External debt statistics that are compiled by the IBRD and OECD provide a potentially useful source of data on debt flows for developing countries, but there are several problems in reconciling them with one another and with balance of payments concepts. Raw comparisons between these debt statistics and Yearbook data suggest that balance of payments borrowing flows are understated. The Working Party could not obtain enough information on the differences, however, to propose specific adjustments to balance of payments data.

National balance of payments compilers in developing countries clearly should be more intimate with the details of debt statistics compiled for their individual countries, because the debt data bases are broad, detailed, and perhaps informative.


Long-term loans are a separate subdivision of the “other capital” category, and Part 2 of the Balance of Payments Statistics Yearbook devotes line entries to gross flows of loan credit. For each of the three sectors in the “other” group—resident official, banks, and other—where underlying data permit, the flows are shown as extensions and repayments of loans separately for assets and liabilities. The loan group could be an important area for comparisons among countries and among sources of statistical data, since loans have been a major concern of international financial policy.

There are major problems in lining up data for relevant comparisons, however, and reconciliation of statistics on lending and borrowing requires detailed inquiries into national statistical methods across many countries. Some countries report loan flows incompletely, even on a net basis, without the extensions and repayments data that appear in the Yearbook. The United States, for example, gives only resident official asset flows and does not show separately lending detail for banks or private nonbanks; it has no separate borrowing figures at all for long-term loans. The United Kingdom also has no separate entry for banks’ long-term lending or borrowing.

Countries are also inconsistent in the timing of entries for loan transactions. Most important for the 1980s, countries have different treatments in their balance of payments accounts for debt delinquencies, reschedulings, and write-downs. Debt-servicing problems give rise to an extraordinary variety of accounting practices to describe a single situation or a single rescheduling agreement. For example, national statistics are affected by a bank’s choice between directly writing down assets for a bad debt and increasing bad-debt reserves while keeping the gross loan in its asset total. Statistical treatment of such events is affected by national accounting practices and regulations.

These are questions of consistency between lender and borrower treatments of specific transactions in balance of payments accounts. The questions do not directly concern correctness, that is, whether the entries conform to balance of payments accounting standards or to the recording of other domestic transactions. Correctness is more difficult to measure than consistency, but both data qualities can be studied only through bilateral studies at a detailed level. In many cases the answers lie inside the books of individual lenders, beyond the reach of statistical inquiries.

International Debt Statistics

An alternative approach to judging the quality of loan data in balance of payments statements lies in the international debt statistics that are compiled by the World Bank and the Organization for Economic Cooperation and Development. The IBRD and OECD statistical systems measure borrowing only by developing countries, and the two systems include nothing about credit flows within the industrial world. For developing countries, the two systems complement one another in that the OECD measures borrowing mainly on the basis of creditor reports, while the IBRD uses reports from borrowers.103 Each has an extensive data base that carries bilateral information on positions, service payments, and reschedulings.

Consistency between the IBRD and OECD systems is substantial but not yet complete. The International Working Group on External Debt Statistics continues to work toward an understanding of the differences between the systems and toward a full integration of the data bases.104 An integrated form of the two data bases would provide a rich source of consistent bilateral transactions data that could be a foundation for balance of payments accounting with regard to debt.

Even without integration, the debt figures can be useful to balance of payments compilers. Most national data sources for debt are independent of balance of payments statistics on borrowing. These sources can perhaps shed light on the completeness of the balance of payments numbers or at least raise questions of consistency. Debt statistics have absorbed a great deal of professional effort over the past decade, and the detailed data bases developed for them afford an external perspective on the quality of balance of payments statements.

