OBJECTIVES SET FOR THE 1997 CPIS

Because of the perceived deficiency in the reporting of both stocks and flows of portfolio investment assets, major objectives for the 1997 CPIS were to ensure that

• all the major investing countries should undertake benchmark portfolio asset surveys at the same time

• all participating countries should follow BPM5 definitions and classifications

• all participating countries should provide disaggregations, by nonresident issuers’ countries of residence, of national stocks of portfolio investment assets

• all participating countries should use the best methods of survey design and implementation.

It was presumed that, if all major investing countries participated and if there were no major gaps in reporting, the 1997 CPIS would provide a reliable measure of the stock of global portfolio investment claims on nonresidents.

The requirement that participating countries provide information on issuers’ countries of residence was intended to permit, with due regard for confidentiality, the compilation of partner country data for portfolio investment liabilities. (To facilitate this activity, the IMF acted as a clearing house.) The advantage of partner country data on liabilities was that such data would be classified, in accordance with the recommendation in the BPM5, on the basis of the creditor principle; that is, according to the country of the owner of the claim. Despite some difficulties in identifying issuers’ true countries of residency, partner country data on liabilities were considered potentially useful to national compilers because of the perception that more serious deficiencies existed in alternative data sources available on the liability side.¹¹

Achievement of the foregoing objectives was expected to facilitate assessments of

• the reliability of data on the global stock of portfolio investment liabilities to nonresidents

• the reliability of sources of data on both outward and inward portfolio investment transactions

• the reliability of source data on income on portfolio investment assets and liabilities.

This information was also expected to be useful for designing improved methods of collecting data on these transactions.

Objectives for the 1997 CPIS were fully consistent with plans by national compilers to implement the recommendations presented in the BPM5—and in the System of National Accounts 1993(1993 SNA)—for the Financial Account and the IIP. When the 1997 CPIS was conducted, all the major investing countries were in the process of developing annual IIP statements, and about two-thirds of the participating countries had already done so (as evidenced by the Balance of Payments Statistical Yearbook published annually by the IMF). Portfolio investment data contained in these IIP statements were compiled from annual asset and liability surveys or periodic benchmark surveys conducted in conjunction with the use of data models to make, for intervening periods, estimates of stocks from flows. The timing of the 1997 CPIS was therefore a factor in obtaining a coordinated response that maximized the benefits of initiatives already being taken by national compilers.

Having decided that the 1997 CPIS should be focused on the collection of data on portfolio investment assets, the IMF Committee on Balance of Payments Statistics also decided that the 1997 CPIS should concentrate on long-term equity and debt securities (that is, on instruments with original maturities of more than one year) rather than on short-term securities and financial derivatives. This decision reflected the perception that the principal gaps in reporting were likely to be in stocks and flows of long-term instruments. Cross-border activity in such stocks and instruments had been very marked. In addition, the committee recognized that designing a coordinated portfolio investment survey to address short-term instruments and financial derivatives would be much more complicated. Therefore, 1997 CPIS participants were required to report data on long-term portfolio investment assets and also encouraged to report additional data on a voluntary basis. It was acknowledged that the collection of the additional data would increase the complexity of the exercise. Data to be provided on a voluntary basis comprised

• geographic attributions of countries’ holdings of short-term instruments and financial derivatives

• disaggregations of countries’ portfolio investment liabilities (that is, short- and long-term instruments and financial derivatives)

• other useful attributes of data on liabilities (such as attribution by currency of nonresident holdings of securities issued by residents and attribution by the economic sector of the domestic investor).

