Euro area balance of payments and international investment ...January 2005 Some aspects of quality...

EURO AREA BALANCE OF PAYMENTS AND INTERNAT IONAL INVESTMENT

POS I T ION STAT I ST I C SJANUARY 2005

ANNUAL QUALITY REPORT

In 2005 all ECB publications will feature

a motif taken from the

€50 banknote.

EURO AREA BALANCE OF PAYMENTSAND INTERNATIONAL INVESTMENT

POSITION STATISTICS

ANNUAL QUALITY REPORT

JANUARY 2005

© European Central Bank, 2005

AddressKaiserstrasse 2960311 Frankfurt am Main, Germany

Postal addressPostfach 16 03 1960066 Frankfurt am Main, Germany

Telephone+49 69 1344 0

Websitehttp://www.ecb.int

Fax+49 69 1344 6000

Telex411 144 ecb d

All rights reserved.Reproduction for educational andnon-commercial purposes is permittedprovided that the source is acknowledged.

As at December 2004.

ISBN 92-9181-533-0 (print)ISBN 92-9181-534-9 (online)

3ECB

Euro area balance of payments and international investment position statisticsJanuary 2005

CONTENTSC O N T E N T SEXECU T I V E S UMMARY 4

I N T R O D U C T I O N 5

1 ME THODOLOG I C A L S OUNDNE S S 6

2 P E R I OD I C I T Y AND T IM E L I N E S S 7

3 R E V I S I ON P R A C T I C E AND PO L I C Y 7

4 S TA B I L I T Y 7

4.1 The directional reliability showsa relative weakness in directinvestment 8

4.2 The mean absolute percentageerror shows an improvement forincome 9

4.3 The improvement in the currentaccount is confirmed by the rootmean square relative error 9

4.4 Stability of the internationalinvestment position 10

5 C O N S I S T E N C Y 1 1

5.1 Internal consistency 11

5.2 External consistency 12

6 C O N C L U S I O N 1 3

A N N E X E S 1 4

4ECB cEuro area balance of payments and international investment position statisticsJanuary 2005

EX E CU T I V E S UMMARYThis report is required by Article 6 of GuidelineECB/2004/151 (hereinafter “the Guideline”). Itfollows the basic principles of the InternationalMonetary Fund (IMF) Data QualityAssessment Framework (July 2003) in terms ofthe different dimensions of data quality to beinvestigated, and includes some quantitativeindicators that are based on the work of a jointECB DG-Statistics/European Commission(Eurostat) Task Force on Quality, alsoinvolving representatives of most of the then 15EU Member States. Key aspects for assessingquality focus on (i) methodological soundness,(ii) timeliness, (iii) revision policy andpractice, (iv) stability and (v) consistency. Thisreport is the first one encompassing variousquality measures for the euro area statistics.It may also be considered as a pilot as such awide set of quantitative indicators have rarelybeen published in a periodic report in any othercountry or for other sets of statistics.Therefore, the quantitative results should beanalysed and interpreted in their appropriatecontext.

The methodologies observed by Member Statesare covered in the country chapters of theECB’s yearly publication “European Unionbalance of payments/international investmentposition statistical methods” (last update:November 2004). The ECB website alsocontains a methodological note on the euro areabalance of payments (b.o.p.) and internationalinvestment position (i.i.p.) which focuses oncommon methodological issues and on theaggregation procedures.

In the first assessment of the b.o.p. currentaccount, full information on goods and servicesis not yet available and estimations are oftennecessary to meet the deadlines. For the timebeing, the results of the stability indicators forthe current account show that these firstestimates systematically underestimate creditsand debits to some extent. Nonetheless, thesepatterns do not seem to affect the net currentaccount.

In general, the picture shown in the euro areab.o.p. and i.i.p. statistics appears credibleand the quantitative indicators show animprovement from 1999 onwards. The resultsfor 1999 and 2000 reflect the fact that nationalb.o.p./i.i.p. compilers were still in the processof adapting their data collection systems tocover the needs of euro area statistics and,hence, revisions were extensive. The level ofrevisions has diminished in more recent years,despite positive errors and omissions in theearly b.o.p. data assessments from mid-2001 tothe end of 2003. However, revisions in net i.i.p.data published in November 2004 amount to€328 billion, or 4.5% of GDP.

While there are differences in the levels forgoods in b.o.p. and external trade statisticsdue to the underlying methodologies, thedifferences in their respective month-to-monthgrowth rates decreased from 0.9 percentagepoint for the period 1999-2001 to 0.7 and0.6 percentage point for exports and importsrespectively for the period 2001-2003.

1 OJ L 354, 30.11.2004, p. 34. This Guideline replaced theGuideline ECB/2003/7 of 2 May 2003, itself an update ofsuccessively the Guidelines ECB/1998/17 and ECB/2000/4.

5ECB


ANNUALQUAL ITYREPORTThe euro area b.o.p. and i.i.p. are based on an

aggregation of statistics provided by individualeuro area countries concerning transactionsand positions between their own residents andnon-euro area residents. The current legalframework for the provision of data to the ECBis established by the Guideline ECB/2004/15,which entered into force on 1 September 2004.This Guideline defines the requirements for thecompilation of the quarterly i.i.p. as from 2005,as well as future requirements related to thevaluation of direct investment positions andto the compilation of portfolio investmentliabilities. In addition, in order to improvethe accuracy and consistency of portfolioinvestment statistics, the Guideline requiresthe collection of quarterly stocks data throughsecurity-by-security reporting as from March2008.

Taking into account, among other things, thevariety of methods and sources at the nationallevel, no simple assessment, whetherqualitative or quantitative, can fully reflect thequality of the euro area statistics. However,some dimensions of quality have been assessedqualitatively, and quantitative indicators2 havebeen designed to help users in the analysis ofthe data.

Key aspects for assessing quality include– methodological soundness,– timeliness,– revision practice,– stability3 and– consistency.

In this report, after the introduction, Section 1concentrates on the methodological soundnessfrom a quality perspective. In Section 2,timeliness is measured by the interval betweenthe end of the reference period and publication.The euro area b.o.p. revision practice andpolicy is explained in Section 3. Quantitativeindicators are calculated for measuring the sizeof revisions in Section 4 on stability, and theconsistency within the b.o.p. (“internal”) andwith other related statistics (“external”) inSection 5. Internal consistency results in low

errors and omissions,4 whereas externalconsistency is measured in relation to externaltrade and to financial flows derived from thebalance sheets of monetary financialinstitutions (MFIs).5 This report encompassesvarious quality measures for the euro areastatistics for the first time. It may also beconsidered as a pilot as such a wide set ofquantitative indicators has rarely beenpublished in any other country or for other setsof statistics, although the Australian Bureau ofStatistics and the Reserve Bank of India haverecently used some quantitative indicators andmore qualitative assessments.6

The calculations of quantitative indicatorswere performed on monthly b.o.p. observationsfrom January 2001 to December 2003 (36observations). Those results are compared withthe periods from January 1999 to December2001 and from January 2000 to December2002.7 These periods were chosen in order toproduce statistically meaningful results whichreflect an average for the whole period. Themost recent observations were excluded toavoid underestimating the indicators ofrevisions. In the initial years of the periodsunder consideration, national compilers werestill in the process of adapting their collectionsystems to cover the needs of euro areastatistics for the conduct of the single monetarypolicy.

