess credits, and the greater part of them is not reported as debits by the partner countries. The counterpart of well over half of the U.S. credits is definitely known to be lacking, and the proportion may well be substantially larger, since no information is available for many countries on whether their income figures include reinvested earnings.

Another kind of direct investment income debits that is missing is the payments by the nonreporting merchant fleets and other international companies. Substantial income must accrue to the owners of such enterprises, who may be considered as residents by reporting countries. U.S. figures, for example, show earnings of some $0.1 billion annually which are attributed to “international” rather than to any country.

U.S. military expenditures are a likely cause of part of the excess of debits in the “government, n.i.e.” account. U.S. records show military expenditures during 1961-64 of $2.8-3.1 billion annually, only part of which can be identified in the statements of other countries. These receipts are probably more difficult for the partner countries to record than for the United States. For example, some $0.8 billion a year is included in the U.S. figures for personal expenditures of troops, civilian personnel, and their dependents. These purchases are presumably not easy for the countries supplying the goods and services to distinguish separately from similar purchases by their own residents, which would not, of course, be balance of payments transactions; in fact, if it were not for the balance of payments convention under which all government employees are treated as residents of their home country, many of the persons involved might be considered as residents by the countries where they are stationed. In addition, military expenditures of an official, rather than a personal, character may present the recipient country with problems of estimation because of the confidentiality of some of the operations. Some direct evidence of this underreporting is available in the fact that total government receipts recorded by the less developed countries are no larger than the government expenditures in those countries shown by the United States alone.

Transfer payments

By far the largest part of the gross entries for “transfer payments” represents intergovernmental economic aid extended and received in the form of goods, services, and cash. Military grants (which in principle should be defined narrowly to comprise only military goods and services with no alternative civilian end uses) are universally excluded from the Fund’s balance of payments statements on pragmatic grounds, since few countries make available enough details about these transactions to permit them to be entered in the balance of payments on an internationally comparable basis. Private transfers, which cover mainly taxes, pensions, and personal and institutional remittances, are important for a few countries but are small in the aggregate compared with government transfers.

Government grants present difficulties of recording, especially for the recipient countries, for several reasons. (1) As intergovernmental transactions, they may fall outside the scope of the usual reporting requirements that are in effect for private transactions (e.g., the customs or exchange records). (2) Because no payment in any form is required, the value of the goods and services may be difficult for the recipient country to ascertain. (3) Since transfer payments are the counterpart of transactions that lack a quid pro quo, their statistical measurement is limited to a single channel. All other transactions involve two offsetting flows and thus two opportunities for recording them: for example, commercial imports are reflected in both the movement of the goods (customs returns) and the payment for them (exchange records); and the statistician may consult both sources in compiling his data.15 Imports under grants, however, have no counterpart in payments and must be recorded mainly on the basis of physical movements. Adding further to these statistical complications is the fact that wrong figures for the flows of goods, services, or cash will not produce net errors and omissions in a country’s global balance of payments statement, when such flows as are recorded are provided with an exact counterpart in the transfer payments account. The statistician thus not only loses one means for checking his figures but may even be unaware that problems of substantial magnitude exist.

In commenting on the recorded excess of payments by developed countries over receipts by developing countries on account of official donations, the United Nations notes several causes of discrepancy, among them (1) the failure of recipient countries to know the overhead expenditures and local costs of the donor countries, (2) leads and lags in reporting, and (3) nonreporting by many less developed countries.16 Donor countries presumably equate the value of grants with the costs to their governments of providing them. These costs may include, in addition to administrative expenditures in connection with the programs, other elements of domestic costs and even subsidies, as in U.S. grants of surplus agricultural commodities.17 Recipient countries, on the other hand, would be more likely to base their estimates of import values on world market prices or the prices being paid for commercial imports of similar goods.

The Fund staff, as part of its procedure in reviewing balance of payments reports submitted by members prior to publication of the statements by the Fund, routinely compares (where possible) partner country data on government grants extended and received, and occasionally even substitutes figures reported by the donor for those shown by the recipient when there are large, unexplained differences that cannot be reconciled in other ways. Nevertheless, the size of the remaining discrepancy makes it apparent that further improvement is needed in these statistics. The errors or inconsistencies now present may not seriously impede the analysis of individual country statements, since they are not likely to affect the accuracy of any of the measurements of surplus or deficit that are usually employed. They are quite important, however, in the context of burden-sharing exercises and studies of the transfer of resources from developed to developing countries.