SCOPE OF 1997 CPIS RESULTS

Twenty-nine countries participated in the 1997 CPIS (see Table 2), and survey coverage encompassed portfolio investments made by more than 4,000 banks; 8,000 nonbank, financial institutions; and 13,000 nonfinancial enterprises. Total reported holdings of portfolio investment assets (equity and long-term debt securities) amounted to nearly US$5.2 trillion at the end of 1997 (see Table 3). The United States, United Kingdom, and Japan (the largest investors) accounted for almost 68 percent of total reported holdings. Holdings of the Kingdom of the Netherlands—Netherlands, Italy, and France each accounted for 4-6 percent of the total; those of Sweden, Ireland, Canada, Bermuda, and Belgium each accounted for 1-3 percent. Sixteen countries also reported cross-border holdings of short-term debt securities amounting to an additional US$100 billion—3.7 percent of the total portfolio assets of these countries. (Short-term debt securities represented large portions of the total portfolio assets of only three countries: Korea, Indonesia, and Venezuela).

Residents in advanced economies had issued about 60 percent of total reported investor holdings of foreign, long-term securities. Emerging market countries—mainly Argentina, Mexico, Brazil, Ghana, Korea, China, and Hong Kong Special Administrative Region (SAR)—had issued a significant fraction (about 12 percent). About 4 percent of the total had been issued through offshore centers (the Cayman Islands, the British Virgin Islands, the Kingdom of the Netherlands—Netherlands Antilles, and Jersey). International organizations had issued about 3 percent. Other countries participating in the survey had issued only 0.2 percent of the total (see Table 4).¹²

AUGMENTATION OF 1997 CPIS DATA

Data from the 1997 CPIS were augmented by supplementary data collected, in conjunction with the survey, from other sources; together, these data are referred to as augmented 1997 CPIS data. Such data were used to prepare more accurate estimates of global aggregates for portfolio investment assets and liabilities and to estimate the global discrepancy between portfolio assets and liabilities.¹³ Augmented 1997 CPIS data were compared with available IIP data and with cumulated balance of payments estimates on portfolio investment assets. The augmented data were then used to adjust the balance of payments estimates and, when possible, the data or estimates of portfolio investment liabilities provided by participating countries.

Sources of the additional data consisted of the Survey of Country Distribution of Long-term Securities Held as Foreign Exchange Reserve Assets (SEFER), which was conducted by the IMF; the Survey of Foreign Equity and Long-term Debt Securities Held by Selected International Organizations, which was also conducted by the IMF and pertained mainly to pension funds of large international organizations; and the Bank for International Settlements (BIS), which provided data (reported on a residency basis) on banks’ holdings of debt securities. The BIS provided data on four countries and economic territories (Germany, Switzerland, Luxembourg, Hong Kong SAR) that did not participate in the 1997 CPIS.¹⁴

Forty-nine countries were asked, through participation in the SEFER, to provide data normally considered highly confidential.¹⁵ Thirty-eight countries representing two-thirds of world foreign exchange reserves agreed to participate. Estimates of foreign securities held by monetary authorities were made for countries that did not participate; these estimates were based on the nonparticipating countries’ published (in International Financial Statistics) holdings of foreign exchange reserves and on the assumption that the instrument composition of reserve assets for the nonparticipating countries was, in aggregate, the same as that for countries reporting information through the SEFER.

The sources of additional data provided geographically detailed information on approximately US$750 billion in holdings of long-term securities issued by nonresidents. These holdings consisted primarily of debt instruments. When the supplementary data were added to data reported in the 1997 CPIS, total reported portfolio asset holdings amounted to US$6.1 trillion.

Because only two-thirds of the 29 countries participating in the 1997 CPIS were, at the time of this survey, compiling IIP statements in which portfolio investment assets and liabilities were identified in accordance with BPM5 definitions, an attempt was made by IMF staff to estimate portfolio investment liabilities for the other one-third. IIP data were also estimated for portfolio investment assets of countries that did not participate in the 1997 CPIS. Augmented 1997 CPIS results were then used to adjust actual and estimated IIP data and to calculate global aggregates for portfolio investment assets and liabilities.

ESTIMATION OF PORTFOLIO INVESTMENT POSITIONS

In the absence of specific collections of relevant position data, the simplest method of estimating portfolio investment positions is—according to the to Balance of Payments Compilation Guide—the unadjusted cumulative flow approach wherein balance of payments data for net transactions in relevant positions are cumulated.¹⁶ Implicit in this approach is the valuation of position data at the prices and exchange rates prevailing at the times of the underlying transactions.

As an alternative, portfolio investment positions can be estimated by cumulating flows at end-of-period prices and applying appropriate price and exchange rate indicators. On the liability side, such indicators could include

• average price indices of shares and bonds traded in domestic markets

• a price index, specifically compiled and weighted by type and by currency, to reflect the composition of internationally traded domestic securities.

Portfolio investment asset positions can also be estimated through the use of price indices that reflect—by type, currency, and country—the composition of foreign securities acquired or sold by domestic investors. Because of the additional detail required, the application (for the purpose of estimating portfolio investment assets) of price indices to data on transactions would, in many cases, entail too great an element of judgment.

To provide a basis for comparison with augmented 1997 CPIS results, the unadjusted cumulative approach was used to estimate asset and liability positions in equity and debt securities for all countries that did not compile IIP statements for the reference date of end-December 1997. Stock and flow information (from the 1970s onward) contained in the IMF database were used in making the estimates. (See the appendix on the IMF database for details.) If not available for the reference date, portfolio investment asset and liability positions were estimated by cumulating flows from the last reference date for which IIP data were available or by simply cumulating flows from the 1970s.¹⁷ Resulting estimates were a mixture of position data (if such data were available from IIP statements), cumulative flow estimates previously included in IIP statements, and the unadjusted cumulative flow estimates described here.

On the basis of the unadjusted cumulative approach and available IIP data, outstanding portfolio investment liabilities consisting of both equity and long-term debt securities were estimated to be about $9 trillion, and outstanding portfolio investment assets were estimated to be about $7 trillion. The difference of $2 trillion represented about 22 percent of total portfolio investment liabilities. Data derived by using this approach are hereafter referred to as unadjusted cumulative data. These data were subsequently compared to estimates, which were based on augmented CPIS results, of counterpart portfolio investment asset and liability positions. The two estimates were compared at global and regional levels (see Table 5). The use of unadjusted cumulative flow data produced an estimate, at the global level, of US$6.9 trillion of total holdings of portfolio investment equity and long-term debt securities; this estimate was comparable to the estimate of US$6.1 trillion produced by using augmented 1997 CPIS data. This outcome suggested that the impact of incomplete coverage by the augmented 1997 CPIS was greater than the downward bias implicit in unadjusted cumulative flow data.

As a second step, unadjusted cumulative flow data for these countries were replaced, on the asset side, with augmented 1997 CPIS data when the value, according to augmented 1997 CPIS data, of the combined holdings of equities and debt securities exceeded the value, according to the unadjusted cumulative flow data, of the combined holdings. Portfolio investment holdings of US$750 billion were thereby identified through surveys conducted (by countries such as Canada, Ireland, Italy, and Spain) to obtain information for the 1997 CPIS and through a new benchmark survey conducted in the United States. These additional holdings were largely attributable to investors resident in European and North American countries. Approximately US$133 billion of the newly identified portfolio investment assets were attributable to Bermuda, the only offshore financial center participating in the 1997 CPIS. The balance of the US$750 billion was attributable to foreign securities held by emerging market countries and by monetary authorities of countries that did not participate in the SEFER.

As a third step, partner country estimates for portfolio investment liabilities (equities and debt securities) were derived, on the basis of geographic attribution, from augmented 1997 CPIS data on portfolio investment assets. Comparisons were then made between partner country data from the augmented 1997 CPIS and portfolio investment liability data derived from unadjusted cumulative flow data. If information provided by the augmented 1997 CPIS on issuers’ countries of residence was reliable, the comparison revealed the identities of countries for which unadjusted cumulative flow estimates of portfolio investment liabilities were manifestly low. For these countries (mainly offshore financial centers and emerging market countries), partner country data (from the augmented 1997 CPIS) on the combined holdings of equities and debt securities were used in preference to unadjusted cumulative flow data. This third step also facilitated the identification of approximately US$500 billion of liabilities. The use of partner country data from the augmented 1997 CPIS to derive portfolio investment liabilities produced, for some industrial countries, monetary totals that were lower than those produced by using available data on portfolio investment liabilities. For the United Kingdom, this outcome was attributed to the nonparticipation of offshore financial centers in the 1997 CPIS.