I N T RO D U C T I O N

2 Documentation on the indicators used in this report areavailable on the Committee on Monetary, Financial andBalance of Payments Statistics (CMFB) website (http://www.cmfb.org/) in: Joint ECB-Eurostat Task Force onQuality: Report on the quality assessment of balance ofpayments and international investment position statistics.

3 This element is identif ied as “revision studies” in the IMF’sData Quality Assessment Framework (DQAF) for Balance ofPayments Statistics, International Monetary Fund StatisticsDepartment, July 2003.

4 Low errors and omissions can provide an indication ofaccuracy, although they can also result from the balancing outof large opposite discrepancies.

5 External consistency for the euro area will also focus on therest of the world account within the quarterly euro areaaccounts. Such consistency would be achieved fully if the restof the world account were built from the euro area b.o.p. andi.i.p., and revised simultaneously.

6 The Reserve Bank of India made an explicit reference to thereport of the DG-Statistics/Eurostat Task Force on Quality,and in particular to the recommended indicators.

7 The results are based on data published in November 2004.


Some aspects of quality are interrelated. Forexample, there is a trade-off betweentimeliness and accuracy of statistics.Publishing early enough to meet user needsmay imply using less information and thusmaking larger revisions when data from othersources, late responses and more robustestimations become available. In addition,achieving external consistency may requireadditional revisions related to differences inthe timing of updates, in concepts, inestimation procedures and in samplingtechniques. Thus revisions increase otheraspects of quality, such as methodologicalsoundness and external consistency.

The IMF has established the Special DataDissemination Standard (SDDS) to guidemember countries in the provision of theireconomic and financial data to the public. 57 ofits member countries have subscribed to thestandard, including a majority of EU MemberStates. The ECB conforms to it as well in orderto foster international comparability of euro areastatistics. References to the SDDS benchmarkare made in this report where appropriate.

1 METHODOLOGICAL SOUNDNESS

The methodologies observed by Member Stateswhen compiling the b.o.p. and i.i.p. are coveredin the country chapters of the ECB’s yearlypublication EU b.o.p. and i.i.p. statisticalmethods. That publication describes thecollection system in each Member State andincludes details about the reporting population,administrative sources, periodicity of surveys,estimations and legal framework. The latestupdate, dated November 2004, covers the 25EU Member States, as well as the two currentAccession Countries (Bulgaria and Romania).The compilation methods for internationalreserves (flows and outstanding amounts) aredescribed in a separate report.8 The agreedmethodology goes somewhat beyond theinternational standards set out in the IMFBalance of Payments Manual, 5th edition

(BPM5) (October 1993), to meet specific userrequirements and to ensure greater consistencywith other monetary and financial statistics.

In 2003, one Member State introduced a morecomprehensive survey in order to increase thecoverage of special purpose entities created bymultinational enterprises. The results of thissurvey were incorporated into the i.i.p. as atend-2002 and end-2003.

In addition, the “statistics” section of the ECBwebsite contains a methodological notespecific to the euro area b.o.p. and i.i.p.focusing on common methodological issues, aswell as on the aggregation procedures at theeuro area level. It is updated whenever changesoccur.

In 2004, the b.o.p. and the i.i.p. haveincorporated a significant improvement,namely a split between “currency and deposits”and “loans” in other investment of the generalgovernment and other sectors. Previously,“currency and deposits” and “loans” wererecorded jointly. This split had been requestedby users in order to monitor the deposits abroadby households and non-monetary companies.Moreover, this split forms part of the IMFb.o.p. standard components, and is required forthe compilation of the euro area quarterlyfinancial accounts.

In comparison with the international standardsset out in BPM5, the euro area b.o.p. and i.i.p.still lack the sector breakdown on the liabilitiesside of portfolio investment, owing todifficulties in obtaining information on thefinal holder of securities (i.e. the actualcreditor).9

The ECB and the national central banks of theEU Member States are involved in an action

8 Statistical treatment of the Eurosystem’s internationalreserves, ECB, October 2000.

9 The data necessary to show this breakdown, important also asa contribution to the breakdown of counterparts to M3 and tothe quarterly f inancial accounts, will be made available byMember States from 2006 onwards.

7ECB


ANNUALQUAL ITYREPORTplan to enhance the collection of data on

portfolio investment, complemented by acentralised securities database10 which will beused to overcome the current difficulties. Forexample, the national compilers of b.o.p. andi.i.p. will use the same criteria from a commondatabase to classify the securities in terms ofmaturity and sector and residence of the issuer.In addition, the database will assist thecompilers when calculating the income onportfolio investment and will allow flexibilityin the compilation of statistics withoutincreasing the burden on respondents.

2 PERIODICITY AND TIMELINESS

The euro area b.o.p. statistics are published at amonthly frequency. Additional details onsector and instrument breakdowns have aquarterly frequency.

Together with the monthly release of thenon-seasonally adjusted data, the ECBpublishes the data resulting from the seasonaladjustment of the b.o.p. current account items.These data ease the interpretation of the latestdevelopments in the current account byremoving the seasonal pattern, as well asdifferences in working days and other calendareffects.

In 2004, the ECB has again met its releasecalendar for publication: monthly data werepublished seven weeks after the end of therespective months, thereby enabling anassessment of the quarterly and annual flowswithin two months (e.g. the full year 2003 waspublished on 24 February 2004).11 Furtherquarterly details were published four monthsafter the end of the respective quarters. Thei.i.p. was released eleven months after therespective ends of year.12

3 REVISION PRACTICE AND POLICY

The euro area b.o.p. is revised according to thefollowing schedule. Quarterly data are revised

with the publication of the following quarter’sdata and thereafter twice a year, at the end ofApril and the end of October. Monthly data arerevised with the publication of the followingmonth’s data, as well as with the revisions ofthe corresponding quarter. The annual i.i.p. isrevised with the publications of data for (atleast) the two subsequent years.

Revisions are necessary to improve the datacoverage, as first assessments of data may bebased in part on estimates due to late or error-prone responses by reporting agents, and toprovide users with more accurate data for timeseries analysis and forecasting. However, largeor biased revisions may signal weaknesses inthe data collection or compilation systems thatneed to be checked and corrected.

Since 2003, euro area and EU b.o.p. aggregates(the former is compiled by DG-Statistics, thelatter by Eurostat) have been revisedsimultaneously in stages, which have alsoenabled the publication of a reconciled euroarea i.i.p. This increases the comparabilityof the data, while also easing the reportingby Member States. Furthermore, a commonrevision policy is under preparation; bothpractices will then converge towards aEuropean framework revision policy.

4 STABILITY

The first release of the monthly b.o.p. for theeuro area occurs seven weeks after thereference period and is based on thecontributions sent by the national compilersone week earlier. The first release is revisedwhen new information becomes available.

10 See “Measuring Euro Area Portfolio Investment – Status FiveYears after Start of the Economic and Monetary Union andOutlook on Future Developments”, by P. Nendorfer; Paperprepared for the 28th General Conference of the InternationalAssociation for Research in Income and Wealth”, Cock,Ireland 22-28 August 2004.

11 The benchmark in the SDDS is three months.12 The benchmark in the SDDS is nine months. However, the

publication date will be reviewed in 2005 when someimprovements in the quality and the availability of quarterlyi.i.p. data have been made.


