Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
THE ROAD TO A MODERN SURVEY SYSTEM: STRATEGIC CHOICES AND FIRST EXPERIENCES 1 Pim Claassen, Bank of the Netherlands, Netherlands 1 INTRODUCTION In many countries, balance of payments compilers are currently in the process of reconsidering their collection strategy. In some countries, this may take the form of an evolutionary process; in other countries that of a radical reform. As in many European countries, the former Dutch (closed) settlement system originated from just after World War II, when foreign exchange regulations were needed to safeguard sufficient levels of foreign reserves. In those circumstances the balance of payments could be considered a free of charge ‘by product’ of the system of foreign exchange regulations. However, with the liberalisation of imports, exports and financial flows (also within the European Union) the balance of payments became a statistic in its own. While the costs of this statistic thus became more explicit, compilers also increasingly faced problems measuring the continuously expanding cross border financial flows. Against this background the Netherlands Bank (DNB) decided, in early 2000, to reform its settlement system in a direct reporting system, which became operational in May 2003. This decision not only aimed at increasing the quality of the balance of payments but, at least as important, also at lowering the reporting burden to society. A new division of tasks between DNB and Statistics Netherlands was agreed, whereby the latter took over the statistics on international trade in services. 2 Part I of this paper heavily draws on the Statistical Bulletin that DNB published in May 2003 . This special issue, among others, provides a full description of the strategic choices for the new system as well as a set of reporting forms and explanatory notes. Part II of this paper reports on the first experiences with direct reporting. 2 2.1 PART I: STRATEGIC CHOICES The reconciliation model DNB made a clear choice for an integrated reporting model with both flows and stocks, including a full reconciliation between them. This preference was reinforced by the prospect that the frequency of the IIP was going to be raised from annual to quarterly. Although there is no formal international obligation to present the reconciliation of flows and stocks as separate data sets, it is not unlikely (for instance in the context of the revision of BPM 5) that this will be the case to allow the quality of data to be tested. The final decision to implement the so-called ‘reconciliation model’ in the Netherlands however was not taken until consultations with external auditors and a representative group of reporting entities confirmed 1 The author, Head of the Balance of Payments & Financial Accounts Department, would like to thank his colleagues for their valuable comments in preparing this paper. 2 Balance of payments in the Netherlands ‘The road to a modern survey system’. De Nederlandsche Bank, Statistical Bulletin, Special Issue, May 2003 (Internet: http://www.dnb.nl) SI/2004/01868/cla its feasibility, notably because it largely corresponds to the data available in the business administration. In other countries with survey systems, like New Zealand, Australia, Ireland and Finland, similar reporting schemes had already been put into operation successfully. As an example, table 1, shows the reconciliation model for the reporting of loans. Table 1 Example of the reconciliation model for loans Eur thousand Asset Type of asset Country of the foreign debtor Changing during the month Position at the beginning of the month Transactions Other changes Position at the end of the month 0 -10 96 Rec Rec/Bop IIP Revaluation Increase asset Decrease asset Exchange rate changes Other price changes 100 35 25 -4 IIP Bop Bop Rec Loans, original maturity > 1 year, to non-banks -Loans US -Loans GB -Etcetera Purposes IIP/B oP/Recl 1 Bop is balance of payments; rec is reconciliation 2.2 Security-by-security reporting In the reconciliation model, reporting agents can choose between security-by-security reporting and security reporting on an aggregated level, i.e. sum totals broken down by type of security, and the country of residence and the sector of the issuer. Security-by-security reporting is based on ISIN codes. Security-by-security reporting implies a large data flow between the reporting entities and the balance-of-payments compiler and a complex compilation process. A prerequisite for processing the collected security-by-security data, furthermore, is a database with updated information on a large number of securities issued worldwide. DNB concluded that the advantages for both the reporting agents and the compiler in terms of the reporting burden and data quality far outweigh the costs. Reporting agents using ISIN codes do not have to classify their portfolio by type of security, country and sector. This classification is generally difficult to provide because it does not suit the business administration. Also they can refrain from reporting price and currency changes and accrued interest, since DNB derives this information from its security database. Finally, reporting by ISIN codes may shield entities from the consequences of new data requirements, insofar as these can be met through the adaptation of DNB’s security database. The security database developed by DNB, DEBBI, contains information on about 450,000 securities, identifiable by their ISIN codes. DNB buys this information from a commercial data provider. The database is not only an important source for the balance-of-payments, but also for the so-called issue market statistics. 2.3 On-line quality checks A close relation of the reporting model for the financial account to the data available in the business administration was an important prerequisite for the new system. It not only reduces the reporting burden, but also enhances data quality which in principle will be equivalent to the quality of the business accounts of the reporting agents. The full reconciliation included in the reporting model also allows for two simple but significant checks of the consistency of the data. Firstly, for any specific item the sum total of all reported changes must be equal to the difference between the opening and closing positions. And secondly, each opening position must be equal to its closing position in the previous reporting period. For data transmission, DNB developed an internet application (e-Line Bop, see below) that carries out both checks on-line. If the results of the tests are negative, transmission is not possible. Since so-called ‘logical errors’ are filtered out, DNB’s efforts to check the data are confined to plausibility checks. The larger effort is focused on the plausibility of the macro data, which may of course facilitate the detection of outliers at the micro level. 2.4 Procedure for non-response Since the beginning of 2000, DNB has an effective legal instrument to keep non-response limited through its authority to impose administrative fines and fees and/or cease and desist orders under penalty. The availability of this instrument made it easier for DNB to decide upon the transition to a survey system. Without it, a significant increase of non-response compared to the settlement system would have been possible. After all, in the settlement system data were largely collected through banks, which generally display more reliable reporting than non-banks. Nevertheless, significant non-response was expected to occur, notably in the first months after the transition to the new system. In the new system estimates for non-response are generated automatically at the start of the production run of each reporting month. These estimates take place at the micro level, i.e. for each missing report a single ‘imputation’ is processed, based on the historical data of the reporting agent involved. When the actual reports are received, the imputations are replaced automatically. The quality of the generated output thus steadily improves as more reports are received and added to the database. 2.5 Cooperation with Statistics Netherlands In a survey system, a variety of sources and grossing up procedures are used, which in itself make it more difficult to trace the origin of implausible macro results. Plausibility checks therefore are based more than before on the extent to which the balance of payments fits into the overall statistical framework of the national accounts compiled by Statistics Netherlands. This requires close cooperation between both institutions. The process of matching data is facilitated by the fact that the new DNB reporting forms comprise transactions, not settlements, which corresponds with the concept used in the national accounts. Also the sector classifications used by both institutions have been geared to one another. To that end, it is necessary to periodically attune register information for the financial account with the general business register of Statistics Netherlands. This register contains over a million legal entities established in the Netherlands and is updated daily with amongst other things information from the Chambers of Commerce. This procedure is aimed at preserving identical sector classifications of single reporting agents, so that distortions of the macro data on these grounds can be prevented. Attuning the registers also enables the comparison of the micro 3 data collected by both institutions from single reporting agents. In general, the overall cooperation with Statistics Netherlands contributes to greater harmonisation of the balance of payments and the national accounts. 2.6 E-Line Bop Data transmission from the reporting agents to DNB takes place over the internet by means of e-Line Betalingsbalans, further referred to as e-Line Bop. This application delivers data to each reporting agent regarding its register data held by DNB, its ‘profile’ determining its reporting obligations and an overview of the reports already transmitted and still due. For each data transmission, the reporting agent must contact the DNB server to start the session and approach the electronic ‘reporting forms’. This procedure also warrants that new requirements, resulting e.g. from international regulations, are implemented simultaneously by all agents. As mentioned before, data transmission to DNB is only possible after the reporting agent has successfully carried out consistency checks of its data. The installation of e-Line Bop requires that the reporting agent downloads a plug-in, a socalled Java-runtime environment (a worldwide standard for which the software is freely available), to his IT environment in order to use the Java-applet provided by DNB. The applet provides for – protected – communication with DNB. To be able to log in on the DNB server, the user must identify himself through his registration number and application code, after which the authenticity of his connection can be checked. The data can be entered manually or electronically, through the import of files in CSV or XML format. Actual transmission takes place through an encrypted so-called SSL tunnel which warrants a high level of security against interception by third parties. By accepting the user agreement reporting agents also acknowledge that their reporting forms are filled out truthfully. 2.7 Uniformity of the new system DNB has aimed for a high level of uniformity in the collection system. Nevertheless specific circumstances had to be taken into account that affects the ability of different types of reporting entities to report the data. For instance the business administration of special financial institutions (SFIs) in the Netherlands usually does not allow for monthly (or even quarterly) reporting of stock data. In such cases, adjustments for were made to the basic reconciliation model for monthly reporting to meet these limitations. Also the requirements for different types of reporting agents can differ because they engage in specific activities. In total, DNB distinguishes 14 separate reporting profiles, each of which comprises the requirements for a specific economic (sub) sector. From a technical point of view, however the reporting profiles are simple variations of the reconciliation model, which warrants a high level of transparency in the elaboration of the collected information. As a result, the automation system for processing the data has remained relatively transparent, too, so that any future adjustments will be less costly to implement. 3 To enable the data exchange with adequate confidentiality guarantees, adaptations of the legal framework in which the statistical tasks of DNB and Statistics Netherlands are laid down, had to be effected. The uniformity of the reporting profiles may also be of benefit to the reporting population, notably those using standardised software for their business administration. The market offers a wide variety of such packages generally referred to as ERP software, short for Enterprise Resource Programming. The close connection of the reporting profiles to the business ledgers in principle provides the possibility to incorporate the reporting requirements into future releases of these packages. To encourage this process, DNB informed ERP producers, amongst others, through presentations and bilateral consultations. 2.8 Frequency of data collection The uniformity in the system also prevails for the frequency of the reporting requirements. In principle all data for the financial account are collected monthly and for international trade in services quarterly. The distinction in the international requirements as regards the frequencies at which different levels of detail have to be provided therefore is not reflected in the new reporting system. This decision was based on the consultations with external parties for whom it appeared inefficient to process diverse data sets at different frequencies. Instead they prefer to preserve one single routine in which all information is processed simultaneously, even if this implies that more detail is provided at a certain frequency than actually required. An exception to this principle had to be made for data that cannot be processed monthly, because they are simply not available. This notably concerns specific items of direct investment and real estate. Profits and losses, market valuation and dividends paid and retained profits are generally assessed, and hence collected, on an annual basis only. The monthly requirements for these direct investment items are therefore confined to available data, notably sales, purchases and dividends having become payable. Similarly for real estate, monthly requirements only concern purchases, sales and net income; the remaining data are collected annually. To accommodate reporting agents further these data can be reported at the end of their financial year. DNB collected information from each reporting entity on the 4 actual month in which their financial year ends. E-Line Bop uses this information to automatically request the yearly data reports in the month in which they have become due. The stock data obtained through the yearly reports automatically replace the stocks previously estimated for the months comprising the entity’s financial year. These estimates are obtained through the accumulation of the reported flows. Similarly, the reinvested earnings included in the annual report are distributed automatically over the twelve months concerned. 2.9 Reporting by banks Data collection from banks is more diverse. For their reporting on portfolio investment and derivatives, the reconciliation model is used. This also applies to banks’ intermediary reporting concerning their customers’ activities in these fields. Banks’ own direct and other investment on the other hand have both been integrated in existing bank reports, notably for money and banking statistics and prudential supervision. An important step to enable this was taken in the 1990s when money and banking statistics were harmonised in preparation for monetary union. One of the measures taken at the time was the introduction of a geographic breakdown of all EU countries, necessary for the consolidation of national money and banking statistics to the level of the euro area. In consultations with the Netherlands Bankers’ 4 It appeared that 20% of the reporting agents has a financial year that differs from a calendar year. Association, it was decided to use the opportunity to adopt a full geographic breakdown vis-àvis all countries of the world to achieve the integration of balance-of-payments data. 2.10 Actual transition The preparations for the new system included a two-year transitional period in which the sampled enterprises were informed extensively about the new reporting requirements. Serious thought was given to incorporating a period of double-reporting during the transitional period. However, due to the fact this would create an additional burden both for reporting agents and for DNB itself, a middle course was adopted. For the large majority of the reporting agents, the transition occurred as a “big bang”, i.e. all at the same date without double-reporting. Concurrently around 85 top reporters, accounting for up to 40% of flows and stocks of direct investment, were offered the opportunity to adopt the new reporting system at an earlier stage. These corporations were supported intensively. After a trial period in which the quality of the new reports was assessed, they could make the definitive switch to the new system. At that moment, their reporting for the settlement system was discontinued. About half these enterprises either made a successful early transition, or at least transmitted approved testreports. To avoid double-counting in the settlement system, arrangements with the banking sector were made at that time to ‘neutralise’ the international settlements of top reporters that had made the early transition. To that end a special neutral code was adopted. DNB also had to adapt its IT infrastructure for the settlement system, still in full operation, to be able to process the data of the top reporters. This adjustment entailed that the reports by the top enterprises were automatically translated into ‘settlement reports’. A relatively minor effort thus enabled simultaneous reporting according to the old and the new collection model. This may be a workable approach also for countries applying settlement systems, if at any time multinationals will be allowed to switch to the so-called multinational reporting model. 2.11 Centralised reporting The sampling process for the new survey was drawn from a population including all economic entities from which data for the financial account were collected in the settlement system. In many cases two or more of the entities selected in the sample appeared to be part of a single conglomerate. If these corporations maintain a centralised business administration for all their group companies located in the Netherlands, it may be inefficient for them to report separate data for each single agent in the sample. Therefore DNB offers the possibility of so5 called centralised reporting. Corporations can decide for themselves whether a centralised report covers all resident group companies, or only part of them, so long as the entities actually selected are included. To be eligible for a centralised report, corporations must however meet the following conditions: • Only resident group companies can be included. 5 As the facility to use centralised reporting has been used by many conglomerates the number of originally selected 2,800 reporting agents (see next paragraph) has been further condensed to about 2,100 reporting agents. As many of these conglomerates, because of their centralised business administration, include all of their group companies in their reporting the actual number of resident companies covered by direct reporting numbers about 4,700. As a result, the actual coverage of the sample might be slightly higher overall than envisaged. • The foreign assets and liabilities of the included companies must be covered for 100%, also if it concerns participations of less than 100%. • To prevent statistical distortions at the macro level, all companies included in a centralised report must belong to the same economic sector. If this is not the case separate reports for each sector have to be produced. 2.12 Cutting-of-the-tail & grossing up procedure For the benefit of the sampling process, DNB compiled a database containing the resident population active in the field of the financial account. This ‘extended register’, comprising around 43,000 domestic entities, contained not only register information, e.g. names, addresses, but also quantitative data on the size of the external transactions and positions of the entities included. The information was to a large extent derived from information available from the settlement system. 6 In addition, public information for the financial sector was used. From the extended register a sample of about 2,800 reporting agents was drawn. The sampling process was straightforward, following the so-called cutting-of the-tail principle: for each of the major items of the financial account those reporting agents were selected that account for around 95% of the total transactions and positions of the population. The results confirm the efficiency that can be achieved by direct reporting: a sample of 6.5% was sufficient to achieve a 95% coverage for the financial account. Although this coverage comes close to full coverage, some procedure for grossing up the reported data was considered necessary to ensure that the macro data give a complete picture of the economic and financial developments. 2.13 Updating the register and benchmarking The sampling process described above will have to be repeated periodically to reassess the coverage and, if necessary, adjust the grossing up ratios, presently set at 5%. Periodic benchmark surveys on the cross-border stocks and flows of the entities not included in the present sample, i.e. the tail, were considered as possible options. A prerequisite for a representative benchmark survey is of course a comprehensive register of the population relevant for the financial account. At that time, DNB envisaged four major sources to keep the present register up to date. Large new direct investments can generally be traced by consulting the financial press. In some cases, these transactions may be so substantial that the resident entity involved will have to be added to the sample immediately. A second source is the socalled ‘register tapes’ a group of large banks will provide to Statistics Netherlands on a monthly or quarterly basis (depending on their own preference). These tapes comprise simple sum totals of all cross-border payments per customer that exceed a threshold of EUR 12,500. DNB receives an overview from Statistics Netherlands comprising the customers responsible for the largest aggregated payments. This information may lead to the identification of enterprises that have newly become relevant for the financial account. Also the periodic attuning of micro data with Statistics Netherlands can result in the detection of new entities relevant for the DNB register. The fourth source for preserving the accuracy of the DNB 6 Due to restrictions on the public accessibility of these sources, DNB in two cases had to carry out small benchmark surveys to obtain the quantitative data needed for the sampling process, notably for investment companies and pension funds. register concerns SFIs. For statistical purposes these institutions are legally obliged to register at DNB upon their establishment in the Netherlands. Their registration is facilitated by the fact that many SFI are administered by trust companies well aware of this obligation. 3 PART II: FIRST EXPERIENCES WITH DIRECT REPORTING The second part of this paper describes the first experiences with the new direct system. More in particular, it intends to assess whether the strategic choices described in part I have worked out well in practice and whether the primary targets of increasing quality and lowering the reporting burden to society have indeed been achieved. The conclusion is that experiences have been very positive and sometimes even above expectations. 3.1 Response rate Already from the start the response rate has been higher than expected. Moreover, the response rate has been improving throughout the period the new collection system is operational. Chart 1 clearly illustrates this continuous improvement. Chart 1: Response rate At the moment, almost 80% of reports have already been received after 15 working days, i.e. the official reporting deadline. This percentage increases to almost 90% at the time the monthly data are processed for dissemination. Towards the end of a quarter, when the quarterly data for the previous quarter are disseminated, the average monthly coverage has increased further to about 98%. In the former system the correspondent response rate for reporting agents’ monthly reports on foreign bank accounts was, at least, 30% lower. Chart 2: the response rate for the different reporting profiles (at end-June 2004). Chart 2 Response rate per profile Profile 2003-Q2 2003-Q3 2003-Q4 2004-Q1 NFV 99,96% 99,63% 98,90% 98,34% OFI 100,00% 100,00% 100,00% 98,20% BLG 100,00% 100,00% 100,00% 100,00% VRZ 100,00% 98,93% 98,39% 98,34% PNS 100,00% 100,00% 100,00% 100,00% BFI 99,10% 99,03% 99,13% 97,97% MFI 100,00% 100,00% 100,00% 99,10% BWB 100,00% 100,00% 100,00% 100,00% CLM 100,00% 100,00% 100,00% 100,00% CSD 100,00% 100,00% 100,00% 100,00% OVH 100,00% 100,00% 100,00% 88,89% SLB 100,00% 100,00% 100,00% 100,00% SLN 100,00% 100,00% 100,00% 100,00% DNB 100,00% 100,00% 100,00% 100,00% Total 99,59% 99,40% 99,15% 98,31% The fact that the initial response rate was high undoubtedly reflects the efforts to inform reporting agents well ahead of the implementation of the new system. All new reporting agents had been extensively informed well in advance in writing about their new reporting obligations. Also, DNB account managers had already been allocated to reporting agents in early January 2003, which enabled intensive and personal support to reporters. In order to help remind reporting agents to their reporting obligations DNB, at the request of reporters, also launched a monthly e-mail signalling service. This e-mail is sent out at the first day the reporting for a new month can start. In addition, DNB has a strict policy of sending out formal (written) reminders. These formal reminders are sent out only few days after the official reporting deadline has passed. In October 2003, after allowing for a certain transition period, DNB informed all reporting agents that from that moment on it would actively use its authority to impose administrative sanctions. In practice this means that if reporting agents do not react properly to these formal reminders, DNB initiates a sanctioning procedure. In the last quarter of 2003, DNB sent out 43 cease and desist orders under penalty to reporting agents. Most of them reacted promptly, but some had to pay up to EUR 11,000 each. 3.2 Use of e-Line BOP Only very few reporting agents turned out not be able to use e-Line Bop. These reporters either had no access to internet or had very strict security policies within their company that did not allow them to download the Java-applet from DNB into their IT environment. As a consequence, at the moment over 97% of the reporting agents reports through internet. In May 2004, DNB held a survey under its reporting agents in order to evaluate their use of eLine Bop. This survey gave, amongst others, information on the platform configurations used by reporters, the use of the test mode of e-line, the sources of the reported data and the actual use of e-Line. When introducing e-Line, DNB had restricted its technical support to a limited number of platform configurations. The survey showed that some of the combinations actively supported by DNB are not used anymore in practice. A vast majority of reporters currently uses the processing systems Windows 2000 Prof and Windows XP in combination with internet browsers IE 5.5 and IE 6.0. The survey also showed that over half of the reporting agents had actively used the test mode of e-Line in the implementation phase. This test-mode has been made available to them as of January 2003. Large reporting agents, such as banks and pension funds, even indicated that they still use this test mode, for instance if they make changes in their reporting system. Another finding is that over 95% of the reporters confirm that their general ledger has become the main source for their reports, whereas in the old settlement system most information originated from treasury systems. This finding confirms that the BOP reporting within companies has shifted to other departments, underlining the importance of finding new contact persons. The survey also revealed that 92% of the reporters entered the data manually into e-Line BOP and only 8% used dedicated software to have an automated import in CSV or XML format. The latter finding might reflect the enormous reduction in the number of data actually reported on a monthly basis by, in particular, non-financial corporations and special financial institutions, as these reporting agents do not have to report individual payments anymore. 3.3 Security by security reporting DNB’s decision, in early 2000, to opt for security-by-security reporting in a direct reporting system was, at that time, unique and meant that the new system had to be expanded to include a complicated security-by-security processing subsystem. The data needed for processing are acquired commercially, adding to the day-to-day costs of the system. In practice, the securityby-security processing subsystem has indeed proven difficult to build. The expertise needed to define all the requirements concern the data needed from the commercial data provider, the structure of the securities database and the algorithms and formulae for the processing. It has therefore been necessary already to deliver a number of successive releases in order to refine the first version of this subsystem according to experience gained since the start of production in May 2003. The adoption rate of security-by-security reporting has been much higher than expected, even though the choice between security-by-security reporting (condensed) and aggregate reporting (fully reconciled) is left to the reporter himself. At the moment, all reporting agents whose profile allows for security-by-security reporting (mfi’s, pension funds, insurance companies, investment companies and custodians) also use this facility. As a result, the share of portfolio investment reported security-by-security would already seem to be much higher than 85%, i.e. the percentage recently agreed upon within the Euro system (e.g. at the moment the ‘coverage’ for foreign debt securities is 98%). In cases where reporters do not use security-bysecurity reporting by ISIN code, it is for practical reasons rather than out of principle. For instance, as the ISIN code is the international standard (ISO 6166), it was decided not to support any of the other codes (e.g. CUSIP and SEDOL). Investment managers based in the UK and the USA, used by Dutch pension funds, have not (yet) accepted the ISIN code and provide data on an aggregate basis only to these pension funds. Close relations with these managers have been forged so that the processing they perform produces aggregate data of adequate quality. The quality of portfolio investment data has improved considerably. Misinterpretations of the reporting guidelines have been discovered with ease. Most misallocations by country of issuer, type of security but also between portfolio investment and financial derivatives could be rectified within a few months after the start of the new system. In the past, these mistakes could not have been singled out so easily. The flexibility of the system has also been proven in preparing for the ECB’s requirement to provide data by currency of denomination for debt securities. The system has, however, turned out to be relatively expensive. Therefore, it is welcome that the costs in the near future can be spread over more statistical products, such as national financial accounts. Also any cost savings that might result from the centralised securities database of the ECB are highly desirable. Although the coverage of securities included in DNB’s own database has steadily been improving since May 2003, the centralised securities database of the ECB could provide further quality improvements in this respect. 3.4 Cooperation with Statistics Netherlands The cooperation with Statistics Netherlands has become much more intensive. It managed to further streamline its process of delivering monthly trade data. DNB now has the trade data available before the monthly BOP is submitted to the ECB. Both institutions also agreed on a timetable to deliver the data on trade in international services to DNB. For the first monthly submissions to the ECB, so called ‘first shots’, DNB uses estimates based on earlier quarterly data. Statistics Netherlands, however, provides the quarterly data on international services to DNB before the quarterly BOP data are due to the ECB. This is done together with data for some other current and capital account items (e.g. compensation of employees and other transfers). The scope of cooperation with Statistics Netherlands has broadened even further. Building on the success of DNB’s direct reporting, its foreseen role as prudential supervisor on pension funds and insurance companies and the division of tasks already agreed on for the BOP, DNB and Statistics Netherlands agreed upon a new division of tasks with respect to the financial accounts. DNB will start collecting and compiling the national financial accounts for all financial sectors, including households. In addition, both institutions agreed that for those sectors DNB would also start collecting current account information needed for the quarterly sector accounts and information on services needed for the BOP. From the discussions both institutions have had so far with the representative organisations on the new reporting framework it is clear that financial institutions see benefits in having their reporting obligations, both supervisory and macro-economic reports, taken care of by one institution (DNB) and as much as possible integrated in a few reports. It has also become clear that the process of compiling financial accounts can benefit very much from security by security reporting. 3.5 Updating register and benchmarking Since the start of the direct reporting system DNB has further developed its strategy with respect to benchmarking. For reporters in the financial sector DNB has decided it will use as much as possible existing supervisory registers and data. This information allows DNB to keep its BOP-register and reporting population up to date and, if necessary, also provides the basis for grossing up purposes. However, DNB introduced annual benchmark reports for all mfi’s, in particular to provide information on their cross border holding of securities and their ‘other investment’ assets and liabilities. Also SFIs, for which no supervisory data are available, deserved special treatment. For this sector DNB also initiated a benchmark survey. On the basis of the results of this survey, sent out early 2004 to over 8000 special financial institutions, DNB will be able to update its register, the reporting population and gross up the results of the sample. For non-financial companies DNB and Statistics Netherlands have entered into an exchange of data to closely attune the reporting population. In this context the general business register of the latter plays an important role. To illustrate the close cooperation also in this field it can be mentioned that the Statistics Netherlands in its new quarterly survey on business statistics does not intend to have non-financial companies to distinguish foreign and domestic transactions and positions: for the foreign part it intends to rely on BOP-data. Except for abovementioned, more structural arrangements, DNB also uses other sources to spot new large players. There are, for instance, strict procedures to follow up any news on mergers and acquisitions in the financial press. Also DNB recently received from Statistics Netherlands for the first time so-called ‘register tapes’ (explained in part I). 3.6 Quality In the paragraph about security-by-security reporting it has already been mentioned that the quality of the data for portfolio investment are felt to have improved considerably. This also holds for the data in the other sub-accounts of the BOP. Some of these improvements result directly from the new reporting system; other improvements are more indirect, but still very noticeable. Against this background, it is not surprising, that the errors & omissions in the new system have so far been within a narrower band than in the previous system (see chart 3). The improvements resulting directly from the new reporting system are mostly related to the use of a full reconciliation between flows and stocks (for most instruments, including accruals, on a monthly basis) in the reporting forms. As noticed in a previous paragraph, most of the