Capital transactions (excluding monetary gold)

It is generally taken for granted that the largest capital flows are those (1) among developed countries and (2) from developed to developing countries. The figures on these flows reported by the developed countries are usually assumed to be much more complete and accurate than the corresponding information available from the side of the less developed countries. Therefore, a world summary of reported data might be expected to show an excess of outflows (debits), since the flows among developed countries should cancel out and these countries should have reasonably accurate figures on their outflows to the developing countries, whose own records of the corresponding inflows are believed to be incomplete.

In practice, however, the world summary (Table 4) yields a net excess of inflows on capital account in each of the years 1961-64. When it is further considered that this recorded excess inflow is understated by the amount of reinvested earnings recorded by the United States but not by some partner countries (see discussion above in connection with investment income), the size and year-to-year variability of the discrepancy, as well as its sign, make the available figures seem a rather insecure basis for generalized statements about the behavior of “world capital flows.”

Any attempt to shed light on likely sources of error in capital flow statistics by comparing individual types of capital transaction meets from the outset with theoretical and practical difficulties. Those types that are generally distinguishable are for the most part not symmetrical by definition for the creditor and debtor countries (as previously explained). On the other hand, the content of the exceptional types that conceptually should be symmetrical, such as direct investment, is found to be quite inconsistent in actual practice.18 The disaggregation of the global summary of capital account statistics into individual categories of transaction thus brings into evidence further problems instead of aiding in the investigation of the over-all discrepancy.

A different form of disaggregation, however, does produce a rather surprising result, although its interpretation is ambiguous to say the least. Beginning with data for 1963, IMF and OECD have issued a joint form for balance of payments reporting by countries that are members of both organizations.19 The form calls for a uniform geographic distribution of transactions between OECD members and others, and most OECD countries have attempted to provide estimates on this basis for 1963 and 1964.20 It is thus practicable to construct a balance of payments statement between the OECD area and the rest of the world in three ways and, by comparing the results, to make some additional comparisons bearing on the global asymmetries.

When balance of payments statements for any group of countries (less than the whole world) are added together, the result conceptually is a statement of those countries’ transactions with the rest of the world. Transactions of countries in the group with each other cancel out, except for statistical errors. A statement of transactions between OECD members and the rest of the world, therefore, can be prepared by adding together the statements for all OECD members. Transactions of these countries with each other should cancel out, and the total should thus represent only transactions of all OECD countries with the rest of the world. This total is shown as column 1 in Table 6. Similarly, the sum of transactions reported by all non-OECD countries should represent the transactions of those countries with OECD members; the total derived in this way is given in column 4. Conceptually, the two columns should show entries that are numerically equal but whose signs are opposite. The differing results obtained by basing the calculation on the alternative statistical sources reflect, of course, the errors and asymmetries in the recorded statistics for the world (column 4).

The third way of arriving in principle at the same result is to add up every OECD member’s record of its transactions with all nonmembers, as is done in column 3 of Table 6. A statement constructed in this way comprises figures that have been reported directly rather than obtained as a residual (through the assumed cancellation of transactions between OECD members). The difference between columns 3 and 1 is the result of the failure, for statistical reasons, of OECD members’ figures for transactions with each other to cancel out as they theoretically should. The discrepancies in OECD members’ records of transactions with each other appear in column 2. Considering the gross magnitudes of the transactions underlying them, the net errors and asymmetries brought out in that column are surprisingly small, but their tendency to be largest in the capital account is perhaps significant.

These comparisons, while provocative, are still largely equivocal. The failure of the OECD countries’ statistics of transactions within the area to balance (column 2) may possibly be the result of a failure to allocate a correct global total for capital movements accurately between the OECD area and the rest of the world. Such a distribution is statistically difficult to make, and the approximate nature of the area breakdown has been stressed by some of the compilers. If indeed the discrepancies stem mainly from this source, the figures in column 1 may well be more accurate than those in column 3. On the other hand, the gross volume of capital transactions among OECD countries is undoubtedly many times as large as that between OECD countries and the rest of the world; furthermore, these transactions are presumably freer in general from restrictions, and therefore perhaps more difficult to record accurately, than transactions with countries outside the group. It is thus at least possible that the statistics of intra-OECD transactions are subject to larger errors than those recorded by OECD with the rest of the world.

