Italian Experience

114. In Italy, the regulations for the collection of data on cross-border and foreign currency activity are contained in Article 21 of Law No. 148/1988:

Article 21—Information on currency movements for statistical purposes

• (1) The Italian Foreign Exchange Office, for the purposes of statistical information, may require authorized banks, other authorized institutions and, according to directives from the Interministerial Committee for Credit and Savings, operators and other parties concerned, to send data and information (periodically or in samples) on foreign exchange and cross-border transactions in which they have, in whatever capacity, been involved.

• (2) Information and data collected shall be processed for each bank, class of transaction and operator without naming the operators themselves and then stored in archives to which the Bank of Italy may have access insofar as the execution of its duties requires.

• (3) Information and data shall be treated as Official Secrets until they are published; they shall be made available to the Treasury Minister who may use them in the preparation of its six-monthly report to Parliament on currency movements and, upon request, to the Minister of Foreign Trade.

• (4) Statistics produced by aggregating data from at least three subjects and excluding references to a single operator may also be made available by the Italian Foreign Exchange Office to the Central Statistics Institute, to national and international public institutions, to authorized banks and, upon payment, to research organizations and other private sectors.

115. The regulations were extended in May 1990 when a Ministerial Decree, put forward by the Interministerial Committee for Credit and Savings, enabled the Italian authorities to collect transactions data from nonbank operators. In October 1995, another directive from the same Committee enabled the Italian Foreign Exchange Office to collect position data from nonbanks.

116. In preparing for their national survey, the Italian authorities needed to address the issue of confidentiality. Article 21 guarantees the anonymity of the data supplied by operators. Confidentiality is not simply entrusted to a generic “official secrecy” or even to a more specific “statistical secrecy” (i.e., the prohibition of disclosure of data referring to a single person) but, rather, is embodied in the very rules of collecting, processing, and storing the data. Under the second paragraph of Article 21, “the information and data collected shall be processed for each bank, class of transaction and operator without naming the operators themselves....” Furthermore, to meet the legal requirements, the Italian Foreign Exchange Office cancels respondents’ names immediately after the data reported have been validated.

117. For the purposes of the Coordinated Survey, the Italian Foreign Exchange Office will have to choose between two alternatives: to require a legislative amendment to the statistical law for changing the principle of anonymity, or to draw up the national survey on portfolio investment in the same way as required for transactions data.

A. Compiling a Register

120. The sources of information on potential survey respondents are varied, and the work required to compile a register will depend on the extent to which a register already exists in the statistical agency. Box 1 provides guidance on how to compile a register, including the possible sources of information, based on the experience of the Australian Bureau of Statistics. In addition, national compilers should read paragraphs 853 to 862 of the IMF’s Balance of Payments Compilation Guide. Indeed, compilers are advised to consult the whole of chapter XVIII, beginning with page 187.

121. The next step, after the initial examination of the sources of information, is to produce a detailed list of potential survey respondents. The national compiler must be careful to select the “correct” population. To do so, the compiler should engage in some preliminary discussions with the largest institutional investors, such as pension funds, mutual funds, banks, insurance companies, etc., as well as the appropriate trade associations. The discussions should make these institutions aware of the national survey and help the national compiler to decide upon the design of the survey.

122. If national compilers decide to conduct an aggregate survey primarily targeted at end-investors, then they must decide whether to set a minimum threshold that excludes units with low-value investments in nonresident securities. For instance, the number of potential survey respondents identified in the review of sources may be very high, but in practice, a significant proportion of investment in nonresident securities may be undertaken by, or channeled through, a relatively small number of large investors or financial institutions. A minimum threshold may reduce the respondent burden without adversely affecting quality; if need be, the compiler might consider requiring reports from a sample of these “smaller investors.” However, if there is to be a minimum threshold, the national compiler must be careful not to exclude potentially important respondents.

123. If the national compiler decides to conduct a survey primarily targeted at custodians, then it is vital that the custodian understand the relationship between major investors in nonresident securities and custodians in order to avoid omission or duplication of data. This point is emphasized in chapter 2. Appendix VI provides a framework, based on the U.S. experience, for analyzing these relationships. One approach to this issue is to require all large investors to identify their domestic custodians, as well as to indicate any nonresident securities holdings they maintain with nonresident custodians. The relationship between end-investors and custodians could also be kept on file, but updating this could prove time-intensive.