From an analytical perspective, users wish toknow how much they can rely on the initialassessment. Although the initial assessmentundergoes several revisions before reachingthe final stage, for practical reasons theindicators developed in Annex 1 assess thestability by analysing how close the firstassessments were to the final assessments.

In the first assessment of the b.o.p. currentaccount, goods and services may be based inpart on estimates in order to meet the deadline.For the time being, the results of the indicatorsof stability for the current account show thatrevisions were systematically positive, andtherefore the first estimates were revisedupwards for credits and debits (see charts inAnnex 2). These patterns largely offset eachother for the net b.o.p. current account whereno systematic pattern can be found.

Revisions to direct investment and portfolioinvestment assets and liabilities present a lesspronounced pattern. The balances did not show aspecific pattern.

Changes in the underlying collection of dataand/or methodological changes in one or a fewMember States, or in compiling the aggregate,may also lead to breaks in the series. While thisaffects previous analyses of the series, it alsoincreases the accuracy of the data and may beexpected to increase the stability of the seriesover time. For example, in early 2001, theaggregation method of the euro area incomedebits was changed, affecting the level of theseries in 1999 and 2000. Moreover, in early2003 one Member State introduced a newsurvey for travel, which triggered revisions tothe back series.

In contrast, the euro area i.i.p. with separateassets and liabilities was published only threetimes, in November 2002 (positions as at end-2001), November 2003 (positions as at end-2002) and November 2004 (positions as at end-2003). Therefore, the analysis of revisions islimited to the end-2001 and end-2002 data.

Owing to recent methodological work carriedout and agreements on direct investment and onportfolio investment (for both b.o.p. flows andi.i.p. stocks) reached by the ESCB StatisticsCommittee, assisted by the Working Group onExternal Statistics,13 new collection methodsare, or will be, implemented by Member Statesby 2007. The methods are designed to increasethe methodological soundness and consistencyof contributions to the euro area aggregate inthe medium term, but may also be a new sourceof revisions in the meantime. Also with regardto direct investment, for example, theInternational Accounting Standards will not beimplemented at the same pace and to the sameextent across Member States and amongcompanies, in particular for their individual(non-consolidated) accounts. This may also leadto some difficulties in data collection and to laterrevisions.

The main results of the stability indicators areincluded in the following sub-sections.

4.1 THE DIRECTIONAL RELIABILITY SHOWSA RELATIVE WEAKNESS IN DIRECTINVESTMENT

The directional reliability summarises howoften the first assessments were able tocorrectly predict a decrease or an increase ofthe final value with respect to the previousobservation. The indicator shows the lowestresults for the direct investment item. Aconsiderable part of direct investment iscomposed of reinvested earnings, which arebased entirely on estimates in the firstassessment of the data because no results ofcompanies are known at that time.

The direction of the month-to-month changesconstitutes a simple measure of stability, whichis applicable to all b.o.p. items. Chart 1

13 See reports of the Task Force on Foreign Direct Investment,ECB (March 2004), of the Task Force on Portfolio InvestmentCollection Systems, ECB (June 2002), and of the Task Forceon Portfolio Investment Income, ECB (April 2003).

9ECB


ANNUALQUAL ITYREPORT

contains the results of the indicator ondirectional reliability for the main items of theb.o.p.

4.2 THE MEAN ABSOLUTE PERCENTAGE ERRORSHOWS AN IMPROVEMENT FOR INCOME

The relative magnitude of revisions was largerfor services and income than for goods. Ingeneral, the initial assessments weresystematically lower than the finalassessments. The impact of the revisions hasbeen reduced in all the current account items,in particular in income debits, where theydecreased by more than 60%. The largerevisions in 1999 were related to the initialcompilation of euro area statistics (in 1999countries compiled data according to the euroarea requirements for the first time).

The mean absolute percentage error (MAPE)was calculated for the gross series of the euroarea current account. The MAPE is equal to theaverage of the absolute revisions in relation tothe size of the respective flow. Chart 2 contains

the results for 1999-2001, 2000-2002 and2001-2003.

4.3 THE IMPROVEMENT IN THE CURRENTACCOUNT IS CONFIRMED BY THE ROOTMEAN SQUARE RELATIVE ERROR

For the net items of the current account and forthe financial account, another type of indicatorwas used: the root mean square relative error(RMSRE). The RMSRE measures the distancebetween the first assessment and the finalassessment in relation to the volatility of eachtime series. This denominator tries tocompensate for the relative difficulty ofestimating a more volatile series. The volatilityof each series was estimated by its standarddeviation, based on the assumption that a seriesfluctuates around the series average.14

Charts 3 and 4 contain the results for 1999-2001, 2000-2002 and 2001-2003. In the current

14 The assumption of stationarity was conf irmed by standardstatistical tests.

Char t 1 Overv i ew o f d i r e c t iona lre l i ab i l i t y

(as a percentage; Jan. 2001-Dec. 2003)

50

60

70

80

90

100

50

60

70

80

90

100

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

1 Goods credits 2 Goods debits 3 Goods net 4 Services credits 5 Services debits 6 Services net 7 Income credits 8 Income debits 9 Income net 10 Total current account credits11 Total current account debits

12 Total current account net13 Direct investment assets14 Direct investment liabilities15 Direct investment net16 Portfolio investment assets17 Portfolio investment liabilities18 Portfolio investment net19 Other investment assets20 Other investment liabilities21 Other investment net

Char t 2 Rev i s i ons o f the cur rent accountas % o f the re spec t i ve f l ow (MAPE)

January 1999-December 2001January 2000-December 2002January 2001-December 2003

0

2

4

6

8

10

12

0

2

4

6

8

10

12

Goodscredits

Goodsdebits

Servicescredits

Incomecredits

Totalcurrentaccountcredits

Servicesdebits

Incomedebits

Totalcurrentaccountdebits


account, the revisions have been reducing overtime.

The net items of the current account(Chart 3) present higher percentages than theitems of the financial account (Chart 4). Theseresults are not due to larger revisions as shownin the average of the revisions (see R

_ and |R

_|

in Annex 2) but to the lower volatility of the

former (used as reference for the calculation ofthe indicator). The impact of the revisions hasalso been reduced in all the financial accountitems except other investment and portfolioinvestment liabilities. Part of the revisions inother investment liabilities was related to theintroduction of a new estimation method for thebanknotes held by non-residents.

Total assets Direct investment Portfolio investment Other investment

Initial assessment 7,277.9 1,937.5 2,270.4 2,581.3

Second assessment 7,260.6 1,877.4 2,302.6 2,578.6

Revision -17.3 -60.1 32.2 -2.7

Revision as % -0.2% -3.2% 1.4% -0.1%

Tab le 1 Rev i s i ons to 2002 i . i . p . a s se t s

(EUR billions)

Char t 4 Rev i s i ons o f the f i nanc i a la c count a s % o f vo l a t i l i t y


0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

1 Direct investment assets2 Direct investment liabilities3 Direct investment net4 Portfolio investment assets5 Portfolio investment liabilities

1 2 3 4 5 6 7 8 9

6 Portfolio investment net7 Other investment assets8 Other investment liabilities9 Other investment net

Char t 3 Rev i s i ons o f the net cur rentaccount a s % o f vo l a t i l i t y


0102030405060708090

0102030405060708090

Netgoods

Netservices

Netincome

Net totalcurrent account

Total liabilities Direct investment Portfolio investment Other investment

Initial assessment 7,567.5 1,512.5 3,026.7 2,897.6

Second assessment 7,878.6 1,673.2 3,181.6 2,875.9

Revision 311.1 160.7 154.9 -21.7

Revision as % 3.9% 9.6% 4.9% -0.8%

Tab le 2 Rev i s i ons to 2002 i . i . p . l i ab i l i t i e s

(EUR billions)

11ECB


ANNUALQUAL ITYREPORTIn spite of high relative revisions of assets and

liabilities, the net direct investment and netportfolio investment were somewhat lessrevised. Some of the revisions of the euro areatransactions were due to transactions by specialpurpose entities which, although high in value,were neutral with regard to the balance. Inaddition, the revision of the initial, error-proneestimates for reinvested earnings affected boththe assets and the liabilities of directinvestment.