Table 6.

Adjusted World Balance of Payments Summary, by Area, 1963-641

(In billions of U.S. dollars)

article image

Positive figures are credits; negative figures are debits.

Japan is included with “Rest of World” for 1963 and with “OECD Members” for 1964.

Details may not add to totals because of rounding and because certain adjustments made to the global totals (col. 1) cannot be allocated regionally.

The figures are the sum of cols. 1 and 4 and agree with Table 4, except that the net U.S. reinvestment of earnings (as reported by the United States) in other OECD member countries has been excluded from services (credit) and capital (debit).

All transactions in nonmonetary and monetary gold are included with net errors and omissions, since the allocation of these transactions by area cannot be estimated. Such transactions net out to zero in col. 5 of this table.

In turn, how much significance should be attributed to the fact that columns 3 and 4 agree better than columns 1 and 421 cannot be determined objectively. In view of the known limitations to the statistics, the fairly close resemblance between the former of the two sets of statements must be in part fortuitous. On the other hand, the occurrence of this pattern in two successive years is rather impressive. While the possibility of systematic errors that might happen to be offsetting cannot be ruled out, the results do suggest at least that there may be a fair degree of consistency from year to year in the underlying statistics.


The aggregation of individual countries’ balance of payments statements into group statements or a world total provides clues to certain inconsistencies and asymmetries present in the figures. The method, however, does not necessarily disclose all the differences and inaccuracies that exist, nor does it clearly indicate generally which figures, by country or by category of transaction, are to be regarded as the more reliable when substantial differences come to light. The comparisons presented in this paper, therefore, bring out areas of probable weakness in the statistics but do not throw a great deal of light on the underlying causes.

The fact that the statistics appear unreliable to an extent and in a manner that cannot always be fully assessed may in itself be a conclusion of considerable importance to analysts who are obliged to work with them. Since the statements as a whole and individual series or combinations abstracted from them are used in many ways and for a variety of purposes, generalizations about their accuracy are not likely to be helpful. A known discrepancy that may be of little or no importance in one analytic context may render the series highly suspect for another use. The Fund staff, working with the national compilers to improve the statistics, have concentrated on producing a record of individual countries’ economic transactions with the rest of the world that can be used as a basis for gauging the impact of these transactions on the countries’ domestic economy and foreign position. Most of the available balance of payments statistics may be judged reasonably adequate to serve this purpose. Even for broader studies—such as those of world imbalance, the adjustment process, or the international payments mechanism—their deficiencies are not likely to limit their usefulness seriously. As a component of the national accounts, also, the quality of statistics on foreign transactions derived from the balance of payments appears to be on a par with other series and is not a serious cause of dissatisfaction.

When separate categories drawn from the full statements are employed for international comparison or aggregation, however, the statistical problems may assume much greater significance. For example, two areas in which figures derived from balance of payments statements are coming under increasing scrutiny are in connection with studies of the role played by service transactions and of the pattern and motivation of capital flows. Both of these purposes often require information in more detail and of a higher degree of accuracy than that presently incorporated in the balance of payments. As discussed above, the best data now available are sometimes conflicting or otherwise obviously deficient and thus require cautious handling. Particular care is needed when indirect or derived figures are substituted for reported information or are used to close gaps in such information. The interpretation of developments may be substantially affected by the choice made between alternative data sources and by the assumptions made about the causes of observed discrepancies.