124. Care is also needed in choosing the respondents to approach; an awareness of the operations, record-keeping, and accounting practices of the target businesses is required. The unit from which data are to be collected—that is, the statistical unit—should reflect the level within business structures at which portfolio investment position data are available and are appropriate, given the consolidated reporting obligations of business. Similar businesses will often have different internal arrangements, which necessitates special reporting arrangements (based perhaps on the degree of centralization or autonomy of related businesses). It may be necessary, therefore, to recognize reporting units inside the standard statistical unit. Again, it is recommended that national compilers consult the IMF Balance of Payments Compilation Guide, paragraph 849.

125. The treatment and capture of data from unincorporated entities needs particular care. Because these entities may not prepare full sets of accounts, their investment in nonresident securities on their own account or as an intermediary may be missed or duplicated in the collection.

126. The register will need to be maintained; otherwise, it can become out-of-date in respect of both the entities to be surveyed and the statistical unit to be approached. Media reports are a good and timely source of updates for significant businesses, but a review of the sources referred to in box 1 might be appropriate at discreet intervals, especially close to the time of dispatch of the survey forms.

B. Details Stored on a Register

127. Ideally, any register should contain more than a simple list of names and addresses; it could also (1) include relevant information on each reporting entity, (2) control the mail-out and receipt of survey forms, and (3) provide information on the status of the response. A possible list of items that can be stored, based on the experience of Statistics Canada and the Bank of England, is furnished in box 2.

128. If compilers wish to computerize the register, it should save them time and effort and reduce the scope for error in dealing with survey respondents, when the survey is in operation. (See box 3 for information on computer databases.) However, if a computerized register is to operate efficiently, careful planning and careful maintenance are required; the importance of this cannot be underestimated. The type of issues and tasks that national compilers need to consider include:

• taking account of the information flows to and from the database (the use of charts and diagrams is highly recommended in the planning phase);

• taking account of the specifications for interrogations and reports to be generated;

• devising a coding structure that embodies the defining characteristics of each respondent in its unique identification number. This is essential because it allows the sorting and analysis of respondents and the tracking of nonresponse;

• ensuring that supporting hardware/software is sufficient for the task, for instance, that the memory and processing capabilities are such that response and retrieval times are acceptable;

• allowing time for a thorough tryout of the system before “live” use;

• backing up copies of data and software on a regular basis and storing them both on-site and at a remote location so that the files can be restored in the event of a system breakdown;

• ensuring appropriate systems security and access authorization;

• producing thorough documentation on the system; and

• providing for a suitable filing system for associated papers, not least the completed

C. Use of the Register

129. Once the national survey is under way, the benefits of an efficient computerized register become apparent. The benefits accrue both at the initial stage of sending out survey forms and at the follow-up stage and beyond. Some of these benefits are presented ahead to provide national compilers with some idea of the capabilities of a fully computerized register.

130. At the initial stage, compilers need to print and dispatch the survey forms and letters. To do so, they can generate the mailing labels from the computerized register. At the same time, a record can be written noting which survey respondents will be sent a questionnaire (e.g., mail status to “yes”). Compilers can also perform a quality check to ensure that the right respondent is receiving the right survey form(s). This is particularly important if the mailing includes more than one type of survey form. It is very useful to include with the survey form a postcard (or use some similar technique) that respondents use to acknowledge receipt of the questionnaire and identify the person to contact. In this way, compilers can quickly identify problems and initiate follow-up action.²³ The national compiler should note, in the response log, those entities that have returned the postcard.

131. National compilers should record the return of the completed survey form in the response log, along with the date of receipt; any changes required to the register, for instance, name, address, etc., could be made. Compilers could run a report to verify that the correct entries have been made. Thereafter, they could distribute survey forms to the appropriate person for data capture and editing.