Moreover, owing to the special aggregationmethod followed for the liabilities of portfolioinvestment, revisions of the extra/intra share ofassets within the country contributions had animpact both on net assets and on net liabilities,while the balance (as the sum of nationalbalances) remained unchanged.

4.4 STABILITY OF THE INTERNATIONALINVESTMENT POSITION

Tables 1 and 2 summarise the revisions ofthe main items of the euro area i.i.p. Therevisions of the total asset positions as atend-2002 amounted to -€17 billion, whichrepresents 0.2% of the total assets. On theliabilities side, the corresponding revisionswere €311 billion (3.9% of the total liabilities).The highest revisions in relative terms were ondirect investment liabilities (9.6%).

The revisions of direct investment were mainlyrelated to the introduction of a new survey inone Member State, which served to bettermeasure activities of special purpose entities.The revisions on the net i.i.p. amount to€328 billion, or 4.5% of GDP.

5 CONSISTENCY

Consistency indicators deal with two aspects:internal inconsistency, as revealed by the itemon errors and omissions; and externalinconsistency, as revealed by discrepancies

vis-à-vis other statistics on the same variable,such as trade statistics and flows derived fromthe balance sheets of MFIs.

With respect to the enhanced recording ofspecial purpose entities, overall consistencywith earlier periods, or with transactionsand positions as reported by other MemberStates has not yet been fully ensured.Discussions are ongoing to remove legalimpediments to an exchange of confidentialinformation across the ESCB that may greatlyassist in this area.

5.1 INTERNAL CONSISTENCY

The b.o.p. has a natural indicator for internalinconsistency: the residual item called “errorsand omissions”. In fact, the principle ofdouble-entry bookkeeping implies that the sumof all transactions vis-à-vis the rest of the worldshould be equal to zero. The size of errors andomissions is a lower bound of the relativeinaccuracy of the b.o.p. In any case, a large orpersistent residual may impede the analysis orinterpretation of the b.o.p. However, errors andomissions close to zero may indicate that thetransactions covered are balanced but this isnot conclusive proof of the full accuracy of theb.o.p.

Char t 5 RMSE o f e r ror s and ommis s ionsby v in tage

(EUR billions; 1999-2003)

variancebias

0

5

10

15

20

0

5

10

15

20

1 2 3Vintage

4 5


The root mean square error indicator (RMSE)was calculated from the time series on errorsand omissions. This indicator is used tomeasure the size of the internal inconsistency,as well as to identify any potential bias.

In the period January 2001 to December 2003,the errors and omissions showed no bias anddid not prove to be significantly different fromzero, according to a standard statistical test.The RMSE was €15.1 billion, which amountedto 5% of the average gross flows in the currentaccount during that period.

Chart 5 contains the results for the RMSE byvintage, for the five years 1999-2003. The firstvintage represents the initial assessment of thedata. The second vintage contains the dataresulting from the revision with the publicationof quarterly data. The remaining vintages referto subsequent yearly revisions. The evolutionof the indicator along the vintages indicatesthat revisions broadly increase the internalconsistency of the b.o.p.

5.2 EXTERNAL CONSISTENCY

The b.o.p. series are compared with thecorresponding data published by Eurostat foreuro area external trade, and with the externaltransactions derived from the balance sheetscompiled in the context of MFI balance sheet

statistics. Although the methodologies ofthose series are not fully consistent with theb.o.p., they broadly reflect the same economicphenomenon. Therefore, the comparisons areuseful for checking whether differences remainstable over time.

Table 3 shows the results of the indicators fordirectional reliability of goods in b.o.p. andexternal trade in the period from January 2001to December 2003 compared with the resultsobtained in the periods from January 1999 toDecember 2001 and from January 2000 toDecember 2002. For the three periods, thedirection of the month-on-month changes wasthe same in more than 90% of cases.

In Table 4, the indicator of differences betweengoods in b.o.p. and in external trade shows thatthe average absolute difference fluctuatedaround 2% for exports and around 5% forimports with a slight improvement in the lastperiod. This indicator is measured in relation tothe average of the discrepancies between thefirst differences of the two series concerned.The differences were systematic.15

Moreover, Table 5 contains the results for theaverage absolute differences between the

Period Exports Imports

1999-2001 2.2 5.1

Average of absolute differences 2000-2002 2.2 5.1

2001-2003 1.8 5.0

Tab le 4 Goods in b .o .p . and exte rna l t rade

(differences as % of average value)

15 As shown by the average of simple values that are identical tothe average of absolute values.


1999-2001 97.1 97.1

Directional reliability 2000-2002 94.3 100.0

2001-2003 97.1 100.0


(percentages)

13ECB


ANNUALQUAL ITYREPORT

growth rates of each series. The indicatorsshow that the average of the absolutedifferences between growth rates was reducedin 2001-2003, both for exports and imports. Nosystematic difference can be found in the sameperiod.

In Table 6, the root mean square relative error(RMSRE) reflects the distance betweencomparable b.o.p. and money and bankingstatistics in relation to the volatility of theb.o.p. series. This indicator increased in the2001-2003 period, while the bias componentreduced to close to zero.

6 CONCLUSION

In the first assessment of the b.o.p. currentaccount, full information on goods and servicesis not yet available and it is often necessary touse estimates to meet the deadlines. For thetime being, the results of the indicators ofstability for the current account show that thesefirst estimates systematically underestimatecredits and debits. Nonetheless, these patternsseemingly do not affect the net current account.

Regression Unsystematic

Period RMSRE Bias component component component(%) (%) (%) (%)

1999-2001 15.6 3.4 0.8 95.8

2000-2002 9.4 1.3 2.0 96.7

2001-2003 10.1 0.2 0.3 99.4

Tab le 6 Depos i t s / l oans o f MF I s - compar i son w i th M&B data

In general, the picture shown in the euro areab.o.p. appears credible and the quantitativeindicators show an improvement since 1999.The results for 1999 and 2000 reflect the factthat national b.o.p./i.i.p. compilers were still inthe process of adapting their data collectionsystems to cover the needs of euro areastatistics and, hence revisions were large.While the levels for goods in b.o.p. and inexternal trade statistics show differences due tothe underlying methodologies, the differencesin their respective month-to-month growthrates decreased from 0.9 percentage point forthe period 1999-2001 to 0.7 and 0.6 percentagepoint for exports and imports respectively forthe period 2001-2003.


1999-2001 0.9 0.9

Average of absolute differences 2000-2002 0.8 0.7

2001-2003 0.7 0.6

1999-2001 0.1 0.1

Average of simple differences 2000-2002 0.1 0.0

2001-2003 0.1 0.0


(percentage points; month-to-month growth rate)


A N N E X 1 M E T H O D O L O G I C A L D O C U M E N TAT I O NFOR QUA L I T Y I ND I C ATOR S 1

This annex contains the methodology usedfor the quantitative indicators developed toassess reliability/stability and serviceability/consistency.

1 RELIABILITY/STABILITY

In the IMF’s terminology, the study ofrevisions is normally referred to as reliability,while some quality works at the European levelrefer to this as stability. The underlyingconcept is, however, the same and can bedefined as referring “to the closeness of theinitial estimated value(s) to the subsequentestimated values.”2 Assessing reliabilityinvolves comparing estimates over time. Inother words, assessing reliability refers torevisions.

The number of revisions observed depends onthe revision policy/practice of a statisticalagency, which normally decides beforehand(sometimes in collaboration with the users)how many times and when the estimates shouldbe revised and communicated to the public.

For example, with reference to a series X withN observations, the statistical agency candecide to publish it k times at predefined timelags {l

1, l

2, ..., l

k}, where the time lag indicates

the time elapsed between the reference periodand the publication period (e.g. if a Junepublication refers to a revision of January data,the time lag is thus five months). Hence kdifferent sets {X(l

1), X(l

2), …, X(l

k)} of the same

variable-series X will be available.Simplifying the notation, the observation tpublished after a time lag i can be representedas

itx .

From the previous k sets of data, revisions canbe easily derived, normally as the differencebetween two subsequent assessments.Therefore a revision variable or series can bedefined as the difference ijij XXR −= , where iand j identify two specific time lags, with j > i.The Task Force on Quality (TF-QA) suggestedmeasuring revisions by the difference between

the first and latest assessments owing tosimplicity and data availability, R = X

k– X

1.

Revisions may also be calculated over atransformation of the original series, such asthe respective first difference or the growthrate, if the transformation is meaningful fromthe user’s perspective.

1.1 SIMPLE MEASURES OF REVISIONS

Simple indicators of revisions express thechanges in original units of the variable X. Thatbrings simplicity to their interpretation butmight hamper the comparison of indicators fordifferent series.

An average of the revisions (R) provides anindication of how far the first assessment wasfrom the latest assessment. The average mayalso be computed on the basis of the absoluterevisions ( R ), which adds a measure ofdispersion. Dispersion can also be assessed bythe standard deviation of revisions.

When useful, the simple indicators of revisionscan also be calculated on the basis of a standardtransformation of the series, such as the growthrate.

1.2 RELATIVE MEASURES OFREVISIONS

Simple calculations of revisions express thechanges in original units of the variable X anddepend on its magnitude, often hamperingcomparability across time, across differentseries and across similar series for differentcountries. Therefore, it is often useful toprovide relative measures which relate therevisions to dimensional measures of thevariable.

1 Based on the report by the joint ECB DG-S/EuropeanCommission (Eurostat) Task Force on Quality.

2 IMF working paper – Statistics Department: “AssessingAccuracy and Reliability: A Note Based on Approaches Usedin National Accounts and Balance of Payments Statistics”,Carson, Carol S. and Lucie Laliberté, February 2002.

15ECB


ANNEX 1

GROSS DATAIn the case of gross data (data which expresspositive quantities), it is straightforward todevise a relative measure of revisionsexpressed in terms of percentage changes from

the initial assessment as 1XRP = , which is

called the percentage revision. As X is a timeseries, an average can be taken across time, P ,the mean percentage error (MPE).

As revisions can be positive or negative, it isusually appropriate to take the absolute value,in order to avoid revisions of opposite signscancelling out in the resulting indicator. So, ifthe average is calculated using the absolutevalues, we get P , the mean absolutepercentage error (MAPE).

As an alternative to the MAPE, the relativeindicators of revisions can also be calculatedafter applying a standard transformation. In thecase of the gross series, the transformation canapply to the b.o.p. current account, a growthrate or a first simple difference (to minimise thetrend), and/or a first seasonal difference (tominimise the seasonal pattern).

As the transformed series contain positive andnegative observations, the MAPE would nolonger be applicable in a direct way becauseobservations can be near zero, and this wouldinflate the indicator. The indicators for net datain the following paragraphs were developed toovercome this.

NET DATAIn the case of net data, the revisions cannot berelated to underlying quantities if therespective series contain observations close tozero because the ratio would inflate theindicator. Therefore, if the calculation of theMAPE overestimates the size of revisions,alternative dimensional measures of thevariable X may be used. A solution to this maybe to choose a measure of the volatility of seriesX for assessing the relative size of the revision.

Ideally, the computation of volatility should beperformed on the basis of the non-deterministiccomponent of the series. This is because theinformation content of a series depends moreon the non-deterministic component, whichchanges from observation to observation, thanon the deterministic component, which reflectsmore structural aspects. Therefore, theindicators of revisions should penalise less therevisions of series whose non-deterministiccomponent portrays high volatility.

The mean absolute relative error (MARE) isthen defined as

)( kXvolR

There are several ways of calculating thevolatility of X. The first issue to be decided ison which series the volatility should becalculated. In principle, the volatility should becalculated on the latest assessment X

k, because

it is potentially the most accurate. A secondissue is to decide about how to measurevolatility. Three options are suggested:

1. Standard deviation. The standard deviationis the classic measure of volatility of aseries and represents the square root of theaverage of the quadratic distances from themean:

∑ ∑= =

−=

N

t

N

ttt

k k

kkx

Nx

NS

1

2

1

11

2. Average distance from the mean. Theaverage distance from the mean representsthe average of the distances from the meanin absolute values. It has the advantage ofexpressing the volatility in the originalunits (not distorted by the application of thesquares).

∑ ∑= =

−=N

t

N

ttt

k k

kkx

Nx

ND

1 1

11

3. Median of distances from the median. Themedian of distances from the median issimilar to the formula in point 2, where theaverages are replaced by medians owing tothe fact that medians are less susceptible tothe presence of outliers.


Using the second option, we obtain

DR

MARE =

UPWARD REVISIONSIn principle, positive and negative revisionsshould occur with roughly the same frequency.If the revisions are systematically positive ornegative, the underlying reasons for this shouldbe analysed. This may be a lack of coverage inearly estimates, and there should be an attemptto correct such a systematic bias. This simpleindicator is the ratio between upward revisionsand the number of observations (N).

upward revisions = (# upward revisions) / N

DIRECTIONAL RELIABILITYTo assess whether the information contained inthe earlier estimates has been altered by therevisions, a 2 x 2 contingency table can be setup. In this contingency table the columnsconsist of positive and negative firstdifferences in the early estimates

111 1 )t(ttxxx −−=∆ , while the rows consist of

positive and negative changes in the latestvalues

kkk )t(ttxxx 1−−=∆ .

there is a total dissociation ( 02211 =+ nn ). Thedirectional reliability indicator (Q) expressesthe percentage of cases in which earlier andlater assessments move in the same direction.High values are optimal in terms of increasingthe reliability of the data.

1.3 DECOMPOSABLE INDICATORS

Other measures are used to detect possible biasor persistent patterns in the revisions.According to the literature on forecast qualitymeasures, the most appropriate for the analysisof revisions is based on mean square errors,which allow a decomposition into bias andvariance.