Asymétries et erreurs dans les données statistiques fournies sur la balance des paiements


La mesure dans laquelle on peut se fier aux relevés de balance des paiements est un sujet qui préoccupe à la fois ceux qui établissent ces statistiques et ceux qui les utilisent. Le fait que ces relevés doivent soit comprendre des données compatibles soit révéler explicitement leur incompatibilité par une écriture au poste des erreurs et omissions nettes permet sans doute d’être raisonnablement sur que les relevés de balance de paiements de la plupart des pays fournissent à ceux-ci des renseigne-ments suffisants pour leur permettre d’établir leur politique économique. Toutefois, les défectuosités des statistiques suscitent de plus grandes préoccupations dans le cas de certains autres types d’analyse pour lesquels on utilise ces chiffres. La balance des paiements est, notamment, la source principale des informations nécessaires à l’étude des flux de capitaux internationaux, et les résultats de ces études sont ainsi largement influencés par les hypothéses faites au sujet des causes profondes des incompatibilités et imperfections observées dans les statistiques.

La consolidation, en un total mondial, des relevés de balance des paiements des divers pays, auquel l’auteur procéde dans cet article, fournit quelques indices quant à certaines des incompatibilités et asymétries que présentent ces chiffres, étant donné que, théoriquement, dans un état mondial, toutes les grandes catégories de transactions, à l’exception de l’or non monétaire et de For monétaire, devraient s’équilibrer. Les importants reliquats que l’on observe dans la pratique indiquent des faiblesses probables dans les statistiques, sans toutefois révéler leurs causes profondes. Cependant, le fait que les statistiques manquent de fiabilité—sans que l’on puisse toujours déterminer dans quelle mesure et de quelle façon—est en soi une conclusion d’une importance considérable dont devront tenir compte en particulier les analystes qui désirent établir des comparaisons à l’échelle internationale de diverses catégories extraites des relevés complets.

Asimetrías y errores en los informes estadísticos sobre balanza de pagos enviados al FMI


El grado de exactitud de los estados de balanza de pagos es materia que preocupa tanto a los compiladores de esas estadísticas como a los que las utilizan. El hecho de que cada uno de esos estados tenga que balancear o revelar su falta de equilibrio mediante un asiento explícito por errores u omisiones netos, ofrece probablemente cierta seguridad de que los estados que individualmente presentan la mayoría de los países constituyen una guía adecuada para la formulación de política en los países compiladores. No obstante, las deficiencias de esas estadísticas adquieren mayor relieve en relación con otras clases de análisis en los que se usan esas cifras. En especial, los estudios sobre las corrientes internacionales de capital suelen basarse en la balanza de pagos como principal fuente de datos, y en los resultados de esos estudios influyen considerablemente los supuestos efectuados en cuanto a las causas fundamentales de las discrepancias y deficiencias observadas en las estadísticas.

La consolidación de los estados de balanza de pagos sometidos por cada uno de los países, en un total mundial—que es lo que se hace en el presente estudio—ofrece la clave de algunas de las incongruencias y asimetrías que presentan las cifras, ya que lo ideal sería que, excepto por el oro no monetario y el oro monetario, todas las demás categorías generates de transacciones deberían neutralizarse mutuamente en el estado mundial. Las grandes partidas residuales que surgen en la práctica indican la probable existencia de fallas en ciertos rubros de las estadísticas, pero no arrojan mucha luz sobre las causas fundamentales. Sin embargo, el hecho de que la exactitud de las estadísticas dé lugar a dudas, en un grado y manera que no siempre se pueden aquilatar plenamente, es de por sí una conclusión de importancia considerable, que habrán de tener en cuenta los analistas que quieran realizar comparaciones internacionales entre categorías separadas obtenidas de los estados de balanza de pagos.


Mr. Smith, Assistant Director in charge of the Balance of Payments Division, Research and Statistics Department, is a graduate of Amherst College and did further work at Harvard Graduate School. Before coming to the Fund, he was with the U.S. Government, specializing in balance of payments problems.


The Bernstein Report, for example, concludes (on p. 1) that U.S. balance of payments statistics “are about as good as other U.S. economic statistics collected by similar methods” ([U.S.] Review Committee for Balance of Payments Statistics to the Bureau of the Budget, The Balance of Payments Statistics of the United States, A Review and Appraisal (Washington, 1965), hereinafter cited as the Bernstein Report). Edward M. Bernstein was the Chairman of the Review Committee.


By the same token, only one party to a transaction—the resident—is usually accessible to the national data compilers, a circumstance which limits the choice of methods for data collection, compared with the more usual situation in which both transactors are domestic parties.