The twenty-nine countries that participated in the 1997 CPIS reported a total of US$5.2 trillion in holdings of equities and long-term debt securities. By using supplementary information from a variety of sources, estimates of the portfolio investment asset holdings of other countries (and international organizations) were made. The addition of these estimates brought the total to US$7.4 trillion. It therefore appeared that countries participating in the 1997 CPIS accounted for about 70 percent of total holdings. However, the revised total remained incomplete because it was not possible to provide estimates for

• some countries and offshore financial centers for which even cumulative flow estimates could not be made

• estimates of portfolio investment assets held by households. (Some countries that participated in the 1997 CPIS considered lack of household coverage to be a critical weakness in the survey.)

• In addition, the revised total was affected by cumulative flow estimates that were made for nonparticipating countries.

Because the use of available data on portfolio investment liabilities produced a significantly larger total (US$9.4 trillion), it could be assumed that this total would be an appropriate benchmark against which unrecorded portfolio investment assets could be measured. (The implication of such an assumption is that countries participating in the 1997 CPIS accounted for about 55 percent of total holdings.) Or, if the inclusion of a substantial element of cumulative flow data in the estimate of portfolio investment liabilities were emphasized, it could be assumed that the underlying estimate of portfolio investment liabilities would be greater. Because of the scale of portfolio investment thought to be undertaken through offshore financial centers, it was quite probable that flows and stocks of world portfolio investment liabilities were underestimated.¹⁸ A summary presentation of global discrepancies in portfolio investment assets and liabilities for end-December 1997 is shown in Table 6.

As indicated in Table 6, although it was possible to reduce the estimated gap between portfolio investment assets and liabilities by about $300 billion, a very large gap (US$1.7 trillion or about 18 percent of total identified portfolio investment liabilities) remained unreconciled. Because, in the absence of position data, the unadjusted cumulative flow approach—which produces a downward bias—was used more frequently on the liability side, the estimate of the gap between global portfolio investment assets and liabilities was likely to be larger. Conversely, it was possible that the gap was overestimated because of deficiencies, on the liability side, in data on transactions. If the estimated asset/liability ratio for all countries was also applicable to offshore financial centers, a significant portion of the estimated gap would have been attributable to the portfolio holdings of offshore financial centers. If the ratio of Bermuda’s estimated portfolio investment assets to Bermuda’s reported banking assets had been applicable for other offshore financial centers, virtually all of the estimated gap would have been attributable to offshore financial centers.¹⁹ If that were so—and if account were taken of deficiencies in 1997 CPIS coverage of portfolio investments made abroad (directly, through nonresident custodians, or to countries that did not participate in the 1997 CPIS) by households—then world portfolio investment liabilities would have been underestimated.

BILATERAL COMPARISONS OF DATA ON PORTFOLIO INVESTMENT ASSETS AND LIABILITIES

Bilateral asymmetries could be addressed only in regard to eight countries (Australia, Indonesia, Israel, Japan, Malaysia, the Netherlands, Portugal, and Spain) that participated in the 1997 CPIS and reported geographically detailed data on national portfolio investment liabilities.²⁰ In principle, differences between reported data for portfolio investment liabilities and partner country data for portfolio investment liabilities could have arisen for the following reasons:

• Portfolio investment assets reported in the 1997 CPIS were attributed by country on the basis of the debtor/creditor principle; portfolio investment

• liabilities reported in the survey were attributed on the basis of the transactor principle.

• Portfolio investment instruments held as reserve assets were reported, on the asset side, by only a few countries, but such instruments were generally reported on the liability side (see Table 2).

• There may have been inherent deficiencies in data sources used, on the liability side, to determine the residencies of transactors; such deficiencies would have been partly due to the use of nominee accounts.