132. Respondents may contact the compiler for a number of reasons. The compiler should keep a record of each phone call and/or correspondence either in a manual file or as a note against the appropriate record in the response log file. If a survey respondent requires more time to complete the survey form, the compiler should note this information so that reminders are not sent. In such a case, the response log status is changed to indicate “in contact,” and follow-up procedures are suspended for a specified period. When new copies of the survey forms are requested, a re-mail is initiated, and additional forms are sent to the respondent with the follow-up procedures kept in place. The compiler must update the register if updated information is provided and run a check to confirm that correct details have been stored.

133. When the closing date for the return of completed survey forms has passed, the national compiler can select the overdue survey respondents from the response log, based on the response status of “outstanding” and the mailing date, and prepare labels for envelopes and follow-up documents.

134. A number of reports can be produced on a regular basis to assess the status of the survey:

• transaction reports: lists of changes sorted by name, ID code, date, etc.

• response log reports: summary counts of survey forms mailed, received, deleted, and percentage outstanding; response log listing all survey respondents, survey status sorted by name, ID code, etc.

A. Relational Databases

138. Relational databases store data in a set of linked two-dimensional tables; within a table, each column represents a distinct attribute, and each row corresponds to a distinct set of observations of the attributes. The idea is to hold the minimum amount of duplicate data: instead of repeating the same data in a table, duplicate data will be placed in a separate table linked to the first. Box 4 defines and exemplifies relational databases.

139. On one hand, a large relational database management system (RDMS) is commendable for processing surveys: it can handle massive amounts of information, have the capacity to operate on many platforms (e.g., DOS, OS2, UNIX), and serve many clients at the same time. Most large RDMSs provide libraries, which, for example, allow the compiler to develop an array of common routines and procedures. On the other hand, depending on the volume of data, the use of a personal relational database (PRD) might be sufficient to process, analyze, and generate the survey output for publication purposes.

140. The RDMSs provide better performance when holding large amounts of data, tend to have better “roll-back” facilities to safeguard against loss of data following system crashes, and have more sophisticated security features. However, the performance gap between PRDs and RDMSs is narrowing. Indeed, PRDs and RDMSs tend to be used in parallel, with the PRD forming the “front-end” to the RDMS. Data are transferred between the two using a “Structured Query Language.”

Box 4.

Relational Databases

Formally, a relational database is a collection of “relations,” whereby a relation is a two-dimensional table in which the entries in the table are single-valued; each column has a distinct name (called an attribute name); all the values in a column are values of the same attribute (namely, the attribute identified by the column name); the order of the columns is immaterial; each row is distinct; and the order of the rows is immaterial.

The following table illustrates a relational database, wherein the middle “block” represents the higher-order database; that is, information from the upper and lower “blocks” feeds into the middle “block.”

Holder reference	Holder		Address	Address of custodian
A	Mutual fund
B	Pension fund
↓
Holder reference	Security reference		Value
A	1		70,000
B	2		40,000
	↑
Issuer	Security reference	Year	Country	Security	Maturity
Government	1	1996	USA	Bond	Jan 99
Corporation	2	1996	Japan	Stock	None

View Table

Holder reference	Holder		Address	Address of custodian
A	Mutual fund
B	Pension fund
↓
Holder reference	Security reference		Value
A	1		70,000
B	2		40,000
	↑
Issuer	Security reference	Year	Country	Security	Maturity
Government	1	1996	USA	Bond	Jan 99
Corporation	2	1996	Japan	Stock	None

B. Time-Series Databases

141. A one-time survey fits the relational database well. However, if the survey is to be repeated regularly, the national compiler would need a system that stores time series data. Relational databases cannot store time series efficiently. Nonetheless, data can be transferred from a relational database to a time-series database.

142. Time series (or temporal) database management systems are “time intelligent”: they incorporate routines that “understand” time. For example, they can turn data on a monthly frequency into data on a quarterly frequency, seasonally adjust data, etc. The choice of a commercial time-series database is extremely limited and often has to be created in-house for the needs of the client.

A. Quality Control Checks Through the Survey Form

148. The sooner errors are spotted and corrected, the better. For this reason, it is recommended that national compilers consider devising a survey form that explicitly includes quality control checks and/or requires extra information, which can be used as a consistency check. However, if extra information is required, it should be kept to the minimum necessary, and the national compiler should be clear as to the purpose for its collection.