The IBRD and OECD debt figures nevertheless differ conceptually from the balance of payments structure that is presented in the fourth edition of the Balance of Payments Manual. The IBRD system includes only loans in foreign currencies, for example, and it does not recognize as a transaction a resident’s repurchase of a country’s foreign-held debt. When a debt cancellation occurs, the IBRD labels it as a stock adjustment, while the Manual calls for a repayment (outflow) entry matched by a transfer inflow in the current account. When debt servicing falls behind schedule, balance of payments accounting shifts the arrearage from long-term to short-term capital. Debt rescheduling is also treated differently in the Yearbook and IBRD data. All these differences mean that the two forms of data can present a single event differently in terms of timing and value. Again, only very detailed inquiries can explain the differences.105

Long-Term Debt

Simple comparisons between the IBRD and balance of payments measurements of net long-term borrowing are presented for 13 countries in Table 38. The first part of the table presents five countries that have particularly large differences. For the five, the IBRD data show a $7 billion average of net borrowing for 1986–89, while balance of payments figures give a reduction of $4 billion. For all years in the table the IBRD data show larger inflows than do the balance of payments, but for all five countries the differences are largest in 1987 and 1988. For the eight countries in the second part of the table, the discrepancies show a wide diversity. Balance of payments amounts for the eight show much larger borrowings in 1986, but in later years and on average the differences are small.

Table 38.

Drawings Net of Repayments in the Form of Long-Term Credit, 13 Countries, 1986–89

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

article image
Sources: IBRD data are taken from the World Debt Tables, 1990–91. The amounts used in these comparisons consist of net flows of public, publicly guaranteed, and private nonguaranteed long-term debt, excluding portfolio debt issues and direct investment inflows. To enhance consistency with balance of payments concepts, they include the amounts that the IBRD reports as the excess of disbursements from rescheduled debt over the amounts of principal and debt stocks rescheduled.Balance of payments amounts are from the 1990 Balance of Payments Statistics Yearbook. They are total net flows of long-term liabilities in “other capital” of resident official, deposit money banks, and other sectors.

In a search to explain the differences in Table 38, the Working Party asked balance of payments compilers in each of the 13 countries to reconcile balance of payments entries, to any extent possible, with the amounts reported to the IBRD. An important problem here is that, in most countries, the government offices that send debt statistics to the IBRD are quite separate from the offices that compile balance of payments accounts. For public or publicly guaranteed debt, at least, the debt figures often are compiled, loan by loan, at a micro level. The data sources for these tabulations can be very different from those that balance of payments compilers routinely use. There may be little interchange between the two compiling offices, rendering a reconciliation of the numbers impossible. It requires a detailed and costly audit of statistical processing through each of the two systems, and most governments are unwilling to undertake such a task.

Responses to the Working Party’s inquiries provided very little information that could be used to reconcile the figures. Some respondents said that they were aware of the discrepancies, and they mentioned the conceptual differences between the two data sets. A few specifically pointed to differences in IBRD and Fund procedures for converting into U.S. dollars the amounts denominated in other currencies. They also said that balance of payments data reported to the Fund and debt data sent to the IBRD are often at different levels of revision or completeness. China, India, and the Philippines provided detailed information about some differences, but they could not give substantial reconciliations.

Debt-restructuring agreements are a special problem in balance of payments accounting for the 1980s. Procedures for recording such agreements are different between the Manual and the IBRD debt statistics. The Manual specifies that a rescheduling should appear as a retirement of the rescheduled debt and a new borrowing to indicate new debt. The IBRD compiles and publishes data on reschedulings, but the net flow numbers in the World Debt Tables do not include or reflect them.106

To learn something about the balance of payments treatment of reschedulings, the Working Party inquired specifically into four major debt-restructuring agreements for two of the countries with large differences in Table 38: for Brazil, the 1988 commercial bank and the 1987 and 1988 Paris Club agreements; for Mexico, the 1987 commercial bank agreement. Table 39 summarizes the agreements. The Working Party sent inquiries to national compilers of debt and balance of payments statistics in Brazil and Mexico and to their six main creditor countries.107 Each country was asked to review the IBRD data, to indicate whether or not the transactions under these agreements were included in their countries’ balance of payments, and, if so, to indicate how and in which components these transactions were recorded.

Table 39.