The indicator is very similar to the MARE but itis calculated as the square root of a ratiobetween average squares of revisions and thevariance of the series. It is designated as rootmean square relative error (RMSRE):

2

2

SRRMSRE =

The value of the RMSRE is 0 when the forecastis perfect, 1 if the forecast is only as accurate asthe reference forecast, and greater than 1 whenthe forecast is less accurate than the average forthe series.3 The square of the RMSRE can bedecomposed as:

[ ]22

1

2

12 )(111 XX

k

XX

k

kkkr

XSXSr

XSXXRMSRE −+

−+

−=

where 1XX k

r is the correlation between the twoseries and

kXS and

1XS are the respective

standard deviations.

3 Other measures of distribution location, like the median andthe trimmed mean, were tested. Assuming that the b.o.p.f inancial net flows are stationary, the average was chosenowing to its simplicity, ease of interpretation, and because itmakes the decomposition of the indicator possible. Althoughnot implemented by the TF-QA, when the series are notstationary the indicator can still be applied using the previousvalue of the series as the reference value, or using the f irstdifference of the series.

∆χ t1 > 0 ∆χ t1 0 n11 n12 n11+n12∆χ tk

17ECB


ANNEX 1

The three components of the square of theRMSRE can also be presented as proportions ofthe RMSRE, which added together will equal 1.

[ ]2

2

2

2

1

2

2

1

2

2 )(11 1

1

RMSREr

RMSREXSXSr

RMSREXSXX

RMSRERMSRE XXk

XX

k

k

k

k

−+

−

+

−

==

The three components can be interpreted asfollows:

1. The unconditional or bias component is anindication of systematic error (revision),since it measures the extent to which theaverage values of the early and laterassessment series deviate from each other.The revisions can be considered biased ifthe mean revision is significantly differentfrom zero.4

2. The conditional or regression component isanother systematic component whichreflects whether the overall pattern of theseries with the early estimates was close tothat of the series with the later estimates. Ifthe forecast correctly reflects the pattern/volatility of the later estimate series, thecorrelation between both series will bequite high and the component close to zero.

3. The unsystematic or disturbancecomponent is the variance of the residualsobtained by regressing the early estimatesdata on the later estimates. It can beassumed to have a random nature withoutany predictable pattern.5

GENERAL INDICATORThe indicators for net flows (MARE andRMSRE) can be expressed in a general formulaas follows:

ρ

ρ

ρ

ρ

1

1

1)(

−Θ=∑

∑

=

=N

ttt

N

t

kx

RMRE

where ρ is the power parameter ( e.g. ρ = 1 forMARE and ρ = 2 for RMSRE).

Instead of the average value of xk, Θ

j can be set

as a forecast to test against the earlyassessments. For instance, it could be theforecast of a benchmark model used to evaluatethe time series.

As explained before, the average is very usefulfrom an analytical point of view, but in the caseof non-stationary series, the average and theMSE decomposition become meaningless. Inthis case, the appropriate use of a relevanttransformation is of utmost importance topreserve the analytical decomposition of theindicator. If the transformation is not feasible,a different reference forecast will be moreconvenient.

2 SERVICEABILITY/CONSISTENCY

In the IMF’s DQAF, consistency is defined aseither (i) over time; (ii) between data collectedat different frequencies; (iii) internationally;or (iv) across variables, vertically (acrosstransactions), horizontally (across institutionalsectors), and/or between flows and stocks. TheTF-QA decided to follow the IMF’s DQAF forb.o.p. 2001 principles by mostly concentratingon this element and dividing it into thefollowing sub-categories:

– internal consistency (e.g. within theintegrated statistics (b.o.p./i.i.p. or nationalaccounts));

– consistency over time (e.g. in the case ofmethodological or institutional changes,such as enlargement, historical data arereconstructed as far back as is reasonable);and

– external consistency (between differentsources of data and/or different statisticalframeworks, including mirror statistics –international statistics should becomparable even when compiled by

4 Normality is assumed for revisions in order to apply the t test;it needs to be tested statistically.

5 This indicator only accounts for linear relationships. Theunsystematic part could still have non-linear patterns within it.


different institutions or by different units ofthe same institution).

2.1 INTERNAL CONSISTENCY

The b.o.p. statistics have a natural indicatorfor internal consistency: the net errors andomissions series (EO). The principle of double-entry bookkeeping used in b.o.p. implies thatthe sum of all international transactions shouldbe equal to zero. Nevertheless, “Data forbalance of payments estimates often arederived independently from different sources;as a result, there may be a summary net creditor net debit (i.e., net errors and omissions inthe accounts). A separate entry, equal to thatamount with the sign reversed, is then made tobalance the accounts. Because inaccurate ormissing estimates may be offsetting, the size ofthe net residual cannot be taken as an indicatorof the relative accuracy of the balance ofpayments statement.6 Nonetheless, a large,persistent residual that is not reversed shouldcause concern. Such a residual impedesanalysis or interpretation of estimates anddiminishes the credibility of both. A large netresidual may also have implications forinterpretation of the investment positionstatement” (IMF, BPM5, 1993, p.17).

According to the IMF’s DQAF for b.o.p. 2001,internal consistency implies checking that“over the long run [the] errors and omissionsitem (1) has not been large and (2) has beenstable over time”.

A measure of the size can be provided by theaverage of the absolute errors and omissions,EO .

As with revisions, an alternative measure of thesize can be provided by the root mean squareerror of the net errors and omissions (RMSE(EO)).

2)( EOEORMSE =

As before, this indicator can be decomposedinto bias and variance components:7

22

2

2

SEORMSE

componentvariancecomponentbiasRMSE

+=

+=

where S is the standard deviation of the errorsand omissions.

Further to the previous indicator, the number ofpositive EOs divided by the number ofobservations can be used to assess the relativefrequency of positive EOs:

NEOCountEOCP t )0()( >=

where N is the timeframe data.

2.2 EXTERNAL CONSISTENCY

Although minor discrepancies arising frommethodological differences can still be presentin two sets of data stemming from differentsources and/or different statisticalframeworks,8 the comparison can still provide auseful measure of consistency and contribute tothe overall increase in quality.

GROSS FLOWS OR STOCKSSimple indicators of external consistencyexpress the differences in original units of thevariables under comparison, X and Y. A simpleindicator measuring the consistency betweenb.o.p. and international trade statistics (ITS)can be computed on the latest assessment ofeach series.

An average of the differences (D ) provides anindication of how far the series deviate onaverage. The average may also be computed onthe absolute differences ( D ).

Yet another alternative indicator is similar tothe MAPE ( )P , but with the percentagedifferences calculated as proportions of an

6 Therefore, the net errors and omissions constitute a lowerbound for overall inaccuracies.

7 Following the simplest MSE decomposition. See Elements ofForecasting, Diebold, Francis X. 2001

8 E.g., the comparison between the euro area goods item (b.o.p.)and Eurostat’s external trade data, or the comparison betweenthe b.o.p. flows of the monetary f inancial institution (MFI)sector and flows derived from the consolidated MFI balancesheet from money and banking statistics.

19ECB


ANNEX 1

average of the two time series.9 The indicatorcaptures the magnitude of the discrepancies inabsolute value, and frames it in proportion tothe average size of the series.

As with revisions, it is also possible to computerelative indicators of external consistency overthe transformed series. For example, the impactof different levels in the series of imports canbe removed by calculating the indicator on thefirst differences.