Aggregate survey forms

151. On an aggregate survey form, a number of checks are possible. In the Australian system, as positions and transactions data are gathered on the same survey form, the reconciliation is done directly by the reporter. (More details are provided in box 5). But this is a system where collection of position data is ongoing. On the other side of the coin, the national compiler of a first-time national survey might consider requesting that survey respondents provide the value of total security assets held or simply total assets held, against which the reported nonresident holdings could be compared. This would help identify severe misreporting, specifically, reporting all security assets as nonresident.

Box 5.

Australia’s Experience with Quality Control Through the Reporting Form

In Australia, the foreign investment reporting forms are structured so that there is a full reconciliation, for each item, between the opening and closing levels of investment, and transactions during the period. An example, for equity instruments, is shown below. Respondents are also asked to specify the reasons behind the data in an “other factors” column in a separate comment box on the form. In the Australian experience, this approach both educates respondents about the relationships between individual data items and forces them to consider the consistency of the data reported.

Equity Investment Abroad

Equity Investment Abroad

Position at market value at start of period	Purchases or sales (net) during period	Exchange rate variations during period	Other factors contributing to changes during the period (e.g., market valuation changes)	Position at market value at end of period
(1)	(2)	(3)	(4)	(5)

View Table

Equity Investment Abroad

Position at market value at start of period	Purchases or sales (net) during period	Exchange rate variations during period	Other factors contributing to changes during the period (e.g., market valuation changes)	Position at market value at end of period
(1)	(2)	(3)	(4)	(5)

The ABS has found that full data on transactions and other reasons for changes in stocks are usually available when survey respondents provide data on their own accounts or client activity—for example, pension funds investing and managing their own assets, investment (fund) managers managing assets on behalf of other entities, and trading entities managing their own portfolios. These entities will collect all relevant stocks and transactions data at market prices.

Within the ABS, individual analysts are responsible for confirming the reported data—either by form type or institution type. They produce editorial notes that identify and explain all large movements at the aggregate level by reference to the entities contributing significantly to those movements. The computer system has already edited and checked the arithmetic and compared large changes in levels with the “reconciliation” data at the individual respondent level. All large amounts (approximately A$100 million or more), reported in the transactions, exchange rate variations, and “other factors” boxes, are validated with respondents.

Independent sources—for instance, published price indices or exchange rate movements over the period—are also used to judge whether the reported data are accurate. For example, if a 20 percent change in the value of a U.K. asset in Australian dollar terms was explained by a negative entry in the “exchange rate variations” column—mplying a decline in the value of the U.K. pound relative to the Australian dollar—and it was known that the actual bilateral rate had not moved in that direction or by that approximate magnitude over the period, then the ABS would question the respondent. Information from share price indices may prove a useful check on entries in the “other factors” box, for equity.

Anecdotal evidence, such as that derived from market or media comment or contact with other data providers, is also used to confirm data. For example, if there is evidence that Australian enterprises in general were selling securities in a specific market but that a particular respondent reported a large increase in assets in that market, owing to positive net transactions, the respondent would be questioned. Comparisons might also be made with the general magnitude, type, and direction of data reported by similar entities to see if they were behaving similarly and providing consistent explanations for movements.

Australia has good experience with checks of this type. Although the basis for estimating exchange rate effects differs among individual enterprises, reported valuation effects are generally in line with independently sourced market information.

152. It is recommended that simple computer-checking procedures be written to process reported aggregate data. For instance, a computer check could be written that sums the individual country data and compares it with the reported total holdings; the difference should be zero. Likewise, a check could be written that highlights zeros or data above a certain size, so that any unusual figures can immediately be brought to the national compilers’ attention. There are many other similar routine checks that can be devised.

153. More generally, the national compiler could require an official of the survey respondent to certify that the information provided is complete and accurate; this could help ensure the data quality and punctuality. Similarly, requesting the name of a contact person helps ensure that follow-up enquiries are dealt with efficiently.