Debt-Restructuring Agreements with Brazil and Mexico, 1987–89

(In billions of U.S. dollars)

article image
Source: IBRD

Compilers in Mexico replied that published balance of payments included net entries for repayments on rescheduled loans and drawings on new loans. On the creditor side, France responded that the flows associated with debt restructuring were not recorded at all in its balance of payments, because the flows would not be in line with that country’s balance of payments concepts. The other creditor countries responded that, as a rule, restructurings were not separately identified in their reporting systems. The United States suggested, however, that it was possible to identify at least some of the U.S. transactions.

Short-Term Debt

Table 40 gives comparisons in net changes in short-term debt for a few countries in a form parallel to the long-term debt comparisons in Table 38.108 The IBRD receives direct reports on short-term debt from a number of countries. Comparisons of IBRD short-term debt statistics with balance of payments flows published in the Yearbook cannot be as specific as those of long-term debt. IBRD short-term data are published only as stocks stated in U.S. dollars and do not include the explicit information on flows that the long-term data include. As a result, flows can be deduced only as first differences in the U.S. dollar positions. As a method to measure flows, first-differencing has many problems that arise from discontinuities in the data and changes in valuation. The IBRD amounts in the table are weak as a result. For the balance of payments side, the table includes total short-term liability flows, which can go well beyond the range of IBRD coverage of debts.

Table 40.

Net Borrowings in the Form of Short-Term Debt, 1986–89

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

article image
Sources: IBRD amounts are taken from World Debt Tables, 1990–91, “net short-term debt.” Flows are first differences in stocks as stated in U.S. dollars. Balance of payments amounts are total short-term liabilities shown in “other capital” in the 1990 Yearbook.

The distinction between long and short term also is blurred in these comparisons. Balance of payments treatment of debt defaults transfers some debts from the long-term to the short-term category, and the maturity break is poorly reported for some of the major countries in financial markets. These considerations suggest that a more useful comparison might combine short and long term together in both IBRD and balance of payments data. There are a number of problems with such a combination, however. Notably, the short-term reporting is thin in the IBRD statistics, and the debt flows must be inferred from position changes.109 (The short-term data base needs further development.) For some countries that are in both Tables 38 and 40, there are slight offsets between differences in long-term and short-term flows. For others, notably Mexico, the differences in both tables are positive and large, and for these countries offsets between maturities evidently cannot explain the discrepancies.

The short-term comparison has problems, but the IBRD data show an almost consistently larger volume of debt growth than do the balance of payments flows, even though the latter cover a broader set of claims. Some of the difference may reflect changing exchange rates for the dollar, but the pattern implies other factors, particularly for Mexico.

Conclusions and Recommendations

The comparisons in Tables 38 and 40 between balance of payments and international debt statistics are broadly consistent with the relationships discussed in Chapter 6 in connection with bank borrowing. There is a large overlap between the two sets of results, but the comparison in this chapter of debt flows with IBRD data has a wider reach. The IBRD statistics include more types of foreign lenders, such as governments and international organizations, and more types of borrowers, mainly public agencies and domestic banks.

Although the debt comparisons call into question the balance of payments entries for a number of developing countries, the Working Party does not propose any adjustments based on the IBRD debt statistics to Yearbook capital flow data.110 Responses to the Working Party’s inquiries did not provide a basis for proposing specific data improvements. Clearly, the conceptual differences between IBRD and balance of payments data cause some of the discrepancies. Nevertheless, the replies from countries suggest strongly that balance of payments compilers are not closely examining the debt statistics as a possible source of information.

Indeed, the replies received by the Working Party from major borrowing countries indicated poor communication between balance of payments compiling offices and debt-compiling offices. In none of the countries did the statistical office provide reconciliations between external debt data published by the IBRD and balance of payments data published by the Fund. The replies did not mention efforts to reconcile the two sets of data. On the creditor side, the OECD countries that the Working Party approached also failed to give useful details on the relation of creditor reports in debt statistics to balance of payments data. There evidently can be much more cooperation between statistical offices than is now maintained.

This experience leads to several recommendations in the area of international debt statistics. As expressed at the broadest level, the Working Party recommends that compilers develop regular contacts with related agencies in their own countries and with international organizations that hold data bases with statistics related to the balance of payments. More specifically: