Executive summary

This annual report provides a quality review of the national balance of payments (b.o.p.), international investment position (i.i.p.) and international reserves template of the Eurosystem (international reserves), as well as the associated euro area aggregates.^[1] The report fulfils the formal requirement for the Executive Board of the European Central Bank (ECB) to inform its Governing Council of the quality of these statistics, as set out in Article 6(1) of Guideline ECB/2011/23 (hereinafter the “ECB Guideline”).^[2] Furthermore, the report provides information supporting the Macroeconomic Imbalance Procedure (MIP) data quality assurance process, as laid down in the “Memorandum of Understanding between Eurostat and the European Central Bank/Directorate General Statistics on the quality assurance of statistics underlying the Macroeconomic Imbalance Procedure” (“the MoU”).

The main principles and elements guiding the production of ECB statistics are set out in the ECB Statistics Quality Framework (SQF)^[3] and quality assurance procedures, which are published on the ECB’s website. This report therefore provides a quality analysis of the statistical output, covering the elements of: (i) methodological soundness; (ii) timeliness; (iii) reliability and stability; (iv) internal consistency (completeness and validation, and net errors and omissions); and (v) external consistency/coherence with other comparable statistical domains (euro area accounts, foreign trade in goods statistics, Monetary Financial Institutions (MFI) balance sheet items, money market funds, investment funds and securities holdings statistics).

The descriptive and quantitative indicators used throughout this report are based on monthly data from September 2016 to June 2019 (unless otherwise indicated) and on quarterly data from the third quarter of 2016 until the second quarter of 2019 (unless otherwise indicated). Data and revisions published up to 23 October 2019 are included. Supporting tables/charts and details of how the indicators are computed can be found in the respective annexes to this report.

Given the specific MIP requirements and the responsibilities entrusted to the ECB in the context of the MoU, the box at the end of the report presents some indicators relating to the fitness for purpose of the data for all EU countries. The box draws on annual data up to 2018 and revisions up to 2017 and focuses on the following quality dimensions: (i) data availability; (ii) revisions; (iii) errors and omissions; and (iv) external consistency with sector accounts.

Statistical developments between 2018 and 2019

In general, euro area countries have broadly implemented the sixth edition of the International Monetary Fund’s (IMF’s) Balance of Payments and International Investment Position Manual (BPM6) and the respective ECB data requirements. This has enabled national compilers (national central banks (NCBs) and national statistical institutes (NSIs)) to both report relevant data and make them publicly available with sufficient accuracy and within the agreed deadlines. Nevertheless, some additional efforts are still needed to disseminate more quality data and improve comparability and consistency with other datasets.

While statistical standards are generally observed, there is still room for improvement in terms of methodological soundness. Luxembourg, the Netherlands and Malta^[4] are encouraged to continue working to increase the coverage and quality of data on special purpose entities (SPEs). While major progress was observed in 2019, Cyprus is still invited to closely monitor the SPE “sector” and continue improving the counterpart – geographical – detail. Greece should start reporting data for financial intermediation services indirectly measured (FISIM). Germany and France should improve their reporting of financial derivatives positions/transactions for general government. Regarding foreign direct investment, some countries should make an extra effort to correctly report transactions – and associated positions – in debt securities and trade credits between companies engaged in foreign direct investment relationships in the appropriate functional category. The Netherlands should correctly identify transactions and positions between fellow enterprises, particularly as regards debt instruments^[5]. Furthermore, national compilers should in general continue their efforts to improve the coverage of assets held abroad by resident households.

The majority of countries have complied on a continuous basis with the deadlines for data transmission, with a few exceptions. Countries that have failed to comply should put in place contingency measures to ensure that such situations do not reoccur. In terms of data availability, Malta should take the necessary steps to start reporting complete monthly and quarterly datasets for “other flows” as soon as possible^[6].

Regarding accuracy and reliability, most countries record regular revisions that do not fundamentally change the economic assessment of first vintages. However, it is also worth noting that 19 countries have implemented major national accounts and b.o.p./i.i.p. benchmark revisions in 2019, which supported the alignment of national accounts (ESA2010) with b.o.p./i.i.p. data. Moreover, countries are encouraged to continue regularly providing the ECB with information on major events and revisions (by means of the “metadata template”) in order to increase both transparency and the analytical value of the data for policy use.

Concerning internal consistency, a large majority of countries provide fully consistent data to the ECB. Austria has recently eliminated inconsistencies between monthly and quarterly data. Ireland has also improved the consistency of monthly and quarterly data, but at the expense of the reliability of the revised monthly data. Belgium and Malta should ensure that quarterly positions and flows are reconciled for all periods.

With regard to consistency/coherence with other datasets, overall b.o.p./i.i.p. data are in line with other datasets, thus ensuring comparability across statistical domains. However, it is of utmost importance that all countries follow the agreed steps to ensure full consistency vis-à-vis sectoral accounts. Regarding other datasets, the ECB encourages b.o.p./i.i.p. colleagues to interact with their counterparts to structurally reduce discrepancies and/or to reconcile and document differences between datasets where there are objective methodological differences.

Within the European System of Central Banks (ESCB), Working Groups on Financial Accounts (WG FA) and the Working Group on External Statistics (WG ES), along with other sub-structures of the Statistics Committee (STC), e.g. the Working Group on Monetary and Financial Statistics (WG MFS), are working closely together on the following common issues:

• securities held with non-resident custodians that are not covered by national securities holdings statistics;
• coverage of the other financial institutions (OFI) sector and, in particular, the timely coverage of SPEs, given the lack of primary statistics;
• coverage of financial derivatives for all sectors, owing to missing data sources and/or counterpart sector details.
• In addition to the collaborative work listed above, the WG ES and WG FA established a joint group on the valuation of unlisted shares and other equity in January 2020.

On the basis of this report, a list of notable issues affecting certain euro area countries, as well as the scope for improvement, is provided in Table 1 below.

Statistical issues affecting MIP indicators

The ECB, in collaboration with Eurostat, has continued to monitor specific quality aspects of the statistical outputs, as required under the MoU. In fact, some of the quality dimensions addressed in the report are also relevant for assessing the quality of data for MIP purposes (e.g. methodological issues A1 to A10, C3, F1 and G1 in Table 1). Some recommendations, such as those related to the functional classification (e.g. A5.1 to A5.3) or to the reconciliation of stocks and flows (C1), do not impact the computation of the main MIP indicators, but do play a role in the calculation and analysis of auxiliary indicators. However, the particularities of the annual data and of the MIP process, as well as the scope of the ECB’s responsibilities in the context of the MoU on the MIP (for those EU28 Member States that have designated the respective NCB to produce the b.o.p./i.i.p. datasets), create special analytical needs. In particular, longer time series (up to 15 years) are necessary for an accurate construction and analysis of the main MIP scoreboard indicators. All necessary data are available for the calculation of the main indicators (with a few exceptions for the goods and services balance). Certain coverage limitations remain, which will naturally vanish in the coming years, as each new release will limit the extent of the backdata required. Nonetheless, further data are needed for the calculation of auxiliary indicators (in particular the new auxiliary indicator, net international investment position excluding “non-defaultable” instruments^[7] as a percentage of GDP) from Malta, Croatia (from 2014), the Czech Republic and Greece (from 2013), Romania (from 2011), and Poland and Bulgaria (from 2010).

All in all, the impact of revisions on the direction (information) of first assessments is relatively minor. National errors and omissions in general remained stable in the last review period, but on average they are still above 2% of GDP in Ireland, Cyprus, Malta, Finland, Denmark and Sweden (see Chart MIP 1).

Last but not least, the analysis shows that discrepancies between b.o.p./i.i.p. statistics and sectoral accounts persist for several countries. This negatively affects the analytical combination of the two datasets and signals a lack of reliability or adequacy of the methodology of at least one of the two statistics. Despite this, the situation has improved compared to the previous quality report, in part due to the benchmark revisions.

For more information on the assessment of data quality for MIP purposes, please see the MIP box at the end of the main body of the report.

1 Introduction

This annual report provides a quality review of statistics on the balance of payments (b.o.p.), international investment position (i.i.p.) and international reserves template of the Eurosystem (international reserves).^[8] It fulfils the formal requirement of the ECB Executive Board to inform the Governing Council of the quality of these statistics, as set out in Article 6(1) of the ECB Guideline.^[9] Furthermore, the report provides information supporting the MIP data quality assurance process, as laid down in the MoU. The report follows the recommendations adopted by the Committee on Monetary, Financial and Balance of Payments Statistics (CMFB) in this domain.The focus of the report is on national data for the 19 euro area countries and euro area aggregates. The data for EU Member States (EU28) are commented on in the MIP box at the end of the report and are also available in the annexed tables^[10].

Scope of data coverage and structure of the report

This report analyses a number of aspects by which data quality can be measured. These include: (i) a review of methodological issues where national compilers diverge from statistical standards or need to enhance statistical procedures; (ii) an assessment of compliance by NCBs with their obligations to transmit data to the ECB, in terms of timeliness and coverage; (iii) the reliability of the statistical data; (iv) the internal consistency of the statistics, particularly as regards consistency over time, across frequencies and between accounts (net errors and omissions); and (v) external consistency/coherence, i.e. consistency vis-à-vis other statistical domains/datasets, namely foreign trade statistics, euro area sector accounts, MFI balance sheet statistics (including money market funds), investment fund statistics and securities holdings statistics.

The analysis covers quarterly and, in the case of euro area aggregates, monthly data. Section 3 (timeliness and punctuality), Section 4 (data and metadata availability) and Section 6.1 (validation/integrity rules) focus on one year of observations (July 2018/Q3 2018 to June 2019/Q2 2019). Section 5 (accuracy and reliability) analyses the impact of three years of revisions (April 2016/Q2 2016 to March 2019/Q1 2019), and the remainder of the sections focus on three years of data (Q3 2016 to Q2 2019).

The last data vintage used throughout the report is the one available as of 23 October 2019 and the country coverage is mostly the euro area, although the annexed tables provide information on the quality of the data for the EU28.

Given the specificities of the MIP process, some indicators on the fitness for purpose of the data are presented in a box at the end of the report for all EU Member States. The need for such a box arises from the fact that annual data display different properties compared with monthly and quarterly data, as well as from the need to assess the quality of data from non-euro area EU countries. The box draws on annual data up to 2018 and focuses on: (i) data availability, (ii) revisions; (iii) errors and omissions; and (iv) external consistency with sector accounts, i.e. MIP-relevant data quality dimensions. All indicators presented in the MIP box relate to national GDP to facilitate the analysis relating to the actual MIP scoreboard indicators.

2 Methodological soundness and statistical procedures

Methodological soundness means that concepts and definitions used to compile b.o.p./i.i.p. statistics broadly conform with the principles and guidelines outlined in BPM6 and take into consideration the agreements of the STC (and respective sub-structures) on the compilation of euro area aggregates.

One of the key elements of compiling consistent data is to adhere to the agreed standards and to transparently describe deviations. A detailed description of the data sources and compilation methods used by all Member States is available on the ECB’s website^[11]. The assessment included in this section is based on this ECB publication, as well as on the regular ECB contacts with national compilers regarding general data quality issues.^[12]

This quality report provides a succinct overview of the methodological soundness of b.o.p. and i.i.p. data for the main dimensions/principles.

2.1 Residency

The residency of institutional units should be defined in conformity with BPM6, particularly taking into account whether they have a predominant centre of economic interest in the country. This applies in particular to “Special Purpose Entities (SPEs)”, which are considered to be resident in the economy where they are incorporated.

Most countries correctly apply the residency concept. In the euro area, several countries host a large population of SPEs and therefore face certain challenges in achieving full coverage, and sometimes even in defining the residency of a certain entity. The current collection of Malta is based on a combination of administrative sources and surveys which are hampered by many limitations (e.g. low response rate, annual frequency with relevant delays, very limited geographical and instrument details).^[13] The 2018 revisions in the geographical allocation of positions for Malta introduced a series break in Q1 2016 that has not yet been solved. Malta should further improve the coverage, frequency and data quality following the established implementation plan.

In 2019, Cyprus substantially improved the coverage of SPEs, and this was reflected in large revisions and a better geographical allocation of external assets and liabilities from the reference period 2008 onwards. However, some limitations still apply to the geographical details, especially for debt instruments in FDI and other investment transactions and positions. These changes are visible in the improvement of the bilateral asymmetry indicators for direct investment positions (see Section 8.2).

Luxembourg’s SPE survey covers all SPEs with a balance sheet of over €500 million. Grossing up is performed for SPEs with balance sheets of between €300 and €500 million, which results in a final coverage of approximately 90% of total assets/liabilities. While additional improvements would represent a major effort for the country, the 10% missing is still quite sizable, taking into account the importance of this sector.

The Netherlands has also improved the accuracy of SPE data compared with 2018 through its efforts to integrate the compilation of b.o.p. and RoW statistics. However, the geographical allocation and the stock/flow reconciliation, as well as the identification of the type of relationship between FDI entities, still require further improvements. Last but not least, the link between the new quarterly data and the monthly estimates needs further attention to safeguard the quality of the more frequent data.

2.2 Functional classification

Most countries classify b.o.p. and i.i.p. data by function in conformity with BPM6 methodology. However, there is still room for improvement.

Regarding foreign direct investment (FDI), a number of countries, including Germany, Greece, France, Luxembourg and the Netherlands, classify transactions and related positions in debt securities between companies in a direct investment relationship as portfolio investment. This deviation creates internal inconsistencies at the euro area level, owing particularly to the residual approach used to calculate euro area portfolio investment liabilities. In the context of the implementation of the ECB Guideline, these misclassifications will become more transparent. Therefore the compiler should assess the potential relevance of the issue and implement a plan to address it. Similarly, trade credits and advances between companies in a direct investment relationship are included in other investment by Belgium and Spain, while Portugal has started to classify them under direct investment consistently for data from 2013 onwards. Germany classifies all transactions and positions in loans/deposits as other investment if at least one of the counterparts under a direct investment relation is an MFI. Malta includes most of the securities assets of the SPEs under portfolio investment, as no information is available regarding the relationship with the debtor.

Transactions and positions between fellow enterprises are not fully recorded under FDI. In particular, the Netherlands does not yet identify transactions and positions in either equity or debt instruments, while Germany, Greece, Austria, Slovenia, Slovakia and Finland do not include transactions and positions in equity.^[14] France records transactions and positions between fellow enterprises in equity from reference period Q1 2019 onwards. Moreover, Belgium, Germany, Estonia^[15], France, Cyprus, Lithuania^[16], Austria, Slovenia, Slovakia^[17] and Finland do not identify reverse direct investment in equity, and Malta shows negative values for liability positions.

2.3 Coverage

Financial intermediation services indirectly measured (FISIM) are not yet classified in the services account in Greece, but are instead still classified with income. Similarly, in a lot of countries service margins on buying and selling financial assets are not recorded or the compilation of this item is not sufficiently sound. Given the complexity of this issue, the WG ES, in collaboration with national compilers and other international organisations, has started investigating approaches to defining best practices and supporting those countries that have not yet estimated this financial service. Work is ongoing and specific guidance is expected during 2020.

According to public metadata, Cyprus^[18] and Luxembourg do not currently estimate employee stock options.

Germany should improve the data quality for transactions in financial derivatives for the government sector, as they are currently reported as zero, while positions are non-negligible. In addition, France does not record any transactions and positions in financial derivatives by the government sector. Ireland and Luxembourg should improve the stock/flow reconciliation for certain sectors. Irish financial derivatives data is quite implausible for investment funds as most of the changes in stocks are reported as other volume changes. Luxembourg also exhibits a reconciliation problem for deposit-taking corporations and financial corporations other than MFIs. Furthermore there is scope for increasing the quality of financial derivatives data in general. The WG ES, in cooperation with the WG FA, mandated a task force to issue recommendations on data sources and data collection and compilation methods. The task force is scheduled to provide guidance in the course of 2020.

In April 2015, the STC approved a new treatment for the recording of transactions and positions in euro currency in b.o.p./i.i.p. statistics. Most euro area countries have been following the new guidance in a timely and accurate manner, at least from reference period January 2014 onwards – with the exception of Ireland, Malta and Finland.

Ireland^[19] , Finland^[20] and Malta^[21] have started to report accurate monthly and quarterly intra-Eurosystem technical liabilities, though they still show diverse problems in the recording of euro currency holdings abroad (stocks).

Finland^[22] does not report insurance, pension schemes and standardised guarantee schemes before reference periods 2016 Q1 on the asset side. Ireland does cover the assets of these instruments but they are reported netted under the liabilities of the financial sectors other than MFIs, while Malta does not cover either assets or liabilities.

Furthermore, Malta does not report the breakdown of equity (into listed and unlisted shares, other equity and investment fund shares).

The WG ES in cooperation with the WG FA continued the work on estimating households’ assets held abroad, and broadly agreed on recommendations regarding the further development and use of data sources.

In general, most countries have difficulties in producing an accurate estimation of cross-border transactions and positions for the non-financial sector (particularly households); this under-coverage is believed to be relevant in particular for assets held outside the euro area, including those held with custodians. Most euro area countries use mirror data from: (i) the locational banking statistics of the Bank for International Settlements (BIS) and MFI balance sheet statistics from other euro area countries, to cover deposits and loans vis-à-vis non-resident banks; and (ii) so-called third-party holdings^[23] collected in securities holdings statistics to improve the estimates for securities. Therefore, NCBs are encouraged: (i) to report the breakdown “vis-à-vis households” by counterpart country to the BIS; and (ii) to integrate available mirror-data (reported by other NCBs) provided for their country and incorporate this information in their national data when appropriate.

Finland^[24] does not include any adjustment for households’ securities held abroad. The Netherlands does not observe securities held by households with custodians outside the Netherlands. Additionally, Germany does not collect information on securities held (positions) in custody abroad by non-bank enterprises and does not attempt any estimation.^[25] Many countries also have difficulties in accounting for real estate holdings, in particular those of resident households abroad. To complement the available information, the WG ES collects bilateral EU data that can be used as mirror data by compilers to cover resident holdings in other EU countries.

The majority of euro area countries estimate, to varying degrees, the impact of illegal economic activities. According to the b.o.p./i.i.p. book and bilateral contacts, Portugal started to include illegal trade in goods in its accounts as requested at EU level (smuggling, trafficking, illegal drugs) with the benchmark revision in October 2019.

Finally, national compilers in general should improve the measurement of reinvested earnings on FDI. They should implement, as much as possible, the recommendations of the Task Force on FDI (TF FDI), which are based on a closer control over the data they are collecting from reporting agents, whether through dedicated surveys or from business accounting data. The valuation of unlisted shares and other equity should also in general be improved and be carried out in a harmonised way. For this purpose a joint WG ES and WG FA^[26] group on unlisted shares and other equity was established in January 2020.

2.4 Other methodological issues

Horizontal and vertical inconsistencies in Malta’s i.i.p. data prevent the ECB from validating the figures reported for net external debt. Only the Maltese net external debt total is available and not its representation by sector, instrument and original maturity.

France and Germany estimate accrued interest for debt securities under portfolio investment income on a security-by-security (s-b-s) basis; however, no equivalent entry is recorded in the underlying instrument in the financial account.

For Ireland, other volume changes owing to changes in methodology or coverage are in some cases reported together with exchange rate and price revaluations, impacting the reliability of its stock/flow reconciliation.

Ireland does not value monthly reserve assets at end-month market prices (not even to account for exchange rate changes), thereby not reporting revaluation changes for inter-quarter months.

In the case of Ireland, the monthly estimate for goods (according to the community concept) does not display the expected seasonal pattern and often shows negative values for exports and imports following the community concept for periods from 2015 to 2017. The quality of the monthly Irish data frequently has a negative impact on the quality of euro area aggregates for goods.

France and the Netherlands systematically report zero monthly transactions in assets for money market funds (MMF) shares in their first estimates (intra and extra-euro area). France also reports zero transactions in MMFs liabilities. They usually fix this misreporting during subsequent quarterly transmissions.

In Belgium and Finland from 2008 to 2012 the resident banks classified some of the liabilities as loans.

3 Timeliness and punctuality

Non-compliance is defined with regard to (transmission) timeliness/punctuality and quality standards vis-à-vis the requirements laid down in the ECB Guideline ECB/2011/23 (as amended).^[27]

In the period under review (reference period July 2018 to June 2019), a persistent non-compliance case was recorded in the case of the Central Bank of Malta for not reporting the complete quarterly “other flows” detail and for recording delays in some monthly and quarterly transmissions.

In addition, the following ad hoc cases of non-compliance were recorded:

• The Central Bank of Ireland failed to transmit the banknote shipments data for the reference period February 2019 within the production window.
• The Central Bank of Latvia transmitted the monthly reserve assets data for the reference period August 2018 one day of delay. Attempts were made to transmit data prior to the deadline, but technical issues prevented their timely sending.
• The Bank of Finland transmitted the complete quarterly balance of payment data for the reference period Q2 2019 three days of delay. Furthermore, several data transmissions were necessary to reach a sufficient level of data quality for the revised periods.

4 Data and metadata availability

4.1 Completeness

For the reference period July 2018 to June 2019, the production of b.o.p., i.i.p. and international reserves statistics was smooth.

In terms of completeness, virtually all countries submitted all the mandatory items, albeit sometimes with delays (thus giving rise to cases of non-compliance – see Section 3 above). While complete datasets were eventually transmitted to the ECB, some delays in the correction of data quality issues detected during the data validation phase adversely influenced the production process. In some cases this created obstacles to the publication of timely and accurate euro area aggregates.

4.2 Accessibility and clarity

Accessibility refers to the conditions by which users can obtain, use and interpret data, ultimately reflecting how straightforward these are to access and the extent to which confidentiality constraints hamper the analytical work.

In line with the ECB legal framework on data confidentiality,^[28] all national data must be transmitted with a flag indicating its level of confidentiality. The ECB encourages national compilers to make as much data available as possible to final users (i.e. by marking observations as “free for publication”) and to ensure that statistical confidentiality flags are appropriately used.

Table 2 below summarises the share of observations marked as “free for publication” for the data requested under Tables 2A and 4A of Annex II to the ECB Guideline^[29] (i.e. the “main items”). The shares are calculated at dataset level for the reference period Q3 2018 to Q2 2019. Table A.1.1 in the Annex shows the same indicator for “all (mandatory) items” transmitted under the ECB Guideline.

The majority of the euro area countries (Belgium, Germany, Estonia, Greece, France, Italy, Latvia, Lithuania, Luxembourg, the Netherlands, Austria, Slovenia and Slovakia) released all “main items” to the general public. Among the remaining countries, Ireland, Spain, Cyprus and Portugal released more than 90% of this dataset, while only Malta and Finland released less than 90% of the observations in the Table 4A of Annex II to the ECB Guideline (i.i.p. data).

The overall situation for Tables 2A and 4A did not change significantly compared with last year’s assessment (small improvements were made by Ireland, Cyprus and Malta, while the share of free observations decreased slightly for Finland). It should be noted that the percentages are calculated based on the number of observations, without taking into account the relative importance (magnitude) of the data.

Full monthly b.o.p. datasets were flagged as “non-publishable” or “confidential” by Ireland, Cyprus, the Netherlands and Austria (generally on the basis of national dissemination policies). Concerning the full extent of quarterly data transmitted to the ECB (of which Tables 2A and 4A are only a small subset), and in line with last year’s results, nine euro area countries have made all the data required in the legal act available to final users for the quarterly b.o.p. and six euro area countries have done so for the quarterly i.i.p. (see Table A.1.1).

Clarity refers to the “information environment” of the data, i.e. whether the data are accompanied by relevant and pertinent metadata, illustrations (such as charts), information on their quality and potential limitations as to their use, and background information (sources and methods).

The ECB publishes monthly and quarterly b.o.p. and quarterly i.i.p. revaluations and other changes in volume for the euro area as a single economic area. Twelve monthly and four quarterly press releases, outlining the latest data and relevant economic developments, are published through wire services and on the ECB’s website. Furthermore, dissemination dates for all these press releases are announced at the beginning of each calendar year in the ECB’s Statistical Calendars.

The concepts and definitions used in the euro area b.o.p. and i.i.p. statistics are in line with international statistical standards. The “B.o.p. and i.i.p. book”, made available on the ECB website, aims at providing users with an overview of the main features of the b.o.p. and i.i.p. methodological framework and of the data sources and compilation methods used by the ECB (for the euro area) and in individual EU Member States.

The data can be accessed via the ECB’s Statistical Data Warehouse or in the External Transactions and Positions section of the Statistics Bulletin. Furthermore, the ECB has a Statistical Information Request facility to help external users of statistics access and analyse the data.

A subset of the statistics produced under the ECB Guideline can also be accessed via the Euro area statistics website. The aim of this dedicated website is to facilitate the understanding, use and comparison of euro area and national statistics by presenting the data in a user-friendly manner. This website also offers the possibility to easily download or share data by embedding the graphics into other websites, emails or social media.

Table A.1.2 in the Annex presents a summary of the national practices regarding data and metadata accessibility. Similarly to the ECB, all euro area countries provide technical facilities for downloading data in different formats (in Excel tables, CSV files, PDF documents or via interactive statistical databases). Furthermore, the majority of the euro area countries have statistical and/or economic bulletins providing a visual representation of the data in the form of charts, graphs and/or tables. Most euro area countries publish regular press release updates on their websites: on a monthly and/or quarterly basis. Last but not least, all countries present extensive information on their institutional environment and statistical processes in the “B.o.p. and i.i.p. book”, as well as on their national websites.

4.3 Availability of metadata

The ECB Guideline requires that the “data shall be accompanied by readily available information on single major events and on reasons for revisions, when the magnitude of the change to data caused by such single major events or revisions is significant […]”. Therefore, national compilers are encouraged to make regular and consistent use of the metadata template in all production cycles and publication means. In general, in the review period, the metadata transmitted by national compilers has been of sufficiently high quality to allow for the production of the euro area aggregates as well as to explain major developments in the aggregate. The ECB welcomes further efforts to improve the accuracy and level of detail in the metadata transmitted to the ECB and also encourages euro national compilers to exchange information with other euro area NCBs within the framework of existing arrangements, for instance in the context of FDI.

5 Accuracy and reliability (including stability)

This section reviews the stability of the data in terms of revisions to the “first assessment” or “first vintage”. In general, revisions are necessary to improve the accuracy of the data as first assessments may be based on incomplete, late or erroneous responses by reporting agents. However, large recurrent (biased) revisions may indicate low quality of data sources and/or methods that need to be addressed. Conversely, minimal or no revisions does not necessarily mean that the first assessment was of high quality; it may simply indicate a national preference for not revising the data.

In this report, quarterly revisions (for all euro area countries and for the euro area as a whole) and monthly revisions (for the euro area as a whole only) are assessed using indicators based on the comparison between first and “last”/most recent assessments.

Different indicators are applied depending on the features of the time series in question. Two basic types of indicators are used:^[30]

Relative size indicators measure the difference between the first and last assessments either in relation to the underlying series (when strictly positive; symmetric mean absolute percentage error – SMAPE) or otherwise in relation to a reference series (e.g. GDP or the underlying outstanding amounts for b.o.p. financial transactions; mean absolute comparative error – MACE). In the case of non-strictly positive (net/balance) time series, revisions cannot be properly related to the series value itself because observations may have different signs and, even more importantly, the value of the series may be close to zero. Therefore, for net/balance series the indicator used is the net relative revisions (NRR). The NRR puts the absolute revisions in relation to the average underlying gross flows for current account items and average positions of assets and liabilities for financial account transactions and positions. Owing to the different denominators employed, the SMAPE, MACE and NRR are not directly comparable.

Directional stability/reliability indicators measure how frequently first assessments are revised in the same direction (the upward revisions ratio and the directional reliability indicator).^[31]

All charts depict the indicators calculated for a revision window of three years (Q2 2016 to Q1 2019 for national and euro area aggregates – quarterly series – and April 2016 to March 2019 for euro area aggregates – monthly series).

In general, the revisions recorded for the period Q2 2016 to Q1 2019 were not fundamentally different from those recorded in the equivalent period analysed in last year’s quality report. However, since the last version of this report was published, 19 countries have implemented major national accounts and b.o.p./i.i.p. benchmark revisions, which has supported the alignment of national accounts (ESA 2010 data) with b.o.p./i.i.p. statistics. While increasing the accuracy, generally these revisions have not fundamentally altered the analytical interpretation of the first assessments.

5.1 Current account

In general, revisions to the euro area current account credits and debits were comparable for monthly and quarterly data as can be seen in Chart 1 below. The euro area aggregates recorded revisions comparable to the euro area country median (1% for the quarterly current account credits and debits), with the monthly data recording slightly higher revisions.

Cyprus^[32] and Malta had the highest revisions among euro area countries for current account credits and debits. Generally, Malta displayed a random pattern of revising its current account both upwards or downwards. By contrast, Cyprus revised its current account upwards in the majority of cases. However, both countries show high directional reliability in their revisions.

In terms of current account sub-items, in particular for monthly data, Ireland displayed a higher number of monthly revisions with weaker directional reliability compared with the quarterly data. These revisions had a negative impact on the quality of the monthly euro area aggregates.

Concerning revisions to the quarterly current account balance (see Chart 2 below), the euro area as a whole recorded comparable revisions to the median of the euro area countries (1%). Monthly revisions were slightly higher than quarterly revisions as assessed by the NRR indicator.

For the current account balance, the most sizable revisions were recorded by Ireland.

Detailed information on SMAPE, upward revisions and directional reliability indicators is available in Tables A.2.1 to A.6.2 in the Annex.

5.2 Financial account transactions

To overcome the fact that transactions in financial assets and liabilities can be either positive or negative, revisions to financial assets and liabilities are related to the respective i.i.p. item for assessing their relative size. MACE is therefore used to assess revisions to the financial account.

For the quarterly euro area aggregates, recorded revisions amounted to 0.2% of the underlying positions for total transactions in financial assets and liabilities, which is slightly lower than the median of euro area countries. Revisions to monthly euro area aggregates were considerably higher, as can be seen in Chart 3 below. Monthly revisions to euro area direct investment data were the highest, at close to 1% for both assets and liabilities, followed by revisions to other investment and portfolio investment.

All euro area countries recorded revisions of less than 1% of the underlying positions for quarterly financial transactions. The highest revisions were recorded by Cyprus^[33], the Netherlands^[34], and Finland^[35].

Concerning revisions to net quarterly financial transactions, the euro area as a whole recorded NRR comparable with the median of euro area countries (0.1%), while revisions to the monthly series were substantially higher (across all functional categories).

In terms of net financial account transactions for individual countries, Lithuania^[36] and Finland recorded the highest level of revisions among euro area countries (see Chart 4 below).

Detailed information on MACE, upward revisions and directional reliability indicators is available in Tables A.2.1 to A.6.2 in the Annex.

5.3 International investment position

Revisions to quarterly i.i.p. (financial account positions) are shown below in Charts 5 and 6. The euro area as a whole recorded revisions (as measured by SMAPE) of approximately 2% for both assets and liabilities, double the median for euro area countries.

At country level, revisions for assets and liabilities were generally comparable (with the exception of Slovenia). Lithuania and Slovenia recorded the highest revisions in the euro area. However, with the exception of revisions to Slovenian assets, this level of revision was comparable to other euro area countries. In the majority of cases, Lithuania^[37] and Slovenia revised upwards their first assessments of total i.i.p. (for both assets and liabilities)^[38]. However, the degree of directional reliability was very high for both Lithuania and Slovenia.

As regards revisions to net i.i.p., the euro area as a whole recorded revisions totalling 1.1% of the underlying average positions during the period under review (comparable to the median level of revisions for euro area countries). Slightly higher revisions (between 1.9% and 3.1%) were recorded in net positions for the various functional categories (direct, portfolio and other investment). At the level of individual countries, the highest NRR for net i.i.p. was recorded in Cyprus. Even if by a considerable distance, Slovenia was next, owing in particular to its revisions to assets.

Detailed information on SMAPE, NRR, upward revisions and directional reliability indicators is available in Tables A.2.1 to A.6.2 in the Annex.

6 Internal consistency

This section comprises two parts, assessing the reported national b.o.p. and i.i.p. data for internal coherence and consistency respectively. This comprises consistency over time (i.e. potential breaks in series), reconciliation across different frequencies (monthly and quarterly data) and an assessment of the arithmetic and accounting identities (including net errors and omissions).

6.1 Validation/integrity rules

This section reviews to what extent the transmitted national datasets were complete and met all basic accounting validation rules. These include linear constraints that apply to the b.o.p., i.i.p. and international reserves template statements, namely whether credits/assets minus debits/liabilities match the respective net flows/positions for each item, and whether sub-items add up to the respective items/totals, etc. Furthermore, it is strongly encouraged that datasets for different frequencies (i.e. monthly and quarterly) or data recorded in different datasets (e.g. reserve assets transmitted in the i.i.p. statement and in the reserve assets template) are kept consistent at all times.

In order to summarise compliance with validation rules, the average share of satisfied validations is used as an indicator (see section “Methodological documentation for quality indicators” for more details). The quarterly data generally had more validation issues than monthly data, but in both cases the failed validations did not impair the overall quality of the national data or euro area aggregates.

These results are fundamentally in line with last year’s assessment. For Ireland, a larger number of validation problems were detected in the monthly data (especially in counterpart and resident sector breakdowns and in the geographical detail) owing to the use of an incorrect methodology to reconcile monthly and quarterly data. Furthermore, Malta recorded inconsistencies in the intra/extra-EU geographical breakdown of the quarterly i.i.p. and reconciliation issues caused by the incomplete reporting of “other flows”. The share of satisfied integrity rules was also below 95% for Belgium (owing to reconciliation issues) and France (owing to issues in the functional detail, the geographical breakdown, and the resident and counterpart sector breakdowns).

Consistency between datasets is very important to ensure the overall quality of the b.o.p. As a result, average time consistency (ATC) and average relative explained changes (AREC) can be used as indicators to summarise consistency problems between frequencies and between positions and flows respectively.

In terms of time consistency, the vast majority of countries exhibit full consistency between monthly and quarterly data, with only a few exceptions. Relative to last year, Austria improved time consistency for most of its current account items, which are now fully consistent with the monthly figures. Meanwhile, Ireland continues to display, for most of the analysed b.o.p. and i.i.p. series, a level of time consistency below the euro area median, with extra-euro area secondary income and services showing consistent monthly and quarterly values in only 83% of cases (see Table A.8.1 in the Annex for more details).

In terms of average reconciled amounts for main items, all countries achieved full reconciliation between positions and flows, with the exception of Malta, which did not provide complete information on other flows (see Tables A.7.5 and A.7.6 in the Annex for more details).

Although the transmission of backdata is not mandatory, greater efforts by national compilers have resulted in the availability of longer time series for analytical use, including in the context of the MIP. While most countries have provided complete and validated datasets for periods before 2013, there are still several cases where these data are either incomplete or have serious validation problems. In general, despite improvements in data coverage and quality, it is of the utmost importance that countries continue their efforts to provide backdata of acceptable quality as agreed by the WG ES.

As regards series breaks, the following issues were identified:^[39]

Germany: Major breaks are observed due to the reclassification of positions between fellow companies from other investment to direct investment in Q4 2012. Major breaks are also present in Q4 2015 due to different estimation methods applied to portfolio investment debt securities liabilities.

The Netherlands: Major breaks are visible for several items from Q1 2015 onwards due to the introduction of new data sources and an updated compilation system.

France: Breaks are observed for secondary income credits of other sectors in Q1 2014, mainly reflecting the inclusion of net non-life insurance premiums and claims, and social benefits (new data collection method for insurance).

Luxembourg: relevant series breaks in foreign direct investment positions are observed in Q4 2011and Q4 2014 as a result of changes in the coverage of SPEs.

Austria: certain breaks apply in primary income credits and debits (from Q1 2013 to Q1 2016) as explained by SPE activity.

It should be noted, however, that countries are making continuous efforts to improve their data. Data transmissions submitted after the review period have already resulted in improved data quality.

Values for the validation indicators are available in Tables A.7.1 to A.7.3 in the Annex.

6.2 Net errors and omissions

Net errors and omissions (n.e.o.) (the difference between net lending/borrowing as compiled from the current plus capital accounts and the financial account) provide an indication of the internal consistency of the b.o.p. In fact, the principle of double-entry bookkeeping implies that the sum of all credit and debit transactions should be equal to zero in the b.o.p. statement (i.e. that n.e.o. are zero). Normal random imbalances commonly result from imperfections in source data and compilation practices. However, if these imbalances are large and/or persistent, they indicate problems in sources and/or methods.

In the context of b.o.p. compilation practices, it is not uncommon that statistical modelling and/or expert judgements are applied with the intent of imposing certain properties on net errors and omissions. This involves using statistical techniques to account for a lack of source data coverage or uncertainty about certain pre-identified items. Such mechanisms are typically incorporated in the compilation system and are applicable during each data production round. At euro area level, a correction mechanism that minimises net errors and omissions is also in place. The assumption behind the adjustment is that certain items in portfolio investment and other investment categories are not appropriately captured in the compilation of national data.

The average relative error for current account provides a measure of the magnitude of net errors and omissions in relation to average gross current account flows. Chart 7 below provides a graphical representation of the situation in euro area countries and the euro area aggregate (Chart A.7.7 in the Annex shows the average absolute n.e.o. in relation to the i.i.p.).

Overall, this year’s results are in line with those presented in last year’s quality report.

As expected (because of the correction mechanism), the euro area as a whole did not exhibit high n.e.o. compared with individual euro area countries. Monthly errors and omissions were substantially higher than quarterly ones. (The average absolute n.e.o. relative to average gross current account flows was 6% for monthly data and less than 2% for quarterly data.)

Quarterly n.e.o. for euro area countries generally exceeded 2% of the average current account gross flows. Over the period under review (Q3 2016 to Q2 2019), Finland displayed the highest average n.e.o. as a percentage of average current account gross flows at 14% (Ireland, which had the second highest n.e.o. in the euro area, recorded a value two times smaller than Finland). Countries are encouraged to continuously monitor the size of their n.e.o. and the underlying causes and address structural problems as soon as possible.

The persistence of the sign of errors and omissions is also relevant as a quality measure as it helps to identify biases in the accounts. Chart 8 below shows the cumulative n.e.o. in relation to current account gross flows.

Neither the euro area as a whole nor the vast majority of euro area countries display a significant statistical bias in their net errors and omissions.

Values for the validation indicators (including n.e.o.) are available in Tables A.7.1 to A.7.7 in the Annex.

7 External consistency/coherence

External consistency is defined as the coherence of b.o.p. and i.i.p. data with other related statistical domains. In this report, the external consistency/coherence of the b.o.p. and i.i.p. is assessed against foreign trade statistics, euro area (sector) accounts, MFI balance sheet statistics (including money market funds), investment fund statistics and securities holdings statistics.

7.1 Coherence with foreign trade statistics

International trade in goods statistics (ITGS) is typically the main data source used to compile the b.o.p. goods account in all euro area countries. However, when comparing the two datasets, important conceptual differences should be taken into account. Differences in concepts and definitions are linked primarily to the fact that b.o.p. follows the so-called change-of-economic-ownership principle, whereas ITGS record physical cross-border movements of goods^[41].

Given the methodological differences between the two datasets, a direct comparison would not convey an accurate picture. Instead, a directional reliability indicator is used to assess whether b.o.p. and ITGS data exhibit consistent developments and can hence be used as complementary analytical data sources. Furthermore, several countries publish reconciliation tables between the two datasets, which are available on the websites of the respective national central bank or national statistical institute.

Table A.8.1 in the Annex shows the individual national directional reliability indicators for the period Q3 2016 to Q2 2019 for the counterpart areas “rest of the world” and “extra-euro area”. The results are comparable to those presented in last year’s quality report.

For the euro area as a whole, there was full directional reliability for both imports and exports. Four euro area countries displayed full directional reliability for both exports and imports for the two counterpart areas analysed. A limited number of countries, including Malta,^[42] showed a lower degree of directional reliability.^[43] On average, data for exports/credits were as directionally reliable as data for imports/debits.

It should be noted that full directional reliability is not necessarily a sign of quality and that inconsistencies in the developments of the two datasets may be explained by the economic structure of the external trade in goods account of the respective country.

7.2 Consistency with euro area sector accounts

Euro area b.o.p. and i.i.p. data constitute one of the “building blocks” of the euro area accounts (EAA) and are widely used at national level for the compilation of the rest of the world (RoW) financial and non-financial accounts as part of the system of national accounts.

The methodological differences between the b.o.p./i.i.p. and the RoW account (national accounts) were removed with the introduction of ESA 2010 and the BPM6, albeit some challenges still remain when it comes to interpretation.^[44] analysis showed that inconsistencies between the two statistical domains persisted in many countries, negatively affecting the combined use of the two datasets and their reliability. Acknowledging this, the ESCB worked to precisely identify the differences and to develop national medium-term work plans to be generally observed by September 2019.^[45] In this context, while the removal of inconsistencies between the two statistical domains has progressed and most countries already compile the two sets of statistics in a consistent manner, a few countries still observe large discrepancies with a substantial impact on euro area and EU aggregates. Such issues are tackled in the context of the MIP quality assurance framework.

7.2.1 Current account

Chart 9^[46] shows the differences between the b.o.p. and RoW current accounts. As an indicative benchmark, relative differences should ideally be no higher than 0.5% of the underlying average b.o.p. and RoW values, as agreed by the STC.^[47]

For the euro area as a whole, the differences were not significant and were broadly unchanged relative to last year, with a high level of consistency between the two datasets. At country level, however, differences above 0.5% were recorded for several countries (Belgium,^[48] Germany (only for debits), Ireland, Greece, France and Luxembourg). Ireland, Greece (only for debits), France (only for credits) and Luxembourg recorded notable discrepancies (above 6%) for their current accounts, with sizeable discrepancies for services (Greece, France and Luxembourg) and primary income (Ireland, France and Luxembourg). In addition, differences above the threshold were also observed for a few other countries (Austria, Portugal, and Slovakia), without affecting consistency between the two datasets.

7.2.2 Financial transactions

Chart 10 shows the differences between the b.o.p. and the RoW account for financial transactions. In this case, discrepancies may be accounted for by time of recording differences, as well as by the reconciliation of national sectoral accounts. Both “vertical” reconciliation (a correction for errors and omissions) and “horizontal” reconciliation (asset/liability equality across sectors) may entail larger adjustments to the financial transactions in the RoW account. Nonetheless, as an indicative benchmark, the relative differences should ideally not exceed 0.3% of the average value of the underlying positions.

For the euro area as a whole, the differences were not significant (a bit higher than last year for assets and broadly unchanged for liabilities) and showed a relatively high level of consistency between the two datasets. At country level, differences of above 0.3% were recorded for several countries (Belgium, Greece, France (only for liabilities) and Malta (only for liabilities)). Greece recorded the highest relative discrepancies, while the largest absolute differences were observed in Germany (only for liabilities) and France. In addition, a difference above the threshold was also observed for Slovakia, but it did not affect the consistency of the two datasets.

7.2.3 Financial positions

Chart 11 below presents the differences between the i.i.p. and the RoW account for financial assets and liabilities (balance sheets/positions). As expected, the differences between the two datasets are larger for positions than for transactions. Relative differences should, as an indicative benchmark, be below 0.5% of the average financial assets/liabilities totals in the i.i.p. and sectoral accounts.

The euro area recorded discrepancies of 4% for both assets and liabilities, similar to last year. These discrepancies arose mostly from differences between the compilation and reconciliation processes for the euro area i.i.p. and the RoW. At country level, differences above 0.5% were recorded by Germany (only for liabilities), Greece, France and Malta. The highest discrepancies were recorded for France (assets) and Malta, with values exceeding 3%. In addition, differences above the threshold were observed for some countries – Italy, Portugal, Slovenia (only for liabilities), Slovakia (only for assets) and Finland – though this did not affect the consistency between the two datasets.

A detailed analysis at the instrument level reveals sizeable differences for equity instruments that are mostly triggered by different valuation practices (e.g. in the case of France regarding unlisted equity instruments)^[49]. Other reasons behind the differences also affecting the remaining instrument types include discrepancies in vintages, data sources and estimation methods.

Further details of these comparisons are available in Tables A.9.1 to A.9.4 in the Annex.

7.3 Coherence with MFI balance sheet data

Data on cross-border transactions and positions of the euro area MFI sector are recorded in the euro area b.o.p./i.i.p. and collected under the MFI Balance Sheet statistics (BSI)^[50].

Consistency between b.o.p. data for the MFI sector and transactions in external assets and liabilities derived from the BSI statistics of euro area MFIs is essential for the construction of the “monetary presentation of the balance of payments” and its use for monetary policy purposes.^[51] Furthermore, this consistency is also paramount for the compilers of euro area accounts, who use both datasets as “building blocks”. On these grounds, the ECB assesses the consistency between the two datasets in every regular production cycle, taking into account details by sector and instrument. Persistent discrepancies between the two datasets are generally explained by methodological differences (see below).

For the euro area as a whole, the discrepancies in comparable monthly figures between the two datasets were in general not significant for the period under analysis. Comparability issues were, however, observed for quarterly data on asset positions in equity. Discrepancies of around €60 billion, representing more than 17% of the average i.i.p. and BSI positions, were recorded for the euro area and explained by French data.

At the level of individual euro area countries, monthly transactions were generally consistent across datasets, representing an improvement relative to the previous review period. Exceptions were the discrepancies in equity transactions that were recorded for Ireland (9%) and Finland (15%). At a quarterly frequency, discrepancies in transactions mostly affected equity assets in Ireland (9.3%) and Luxembourg (7.3%).

In terms of positions, Ireland, France, Luxembourg and Slovenia all recorded discrepancies exceeding 25% of average positions for equity securities, but the French discrepancy of over €60 billion was the driver of the euro area discrepancy. In the case of loans and deposits, the highest discrepancy was found in Maltese liabilities, averaging 12% over the reference period. The highest discrepancy in debt securities was seen in Cyprus (5.7%).

The main reasons for these inconsistencies include: (i) differences in the classification of instruments (e.g. the b.o.p./i.i.p. may classify an instrument as a deposit, whereas it is classified as “remaining assets and liabilities” in BSI statistics);^[52] (ii) differences in the treatment of short-selling of securities in certain countries (off-balance-sheet treatment instead of a reduction in assets); (iii) reliance on distinct data collection systems, namely s-b-s for the b.o.p. and monthly aggregated sources for BSI, which leads in particular to differences in valuation criteria (i.e. the b.o.p./i.i.p. are calculated at transaction/market prices, while BSI transactions are derived from positions reported at fair, cost or nominal value, depending on accounting practices).

Eurosystem

Most of the discrepancies in the data for the Eurosystem as a whole (i.e. euro area aggregates) are related to the inclusion in the b.o.p. of estimates for foreign holdings of euro banknotes,^[53] while in BSI statistics all holdings of euro banknotes are deemed in circulation in the euro area.

At country level, the treatment of intra-Eurosystem technical claims is also a source of discrepancies, as these are included under remaining assets and liabilities without geographical breakdown in BSI, and under currency and deposits in the b.o.p./i.i.p. Additionally, the b.o.p. estimations for foreign holdings of euro banknotes are not included in BSI statistics.

Further details of these comparisons are available in Charts A.10.1 to A.10.6 in the Annex.

7.4 Coherence with money market fund statistics

Data on cross-border investment in euro area money market fund (MMF) shares are recorded within the portfolio investment account of the euro area b.o.p./i.i.p. Data on assets and liabilities of euro area MMFs are also collected under BSI statistics^[54], as MMFs is a sub-sector of MFIs.

At the euro area level, the i.i.p. consistently exceeded the BSI outstanding amounts of MMF shares issued by euro area residents and held by non-euro area residents. At country level, small discrepancies were recorded in the period under review for Ireland, France and Luxembourg (the only countries in the euro area with relevant MMF activity).

The discrepancies between the two sets of statistics were related to the use of different compilation methods in b.o.p., i.i.p. and MFI balance sheet statistics. While the “residual approach” is used to calculate b.o.p. and i.i.p. portfolio investment liabilities,^[55] MMF liabilities are allocated geographically by respondents in BSI statistics. Although in the case of MMF shares there is, in principle, no significant trading in secondary markets, the intervention of intermediaries buying, holding and selling shares on behalf of their clients can make it difficult to identify the place of residence of the actual holders. In such cases, the first counterpart – the custodian or other intermediary – may be known, but the final investor often is not. Identifying the place of residence becomes increasingly complicated as the length of the chain of intermediaries increases, so the residual approach of the b.o.p. and i.i.p. may be more accurate.

Further details of these comparisons are available in Charts A.11.1 and A.11.2 in the Annex.

7.5 Coherence with investment fund statistics

Details on cross-border investments in non-MMF investment fund (IF) shares are recorded in the b.o.p. and i.i.p. statistics within portfolio investment. Data on IF assets and liabilities are collected under the Regulation on Investment Funds^[56] (IF dataset).

At the euro area level, the i.i.p. consistently exceeds the IF dataset in terms of euro area investment fund liabilities. The average absolute discrepancy reached a value close to €170 billion for positions and €3 billion for transactions throughout the period under analysis. The discrepancies at the euro area level are partly explained by the use of the residual approach to calculate portfolio investment liabilities (see Section 7.4 above).

At country level, Greece (133%) has the highest relative inconsistencies for IF shares held by non-residents, although outstanding amounts are small. The i.i.p. data reported by Malta are zero, whereas the IF dataset shows positive (although not very sizeable) outstanding amounts. In addition, while France displays a relative discrepancy of 5%, the average absolute discrepancy stands at €7.3 billion for stocks. The two datasets are fairly consistent as regards transactions, with the largest discrepancy affecting Irish data (€7 billion).

Further details of these comparisons are available in Charts A.12.1 and A.12.2 in the Annex.

7.6 Coherence with securities holdings statistics

The ECB Guideline stipulates that portfolio investment collection systems of euro area countries shall as much as possible rely on s-b-s information (see Annex VI of the ECB Guideline). In particular, it is stated that “the target coverage is defined as follows: stocks of securities reported to the national compiler on an aggregate basis, i.e. not using standard (ISIN or similar) codes, should not exceed 15% of the total portfolio investment stocks of assets or liabilities”. Therefore, it is expected that b.o.p. and i.i.p. statistics and SHSS^[57] provide consistent results, mainly because national portfolio investment assets and SHSS should rely on the same s-b-s sources of information.^[58]

This section compares the positions at market value of (i) debt securities and (ii) listed shares and investment fund shares/units as available in the SHSS dataset.^[59] This analysis considers, on the SHSS side, the cross-border holdings by residents of each euro area country as collected by the respective country, as well as holdings by non-financial investors of each euro area country held in custody in other euro area countries (i.e. the so-called third-party holdings).

7.6.1 Debt securities

Taking into account the scope of the compilation of portfolio investment on an s-b-s basis as indicated above, the focus should be on discrepancies that are above 15% of the respective position.

For the euro area as a whole, the level of discrepancies for debt securities was 7% of the underlying i.i.p., which signals a good degree of consistency with SHSS. At the level of individual countries, there were, for the first time, no cases of relative discrepancies above 15% owing to SHSS under-coverage. Conversely, Cyprus recorded a difference slightly above 15% owing to over-coverage of SHSS amounts. This reflects the inclusion of third-party holdings data in SHSS in relation to long-term debt securities held by non-financial investors.^[60]

The decline in SHSS holdings by financial corporations other than MFIs of long-term debt securities issued by non-euro area countries explains, to a large extent, the (positive) b.o.p.-SHS gap. The lack of comprehensive coverage of non-ISIN securities data in SHSS,^[61] the different revision policies for SHSS and the i.i.p., and the i.i.p.’s attempts to cover securities held with custodians outside the euro area explain a significant part of this discrepancy.

Further details of these comparisons are available in Chart A.13.1 in the Annex.

7.6.2 Listed shares and investment funds shares/units

For the euro area as a whole, the total discrepancy as a percentage of the underlying i.i.p. was 5%. At country level, discrepancies above the 15% threshold owing to SHSS under-coverage^[62] were recorded in Italy, Portugal and Finland. Some countries also recorded over-coverage of SHSS amounts – which was fairly significant, in absolute terms – in respect of investment fund shares held by German financial corporations other than MFIs and non-financial investors, which were issued mainly by other euro area countries (with the latter being linked to the inclusion of third-party holdings data in SHSS). Finally, Malta continued to report zero holdings of listed shares and investment fund shares within its b.o.p. and i.i.p. statistics, meaning that indicators were not calculated for this country despite relevant amounts being reported in the context of SHSS for these instruments.

To a large extent, the decline in SHSS holdings by financial corporations other than MFIs of listed shares and investment fund shares issued by non-euro area countries explains the positive b.o.p.-SHSS gap. The caveats mentioned for debt securities also hold when it comes to explaining this discrepancy.Further details of these comparisons are available in Chart A.13.2 in the Annex.

8 Asymmetries

Asymmetries are an inherent feature of all statistics for which “mirror” data are collected, i.e. for which two countries collect the same type of information in relation to each other. They occur when one country’s data do not correspond to the data for the same transaction reported by its partner country. In reality, however, for a variety of reasons it is rarely the case that two data sources provide exactly the same results, and this leads to the emergence of asymmetries.

Asymmetries can be observed at the level of the global economy (where total world assets should equal total world liabilities), at the level of geographical aggregates (where total intra-euro area assets should match total intra-euro area liabilities) and at the level of bilateral pairs (where flows and positions between pairs of countries should match perfectly).

8.1 Intra-euro area asymmetries

Charts 12 and 13 provide an overview of intra-euro area asymmetries in the current and capital accounts and the financial account respectively.

Current and capital account asymmetries (credits minus debits) were always positive over the period under review. The main contributors to the overall asymmetries show structural biases: consistently positive asymmetries in goods and services accounts, with negative contributions being made by the primary income account. The secondary income and capital accounts only contributed to overall asymmetries in particular quarters.

In the financial account, asymmetries were mainly recorded in direct and other investment. Portfolio investment and related income do not show asymmetries by construction, owing to the residual compilation approach at the euro area level. Financial account asymmetries were fairly volatile in the period under review, with periods where asymmetries in direct and other investment offset each other alternating with periods where they both contributed in the same direction to the overall asymmetry.

8.2 Bilateral asymmetries

Quarterly bilateral transactions and positions between euro area countries are transmitted to the ECB on a voluntary basis, hence a full bilateral dataset is not yet available. Owing to data availability, the analysis of bilateral asymmetries between euro area countries is performed only for direct investment. The analysis will be extended to other items as data becomes available.

The internal and external country geographical quality indicators (ICGQ and XCGQ respectively) are measures that summarise the quality of the geographical breakdown. The ICGQ aims to assess the accuracy of individual countries’ geographical classification within the sample of countries for which bilateral data are available by aggregating absolute bilateral asymmetries. Meanwhile, the XCGQ aims to show how well a country’s reported intra-euro area aggregate matches its mirror data, calculating the difference between the intra-euro area figure reported by the country under consideration and the corresponding figure derived from counterpart data. More information on these indicators can be found in the section on “Methodological documentation for quality indicators”.

The results of the ICGQ indicator for FDI transactions were characterised by significant variability across countries and over time. Several countries consistently recorded high scores across the entire time period, indicating structural problems in matching counterparties’ transactions. Meanwhile, the majority of countries experienced high volatility in the measures over time, pointing to quarter-specific problems in capturing the geographical detail of transactions, rather than structural issues.

Results for the XCGQ indicator were generally better than those recorded for the ICGQ, as that indicator is less specifically about matching up individual country counterparts and merely measures how well the counterparts as a group match a country’s estimate for that group. Consequently, most of the countries performed relatively well across the entire time period. This finding is obviously welcome from the point of view of the quality of overall euro area data. Nonetheless, several countries still recorded fairly poor results in several quarters.

For both quality measures, the results recorded for FDI positions were better than those observed for transaction data.

Overall, it appears that countries that are characterised by large numbers of SPEs and face well-known challenges when it comes to capturing and measuring the activities of those institutions were found to have structural problems matching the figures provided by their euro area counterparts.

Further information on summary indicators of bilateral asymmetries is available in Tables A.14.1 to A.14.4 in the Annex.^[63]

Box 1 Quality indicators on b.o.p. and i.i.p. statistics underlying the MIP

The MIP scoreboard for the Alert Mechanism Report (AMR) consists of 14 headline indicators with thresholds (complemented by auxiliary indicators with no thresholds). The composition of the MIP indicators is subject to review and evolves over time in order to reflect the latest developments or increased data needs. Most of these indicators are composite, i.e. they make use of at least two data sources.

Balance of payments (b.o.p.) and international investment position (i.i.p.) data underpin the construction of the following three headline indicators:

• current account balance (percentage of GDP), three-year backward-moving average (up to 13 years of data required);
• net international investment position (percentage of GDP) (up to ten years of data required);
• export market share (percentage of world exports), five-year percentage change (up to 15 years of data required).

Additionally, b.o.p. and i.i.p. data are also used for five auxiliary indicators:

• current plus capital account balance (net lending/borrowing) (percentage of GDP) (ten years of data required);
• net international investment position excluding “non-defaultable” instruments^[64] (NENDI) (percentage of GDP) (ten years of data required);
• foreign direct investment in the reporting economy, flows (percentage of GDP) (ten years of data required);
• foreign direct investment in the reporting economy, positions (percentage of GDP) (ten years of data required);
• export performance against advanced economies (percentage of OECD exports), five-year percentage change (15 years of data required).

Together, these indicators provide analytical evidence of possible vulnerabilities and risks that would require further investigation at country level.

The following sections assess the fitness for purpose of b.o.p. and i.i.p. data used for the MIP, analysing the data vintage used in the 2019 Alert Mechanism Report.

Institutional set-up

B.o.p. and i.i.p. data are transmitted to the ECB on the basis of Guideline ECB/2011/23 and to Eurostat on the basis of Regulation (EC) No 184/2005. This annual quality report follows the basic principles of the “Public commitment on European statistics by the ESCB” and is a requirement under Article 6(1) of Guideline ECB/2011/23. This report is fully coordinated with the report produced by the European Commission (Eurostat) on the basis of Article 4(4) of Regulation (EC) No 184/2005. The quality assessment of the Eurostat report is conducted in accordance with the “European Statistics Code of Practice”.

The indicators used for the MIP are provided by Eurostat on the basis of statistics compiled in the Member States by either NSIs or NCBs. The MoU was therefore signed in November 2016. In the MoU (and the related letters that were exchanged), the European Commission and the ECB mutually recognise the quality assurance frameworks in place in the European Statistical System (ESS) and the ESCB and establish practical working arrangements for cooperation with regard to the quality assurance of statistics underlying the MIP.

The MoU specifies that Eurostat and the ECB’s Directorate General Statistics (DG‑S) should regularly conduct assessments of the quality of national datasets. In particular, the ECB/DG‑S should run its quality procedures for the datasets reported by NCBs and provide Eurostat with quality-assured datasets and/or information on the quality of the data after the regular data transmission in September/October each year. The MoU also envisages visits by the ECB/DG‑S and Eurostat to NCBs and/or NSIs to help assess the output quality of MIP-relevant data. In 2019, country visits to Germany, Malta and Ireland took place, and as a result of those visits, recommendations for improving data quality were included in the relevant sections of this report.

To ensure full transparency with regard to the quality of MIP-related statistics, a three-level quality reporting system has been set up over the last few years with the support of the CMFB. That system consists of national self-assessment reports (Level 3), which, in turn, feed into the domain-specific quality reports (Level 2) – including this report – which are coordinated between the ECB and Eurostat. Finally, a joint Eurostat/ECB summary report assessing the quality of all statistics underpinning the MIP (Level 1) is published each year on the CMFB’s website.

Data availability and confidentiality

The relevant ECB and European Parliament and Council legal acts do not impose backdata requirements in compliance with the BPM6 statistical standard. Despite this, the majority of national compilers have provided the thirteen years of current account backdata and ten years of net international investment position backdata that are required for the calculation and analysis of the main indicators. Certain coverage limitations remain, but these will naturally vanish in the coming years as each new release will reduce the required backdata.

As regards the auxiliary indicators, there are coverage limitations affecting the calculation of the indicator added in 2018 (net international investment position excluding “non-defaultable” instruments – NENDI), which uses positions in equity securities. Information from the following countries is unavailable or of poor quality (as it also includes investment fund shares): Malta, Croatia (from 2014), the Czech Republic (from 2013), Greece (from 2013), Romania (from 2011), Poland (from 2010) and Bulgaria (from 2010). In general, all available MIP-relevant data are free for publication.

Sources and methods

The introduction of BPM6 provided an opportunity for a large group of countries to move over to survey-based systems as an alternative to traditional international transaction reporting (“settlement”) systems. B.o.p. and i.i.p. statistics are, by nature, based on a multitude of data sources, relying on micro datasets (e.g. the Centralised Securities Database (CSDB)), macro datasets, direct reporting and counterpart information, statistical surveys and administrative datasets (e.g. for the general government sector).

While the compilation of b.o.p. and i.i.p. data in EU Member States is deemed methodologically sound, there are challenges when it comes to measuring some components and complying with all EU recommendations and/or BPM6 standards. In particular: (i) Luxembourg, the Netherlands, Cyprus and Malta would benefit from further improvements in the coverage of resident SPEs; (ii) all EU countries should make an effort to implement estimates for service margins as regards the buying and selling of financial assets (financial services); (iii) some countries need to follow EU recommendations and include estimates for certain illegal economic trade activities (illegal drugs, prostitution services, and smuggling of tobacco and alcohol); (iv) most countries have difficulties capturing households’ assets held abroad; and (v) all countries should improve the measurement of reinvested earnings on foreign direct investment and the valuation of unlisted shares and other equity. For more detailed information, see Table 1 in the executive summary and Section 2.

Accuracy and reliability

Since the last review period, 19 countries have implemented major national accounts and b.o.p./i.i.p. benchmark revisions, which have supported the alignment of national accounts (ESA 2010 data) with b.o.p./i.i.p. statistics. These revisions and other regular revisions have not significantly altered the analytical interpretation of the indicators, with the exception of the 2017 current account indicator for Cyprus, which now records a value between the threshold limits, in contrast with the last review period.

Internal consistency

For the quarterly b.o.p., most countries fulfil all validation (accounting) rules. One of the most common issues among countries concerns the reconciliation of positions and flows, which is very important for confirming the plausibility of the net i.i.p.

As regards series breaks, in addition to the issues mentioned in Section 6.1 (validation/integrity rules), the following breaks apply to periods before 2013 (transmission of data for periods before Q1 2013 is not mandatory):

Czech Republic: Breaks apply to the services balance between 2004 and 2014, primary income credit in Q4 2005, and financial account assets and liabilities in Q1 2017.

Germany: Breaks are present in Q2 2012 owing to the introduction of a new survey that allows positions between fellow companies to be reallocated from “other investment” to “direct investment”. Breaks are also present in Q4 2015 owing to a change to the estimation method applied for portfolio investment debt securities liabilities.

Ireland: Available foreign direct investment position data before 2008 follow the directional principle.

Croatia: Breaks are observed for stocks of financial derivatives assets (2014), stocks of direct investment assets (between 2010 and 2014), and secondary income (2013).

Italy: Breaks in the series for financial derivatives (assets and liabilities) are observed in 2008 as a result of the introduction of a more accurate quarterly i.i.p. data source for financial derivatives held by resident deposit-taking corporations;

Luxembourg: relevant series breaks in foreign direct investment positions for 2011 are related to improvements in the coverage of SPEs;

The Netherlands: a break was recorded in 2015 due to the introduction of an update compilation system;

Austria: some breaks apply in foreign direct investment positions (in 2005, due to the introduction of data for SPEs) and financial derivatives (before 2006 the reported value for positions in financial derivatives is zero, whereas non-zero values are reported for transactions);

National net errors and omissions in general remained stable in the last review period, however they are still above 2% of GDP in Ireland, Cyprus, Malta, Finland, Denmark and Sweden (see Chart MIP 1). In this context, it is important to highlight that some euro area countries have formal correction mechanisms to address this problem, naturally leading to reduced levels of errors and omissions.

Chart MIP 1

Average absolute net errors and omissions

(percentage of GDP)

Source: ECB.

In cumulative terms for the period 2016‑18, a bias (at least 2% of GDP) can be statistically identified in Cyprus, Bulgaria and Sweden.

External consistency

The methodological differences between the b.o.p./i.i.p. and the RoW account (national accounts) were removed with the introduction of ESA 2010 and BPM6. However, analysis shows that inconsistencies between the two statistical domains persist in several EU Member States, negatively affecting the combined use of these two datasets, as well as their reliability. The CMFB endorsed a medium-term work plan designed to eliminate most discrepancies by September 2019. The remaining discrepancies will be analysed in depth by ECB and Eurostat, and the most relevant outstanding differences will be addressed.

Discrepancies above 0.5% of GDP are still recorded (for either credits/debits or both) for the current accounts of one-third of EU countries (Ireland, Greece, France, Luxembourg, Malta, Portugal, Slovakia, the Czech Republic and Denmark).^[65] Nonetheless, with one minor exception (Ireland), none of the discrepancies recorded was above 2% of GDP. For financial account positions, the discrepancies between the i.i.p. and the RoW account are more pervasive, totalling more than 10% of GDP in three cases: France (assets only), Malta and Croatia (assets only).

MIP Annex Table 1

Annual absolute revisions – balance/net items for 2017

(percentage of GDP)

Source: ECB.Note: All indicators are compiled using neither seasonally adjusted nor calendar adjusted data.

Annexes

See more.