data directly originates from the reporter’s general ledger and is therefore likely to be of good quality. Moreover, reports also have to be internally consistent, something that can not be achieved in a settlement system. At the macro-level there is the possibility of a monthly reconciliation of flows and stocks, which in the former system (where flows came in on a 7 monthly basis and stocks, to some extent, were only reported on a yearly basis ) could only take place at an annual basis. The new collection system thus also provides a much better basis for the quarterly IIP and external debt statistics. The other, more indirect, improvements result from the fact that it seems much easier to control a direct reporting system than a settlement system. One illustrative example is that the department’s account managers in the new system all have a clear responsibility for the timeliness and quality of the reports of a limited number of reporting agents (in the former system they had to deal with and code individual transactions of ten thousands of reporters). The new system is also much more transparent. For instance, outliers can be easily traced and dealt with in a context in which the reporter has submitted a full report. Moreover, the new system in general would appear to allow for more timely management information. One of the results is that the reporting discipline in the new system is much higher than in the former system. The latter could be achieved because it is much easier to have equal treatment to reporters (which is indeed essential if you decide to sanction reporting agents not fulfilling their obligations) in the new system. The fact that in the new collection system different reporting profiles are distinguished, has also made it easier for DNB to organize an active feedback of aggregated data at a sector level to the several groups of reporting agents (e.g. pension funds, custodians, insurance companies and trust companies) that initially had been set up by DNB to discuss the new reporting framework. This feedback, in which DNB presents the aggregated to the sector, has already proven to be very useful, also in terms of checking the plausibility of some observed developments. These sectoral analyses could be further strengthened on the basis of the information DNB is going to collect for the national financial accounts. One of the few problems that initially caused problems with respect to errors & omissions were the continuously expanding activities of so called Special Purpose Vehicles (SPVs). Many of these SPVs, which issue very large amounts of debt securities in the international market for securisation purposes, were set up recently and were at that time not registered as direct reporters (causing omissions). Their issues could fortunately be traced in the security issue statistics, but it is clear that, because of the amounts involved, better coverage of their business is warranted. 3.7 Costs to society The reduction of the reporting burden to society has been an important aim of the change in the reporting system. In this respect, it should be noticed that in the Netherlands the reduction in the reporting burden in the meantime also has become a political issue, since the government has committed itself to a lowering of the administrative burden by a minimum of 25% in four years time. In this context the Ministry of Finance, early 2003, ‘zero-measured’ the estimated costs to society related to the balance of payments. These costs, applying to the former closed settlement system, were then estimated at EUR 75 million per year. 7 In the former settlement system there were yearly reports for stocks of direct investment (equity), portfolio investment and derivatives. For other instruments, in particular intercompany loans and other investment (of other sectors) stocks had been estimated by cumulating flows. For those instruments improved collection in the new system has caused breaks in time series. In the survey DNB held in May 2004, reporters were also asked for the time they actually spent on the initial installation of e-Line BOP and the initial retrieval of the data to be reported to DNB. Together those two activities took reporters on average 23 hours, equivalent to about EUR 3,6 million in total one-off costs to society. The survey also gave information on the time reporters spend on fulfilling their regular reporting BOP-obligations. This turned out to be 3,25 hours per month, almost regardless of the reporting profile of the reporting agent. Based on this outcome the running costs to society for the BOP-reports to DNB could be estimated at EUR 6,8 million per year, indicating large cost savings as compared to the 8 former settlement system. In making this comparison some (specific Dutch) features of the former system should be kept in mind. For instance, in the former system all reporting agents had to provide the information themselves on the economic nature of the transaction for each single payment above a threshold, both on the payments made through domestic banks as well as on payments settled through foreign banks. 8 The costs to society of the new survey for international services should be added to this in order to come to fully comparable data. Although these costs are as yet not known, they will change very little to this favourable outcome.