The qualifications that should be attached to this statement, omitted here, may be found in the pamphlet by Poul Høst-Madsen, Balance of Payments: Its Meaning and Uses, International Monetary Fund, Pamphlet Series, No. 9 (Washington, 1967).


International Monetary Fund, Balance of Payments Manual, 3rd ed. (Washington, 1961), hereinafter cited as the Fund’s Manual.


This convention is common to the Fund’s Manual and to the United Nations (UN) and Organization for Economic Development (OECD) systems of national accounts.


The percentages in this paragraph are based on the world trade figures given in the Direction of Trade, published by the Fund and the International Bank for Reconstruction and Development.


The Fund’s Manual, pars. 45 and 46.


For an aggregation of the balance of payments statistics on private capital, together with a discussion of the technical problems, see Marcus Diamond, “Trends in the Flow of International Private Capital, 1957-65,” Staff Papers, Vol. XIV (1967), pp. 1-42.


This grouping also was in part selected so that the figures would be useful as basic data for purposes other than that of appraising their reliability.


Through 1964, other Soviet countries and Mainland China were not known to be involved in gold transactions.


The alternative derivation of the balance of payments entry for nonmonetary gold, as the difference between production and consumption (including hoarding), is also shown in Table 2. This derivation is entirely a pro forma one, however, since independent estimates of consumption are not available.


Strictly speaking, some Fund charges are received in gold and thus some gold transactions have their counterpart in the services account. The net effect, however, is as described here, as may be seen from Table 3.


Some countries base their merchandise item in the balance of payments on the exchange record rather than on the trade returns. The foregoing line of reasoning would not necessarily apply to these countries, and the illustrative figures that follow may thus be an overstatement of the case. However, by far the largest part of the merchandise trade recorded in the balance of payments is based on trade returns.


Op. cit., p. 43.


This statement does not contradict footnote 2, which refers to the existence of a single resident transactor; most transactions nonetheless involve two flows, both of which may often be subject to measurement from the resident end.


United Nations, Department of Social and Economic Affairs, International Flow of Long-Term Capital and Official Donations, 1961-1965, E/4170, UN Publication Sales No.: 66. II.D.3 (New York, 1966), p. 37.


The Bernstein Report (p. 53) cites instances where U.S. exports under grants are recorded in the U.S. balance of payments at values that are as much as 40 or 50 per cent above their probable export market value.


When the statistics underlying this paper were being compiled, it seemed logical to keep all official reserve holdings (rather than just monetary gold) and the corresponding liabilities separate from other kinds of capital. The discrepancy between reported official foreign exchange assets and their identifiable liability counterparts, however, was found to be very large and variable. It would seem likely that the amounts of officially held foreign assets of all types and of liabilities to foreign official holders could be established with a fair degree of accuracy by balance of payments compilers, and that changes in these amounts would be entered in the balance of payments. It is not so certain, however, that the assets would all be consistently divided between “reserves” or “official monetary movements” and other types of capital, or that, for the liabilities, the sector of the foreign claimant could be correctly identified by the debtor (let alone whether the creditor would consider a particular claim as “reserves”). If the apparent discrepancy in the statistics between reserve assets and the corresponding liabilities is in fact mainly a problem of identification, the difficulty disappears automatically when a global summary is made, since the transactions will cancel out even though they cannot be separately identified. On the other hand, if amounts for “reserves” are deliberately eliminated from the statements of the creditors and debtors, and these amounts are not symmetrical, the differences will appear to be attributable to nonreserve-type capital. For purposes of this paper, it was felt that a better evaluation of asymmetries in the recording of nonreserve-type capital could be made from statistics including official foreign exchange reserves than from figures from which such reserve movements had been only partly eliminated.


Switzerland, which is not a Fund member, nonetheless reports such information as it has available.


The breakdown was not reported by Spain, Switzerland, and Turkey. The largest part of the merchandise trade of these countries, as shown in the trade statistics, was with other OECD members, and it has seemed safe to assume (when no other information was available) that other transactions, including those on capital account, were also mainly with those countries. In any case, the total transactions of Spain, Switzerland, and Turkey are small compared with those of the aggregate for other OECD members, so that the error arising from this allocation procedure could hardly affect the result significantly.


This statement is directed to the capital item but is also true of all other items (except net errors and omissions).