B. Quality Control Checks Against Other Reported Data

154. Different possibilities exist for checking data against other reported data—especially for assets of financial intermediaries like banks or investment companies—both at the level of the individual reporter and the aggregated level. These financial intermediaries might separately report statistics on portfolio investment position data for supervisory and/or statistical requirements. To give some examples:

C. Quality Control Through Analytical Checks

162. The degree to which analytical checks can be developed depends on the availability of comparable data. If data are available, the following checks can be devised:

• If stock data and flow data are independently assembled, then an attempt should be made to reconcile these data. To do so requires consideration of effects caused by different prices, exchange rates, other adjustments like write-offs, or thresholds. The method by which this type of analytical check can be developed is focused on in appendix VIII; the check can be applied at the individual respondent and aggregate data levels.

• In countries where the position data will be produced for the first time for a number of years, but transactions data are available, an attempt could be made at reconciling position data, estimated by accumulating transaction data with reported position data. The method of calculating position data from transactions data is illuminated in paragraphs 740 to 743 of the Balance of Payments Compilation Guide. The degree of deviation of the cumulated transactions and reported position data could depend on the composition of the holdings, the variability of the relevant exchange rates and market prices of securities, and the length of time over which transactions have been cumulated. However, if the national compiler judges the discrepancy as being significant, it should be investigated. Given the number of assumptions and potential timeframe of observations, this check is probably more appropriate for aggregate data, although it could be developed for checking individual reporters’ data. The more frequent the surveys of positions, the more reliable the reconciliation.

163. One important piece of information required to develop these analytical checks is the currency composition of the portfolio assets, without which the currency and security price revaluations are, at best, an estimate. Therefore, the national compiler might ask survey respondents to provide a breakdown of investments by currency—although this is not a mandatory requirement of the Coordinated Survey. As a second best option, the national compiler might use data on the currency composition of holdings provided by different financial intermediaries (e.g., banks) as a proxy for the currency composition of all survey respondents’ holdings.

164. Some compilers will face problems, particularly when checking data for claims vis-à-vis individual countries, if their transactions data are based on the “transactor” principle, as opposed to the debtor/creditor principle used for position data (see BPM5 paragraphs 484 to 488). In other words, it might not be possible to reconcile cumulated transactions data with the position data from the national survey because different principles of geographic allocation are being used.

D. Quality Control from the Viewpoint of the Recipient (Debtor) Country

165. Because comparable data are to be exchanged between participating countries, national compilers shall receive data on their residents’ securitized liabilities to nonresidents. While quality control is primarily the responsibility of the national compiler in the creditor country, the recipient compiler could be in the position to check the data received. Checks could be carried out on data vis-à-vis a single counterpart (creditor) country and/or on data vis-à-vis a “group of reporting countries.”

166. Regarding individual counterpart countries, the recipient compiler (debtor country) could check the received data by:

• comparing the data with that previously published. At present, several countries (e.g., the Netherlands) publish data on their investment in nonresident securities, which can be used as an approximated range for the position data at end-December 1997;

• using information about interest payments vis-à-vis single counterpart countries. The country attribution of the outstanding volume of domestic securities could be estimated. However, care is needed with this check, because the residency of recipients of interest payments on bearer securities may itself be estimated; and

• using data published in the BIS statistics on banks’ investments in nonresident securities. These data give a partial picture of nonresident investment in domestic securities. If information is available on the banks’ percentage share of total holdings of nonresident securities in the creditor economies, then from the partial BIS data, some estimate of the country attribution of the nonresident investment in domestic securities could be made.

167. The recipient compiler should be aware that each of the aforementioned estimation methods is only a broad check. Nonetheless, these estimates would provide the recipient compiler with a benchmark against which to judge the data received. For instance, Austria has calculated that, at the end of 1994, approximately 55 to 65 percent of the outstanding volume of domestic securities are held by nonresident investors within the European Union member states. This estimate is based on a mixture of the estimation methods mentioned above.

168. It is also possible to devise checks on the aggregate data received:

• Nonresident investment in domestic securities could be estimated by residual, that is, the outstanding volume of equity and long-term debt securities issued by residents, adjusted for residents’ investments and compared against the aggregate data received. Data on the outstanding volume of domestic securities issued could be available from domestic or international publications (see appendix VII). Residents’ holdings of resident securities may be published by the domestic statistical agency or, at least, be available for the banking sector from published banking statistics.

• The analytical check of comparing cumulated transactions data with the reported holdings data (see paragraph 162 above) could also be developed for liabilities data.