Another simple measure is based on theaverage differences of the series aftertransforming them into growth rates. This hasthe advantage of abstracting from differencesin levels between the time series: the imports ofgoods are measured on a c.i.f. basis for ITS andon an f.o.b. basis for b.o.p., while both exportsare measured on an f.o.b. basis. A simpleindicator of external consistency thenbecomes:

yx GGG −=

NET FLOWSDifferences between b.o.p. transactions andsimilar transactions derived from the MFIbalance sheet can be attributed to a variety offactors: dissimilar timeliness in terms ofrecording and reporting, different revisionpolicies and different valuation methods.

It is proposed that the relative indicators forassessing reliability will also be used forassessing consistency between comparablenet flows. The indicator similar to theMAPE, “C”10, the MARE and the RMSRE arecalculated over the latest assessment of eachseries. The volatility can be assessed over themean of both series.

DIRECTIONAL CONSISTENCYNo less important is the consistency of theinformation provided by the two sources, i.e. ifthe signs of the first differences coincide inboth sources. In the contingency table forexternal consistency, the columns are thepositive and negative changes for b.o.p. series

( ), and the rows are thepositive and negative changes for the mirrorseries ( )

9 ( )∑−= +−

=T

aTt tt

tt

yxyx

aC

2

1

Based on Some elements of a quality framework for CMFBstatistics, Keuning, S. and Algera, S.

10 As with the MAPE, the “C” indicator may become inflated inthe presence of observations close to zero.

Subtotal

n11

n12

n11

+n12

n21

n22

n21

+n22

Subtotal n11

+n21

n12

+n22

N

Cont ingency tab l e fo r d i r e c t iona lcon s i s t ency

where n11

is the number of cases when both ∆xt

and ∆yt are positive, and n

22 the number of cases

when they are negative.

For a maximum directional consistency, onewould expect a high sum for the main diagonal(n

11+ n

22). As before, the directional

consistency indicator (QC) is constructed as

follows:

NnnQC 2211

+=

1−−=∆ ttt xxx

1−−=∆ ttt yyy

0>∆ ty0≤∆ ty

0>∆ tx 0≤∆ tx


ANNEX 2

Chart 1 Euro a rea goods – c red i t s

(EUR billions)

-100

102030405060708090

100

-100102030405060708090100

revision first assessment final assessment

Jan. July1999

Jan. July2000

Jan. July2001

Jan. July2002

Jan. July2003

Char t 2 Euro a rea goods – deb i t s

(EUR billions)


Jan. July1999

Jan. July2000

Jan. July2001

Jan. July2002

Jan. July2003

-100

102030405060708090

100

-100102030405060708090100

Char t 3 Euro a rea goods – net

(EUR billions)


Jan. July1999

Jan. July2000

Jan. July2001

Jan. July2002

Jan. July2003

-10

-5

0

5

10

15

20

-10

-5

0

5

10

15

20

Tab le 1 S tab i l i t y i nd i ca tor s f o r goods

Quality Reference Goods

indicator period Credits Debits Net

R Jan99 - Dec01 1.53 3.17 -1.64

Jan00 - Dec02 1.06 1.95 -0.90

Jan01 - Dec03 0.60 0.72 -0.12

R Jan99 - Dec01 1.88 3.60 2.11

Jan00 - Dec02 1.50 2.44 1.84

Jan01 - Dec03 1.16 1.28 1.18

Jan99 - Dec01 0.03 0.05 0.56

Jan00 - Dec02 0.02 0.03 0.51

Jan01 - Dec03 0.01 0.02 0.32

Q Jan99 - Dec01 97.14% 94.29% 88.57%

Jan00 - Dec02 97.14% 94.29% 88.57%

Jan01 - Dec03 100.00% 94.29% 88.57%

MAPE/

RMSRE

21ECB


ANNEX 2

Tab le 2 S tab i l i t y i nd i ca tor s f o r s e rv i ce s

Quality Reference Services


R Jan99 - Dec01 1.94 2.13 -0.19

Jan00 - Dec02 1.81 1.60 0.21

Jan01 - Dec03 1.39 0.93 0.46

R Jan99 - Dec01 1.94 2.13 0.75

Jan00 - Dec02 1.83 1.60 0.80

Jan01 - Dec03 1.41 1.02 0.75

Jan99 - Dec01 0.09 0.10 0.57

Jan00 - Dec02 0.08 0.07 0.59

Jan01 - Dec03 0.06 0.04 0.56

Q Jan99 - Dec01 91.43% 82.86% 85.71%

Jan00 - Dec02 94.29% 88.57% 77.14%

Jan01 - Dec03 88.57% 91.43% 82.86%

MAPE/

RMSRE

Char t 5 Euro a rea se rv i ce s – deb i t s

(EUR billions)


Jan. July1999

Jan. July2000

Jan. July2001

Jan. July2002

Jan. July2003

-5

0

5

10

15

20

25

30

35

-5

0

5

10

15

20

25

30

35

Char t 6 Euro a rea se rv i ce s – net

(EUR billions)


Jan. July1999

Jan. July2000

Jan. July2001

Jan. July2002

Jan. July2003

-10

-5

0

5

10

-10

-5

0

5

10

Char t 4 Euro a rea se rv i ce s – c red i t s

(EUR billions)


Jan. July1999

Jan. July2000

Jan. July2001

Jan. July2002

Jan. July2003

-5

0

5

10

15

20

25

30

35

-5

0

5

10

15

20

25

30

35


Char t 7 Euro a rea income – c red i t s

(EUR billions)


Jan. July1999

Jan. July2000

Jan. July2001

Jan. July2002

Jan. July2003

-5

0

5

10

15

20

25

30

35

-5

0

5

10

15

20

25

30

35

Char t 8 Euro a rea income – deb i t s

(EUR billions)


Jan. July1999

Jan. July2000

Jan. July2001

Jan. July2002

Jan. July2003

-5

0

5

10

15

20

25

30

35

-5

0

5

10

15

20

25

30

35

Char t 9 Euro a rea income – net

(EUR billions)


Jan. July1999

Jan. July2000

Jan. July2001

Jan. July2002

Jan. July2003

-15

-10

-5

0

5

10

15

-15

-10

-5

0

5

10

15

Tab le 3 S tab i l i t y i nd i ca tor s f o r i n come

Quality Reference Income


R Jan99 - Dec01 0.71 1.43 -0.72

Jan00 - Dec02 0.53 0.45 0.08

Jan01 - Dec03 -0.16 -0.40 0.24

R Jan99 - Dec01 1.54 2.14 1.44

Jan00 - Dec02 1.38 1.24 0.93

Jan01 - Dec03 0.91 0.86 0.90

Jan99 - Dec01 0.08 0.11 0.64

Jan00 - Dec02 0.06 0.05 0.43

Jan01 - Dec03 0.04 0.03 0.43

Q Jan99 - Dec01 85.71% 85.71% 88.57%

Jan00 - Dec02 85.71% 88.57% 88.57%

Jan01 - Dec03 82.86% 94.29% 94.29%

MAPE/

RMSRE

23ECB


ANNEX 2

Chart 11 Euro a rea cur rent account –deb i t s

(EUR billions)


Jan. July1999

Jan. July2000

Jan. July2001

Jan. July2002

Jan. July2003

-20 -200

20406080

100120140160

020406080100120140160

Chart 12 Euro a rea cur rent account –net

(EUR billions)