Although the SEFER was intended to address the latter deficiency, SEFER data were provided to the IMF with the understanding that no individual positions in instruments issued by specific partner countries would be disclosed. The confidentiality of reserve data at the individual holder level limited the bilateral comparability of the survey results—especially for countries issuing debt securities traditionally held by the monetary authorities of other countries. Bilateral discrepancies could also have arisen from 1997 CPIS coverage of direct investment securities, the use of different prices and exchange rates for the in valuation of portfolio investment positions, and inconsistent treatments of repos.

In Chart 2, bilateral differences between total portfolio investment liabilities of the eight countries and corresponding assets reported in the 1997 CPIS by their partner countries are presented in a decreasing order of magnitude (from positive to negative values) and scaled by the total external liabilities of the relevant countries. Positive values indicate that a portion of the liabilities attributed to some partner countries could not be identified by these countries. Negative discrepancies indicate an opposite situation. It is interesting that Japan allocated more liabilities to the UK and Singapore than were identified by these countries, while a negative imbalance was observed vis-à-vis the United States. ²¹ A similar pattern was demonstrated by the Netherlands, for which negative imbalances vis-à-vis Belgium and the UK were mirrored by an opposite gap with respect to the United States, Japan, France, and Italy. Malaysia and Indonesia presented a positive discrepancy with respect to Singapore, the UK, and Belgium. An opposite imbalance was observed for Japan, Korea, and the United States. On the whole, the identified imbalances seemed to indicate that a substantial portion of external liabilities was attributed to intermediary countries in which large international financial markets operated. Therefore, it was probable that these countries did not represent the countries in which the holders resided.²² These results confirmed the view that a reliable disaggregation, by country, of external liabilities could not be prepared from the liability side and underscored the importance of developing more reliable partner country data for the disaggregation, by country, of external assets.

BENEFITS OF THE 1997 CPIS

Results of the 1997 CPIS constituted a rich source of data that could be used for numerous purposes—such as an analysis of portfolio investor behavior or an assessment of the adequacy of countries’ IIP data on portfolio investment liabilities. In addition, 1997 CPIS results provided extremely important information—such as an indication of the scale of portfolio investment conducted through offshore financial centers and evidence that world portfolio investment liabilities may have been underestimated rather than over-estimated as previously assumed. Although the size of the estimated global discrepancy between portfolio investment assets and liabilities remains substantial, a significant number (and monetary amount) of portfolio investment assets were newly identified through the 1997 CPIS. The newly identified assets were largely attributable to investors resident in European and North American countries. These assets were initially reported in new surveys conducted in some countries and in a new benchmark survey conducted in the United States. Bermuda, the only offshore financial center participating in the 1997 CPIS, also made an important contribution to the identification of portfolio investment assets. The use of 1997 CPIS data in conjunction with available and estimated data for countries’ international investment positions also permitted the identification of new portfolio investment liabilities.

The following list covers some other outstanding benefits of the 1997 CPIS.

• The 1997 CPIS demonstrated that an organized effort towards standardizing the scope, coverage, timing, definitions and concepts used in the compilation of data could be successfully coordinated among a large number of countries.

• The 1997 CPIS proved an effective vehicle for establishing and disseminating, on a worldwide basis, appropriately high methodological standards.

• The higher visibility of the coordinated, cross-country approach of the 1997 CPIS resulted in improved budgetary support for the reporting of data on portfolio investment.

• The 1997 CPIS facilitated a greater understanding of country practices with respect to survey design and alternative approaches to data collection and encouraged exchanges of experience in these matters.

• The 1997 CPIS increased the confidence of countries in portfolio investment data.

• The 1997 CPIS facilitated data exchange.

• The 1997 CPIS encouraged awareness of the BPM5 and promoted implementation of the methodology and standards presented therein.

• Although there were some difficulties relating to issuers resident in offshore financial centers and to multinational companies with headquarters located in territories of convenience, for the most part, the 1997 CPIS successfully facilitated the attribution of portfolio investment assets by country.