Jan. July1999

Jan. July2000

Jan. July2001

Jan. July2002

Jan. July2003

-20

-15

-10

-5

0

5

10

15

20

-20

-15

-10

-5

0

5

10

15

20

Tab le 4 S tab i l i t y i nd i ca tor s f o r cur renta c count

Quality Reference Current account


R Jan99 - Dec01 4.48 7.47 -2.99

Jan00 - Dec02 3.76 4.80 -1.04

Jan01 - Dec03 2.14 1.93 0.21

R Jan99 - Dec01 4.59 7.68 3.86

Jan00 - Dec02 3.87 5.02 2.83

Jan01 - Dec03 2.37 2.32 2.05

Jan99 - Dec01 0.04 0.07 0.79

Jan00 - Dec02 0.03 0.04 0.62

Jan01 - Dec03 0.02 0.02 0.46

Q Jan99 - Dec01 88.57% 94.29% 82.86%

Jan00 - Dec02 94.29% 97.14% 91.43%

Jan01 - Dec03 94.29% 97.14% 82.86%

MAPE/

RMSRE

Chart 10 Euro a rea cur rent account –c red i t s

(EUR billions)


Jan. July1999

Jan. July2000

Jan. July2001

Jan. July2002

Jan. July2003

-200

20406080

120100

140160

-20020406080

120100

140160


Char t 13 Euro a rea d i r e c t i nves tmentabroad

(EUR billions)


Jan. July1999

Jan. July2000

Jan. July2001

Jan. July2002

Jan. July2003

-140-120-100

-80-60-40-20

02040

-140-120-100-80-60-40-2002040

Char t 14 D i rec t i nves tment in the euroarea

(EUR billions)


Jan. July1999

Jan. July2000

Jan. July2001

Jan. July2002

Jan. July2003

-50

0

50

100

150

200

-50

0

50

100

150

200

Char t 15 Euro a rea d i r e c t i nves tment –net

(EUR billions)


Jan. July1999

Jan. July2000

Jan. July2001

Jan. July2002

Jan. July2003

-50

-25

0

25

50

75

100

125

150

-50

-25

0

25

50

75

100

125

150

Tab le 5 S tab i l i t y i nd i ca tor s fo r d i r e c ti nve s tment

Quality Reference Direct investment

indicator period Assets Liabilities Net

R Jan99 - Dec01 -10.40 10.21 -0.20

Jan00 - Dec02 -8.37 8.23 -0.14

Jan01 - Dec03 -5.73 5.99 0.26

R Jan99 - Dec01 11.09 11.18 5.41

Jan00 - Dec02 10.68 10.42 5.33

Jan01 - Dec03 8.31 8.48 5.19

RMSRE Jan99 - Dec01 0.96 0.83 0.30

Jan00 - Dec02 0.82 0.67 0.29

Jan01 - Dec03 0.70 0.57 0.30

Q Jan99 - Dec01 77.14% 62.86% 77.14%

Jan00 - Dec02 80.00% 65.71% 85.71%

Jan01 - Dec03 88.57% 57.14% 85.71%

25ECB


ANNEX 2

Char t 17 Euro a rea por t fo l i oinves tment – l i ab i l i t i e s

(EUR billions)


Jan. July1999

Jan. July2000

Jan. July2001

Jan. July2002

Jan. July2003

-80-60-40-20

020406080

100

-80-60-40-20020406080100

Char t 18 Euro a rea por t fo l i oinves tment – net

(EUR billions)


Jan. July1999

Jan. July2000

Jan. July2001

Jan. July2002

Jan. July2003

-200

-150

-100

-50

0

50

100

-200

-150

-100

-50

0

50

100

Char t 16 Euro a rea por t fo l i oinves tment – a s se t s

(EUR billions)


Jan. July1999

Jan. July2000

Jan. July2001

Jan. July2002

Jan. July2003

-100

-80

-60

-40

-20

0

20

40

-100

-80

-60

-40

-20

0

20

40

Tab le 6 S tab i l i t y i nd i ca tor s fo r por t fo l i oi nve s tment

Quality Reference Portfolio investment


R Jan99 - Dec01 -4.19 4.86 0.68

Jan00 - Dec02 -3.28 5.11 1.82

Jan01 - Dec03 -2.17 5.02 2.85

R Jan99 - Dec01 5.96 7.69 7.72

Jan00 - Dec02 5.25 7.29 7.26

Jan01 - Dec03 4.44 9.09 7.41

RMSRE Jan99 - Dec01 0.54 0.38 0.32

Jan00 - Dec02 0.47 0.33 0.29

Jan01 - Dec03 0.39 0.45 0.32

Q Jan99 - Dec01 74.29% 94.29% 91.43%

Jan00 - Dec02 88.57% 94.29% 91.43%

Jan01 - Dec03 91.43% 82.86% 85.71%


Char t 19 Euro a rea other i nves tment –a s s e t s

(EUR billions)


Jan. July1999

Jan. July2000

Jan. July2001

Jan. July2002

Jan. July2003

-100

-50

0

50

100

150

-100

-50

0

50

100

150

Char t 20 Euro a rea other i nves tment –l i ab i l i t i e s

(EUR billions)


Jan. July Jan. July Jan. July Jan. July Jan. July-100

-75-50-25

0255075

100125150

-100-75-50-250255075100125150

1999 2000 2001 2002 2003

Char t 21 Euro a rea other i nves tment –net

(EUR billions)


Jan. July Jan. July Jan. July Jan. July Jan. July1999 2000 2001 2002 2003

-80

-60

-40

-20

0

20

40

60

80

-80

-60

-40

-20

0

20

40

60

80

Tab le 7 S tab i l i t y i nd i ca tor s fo r o therinve s tment

Quality Reference Other investment


R Jan99 - Dec01 0.21 1.37 1.58

Jan00 - Dec02 -0.68 2.28 1.61

Jan01 - Dec03 -1.00 1.65 0.66

R Jan99 - Dec01 6.50 5.47 7.30

Jan00 - Dec02 4.00 5.34 5.47

Jan01 - Dec03 4.37 6.44 5.10

RMSRE Jan99 - Dec01 0.25 0.17 0.35

Jan00 - Dec02 0.13 0.16 0.23

Jan01 - Dec03 0.14 0.18 0.24

Q Jan99 - Dec01 85.71% 91.43% 94.29%

Jan00 - Dec02 91.43% 91.43% 85.71%

Jan01 - Dec03 94.29% 91.43% 85.71%

27ECB


ANNEX 2

Char t 22 Euro a rea e r ror s and omi s s ions

(EUR billions)


Jan. July Jan. July Jan. July Jan. July Jan. July1999 2000 2001 2002 2003

-50-40-30-20-10

0 01020304050

-50-40-30-20-10

1020304050

Tab l e 8 S tab i l i t y i nd i ca tor s f o r e r ror sand omi s s ions

Quality Reference

indicator period

R Jan99 - Dec01 0.64

Jan00 - Dec02 -2.67

Jan01 - Dec03 -4.55

R Jan99 - Dec01 9.77

Jan00 - Dec02 10.05

Jan01 - Dec03 9.68

RMSRE Jan99 - Dec01 0.80

Jan00 - Dec02 0.82

Jan01 - Dec03 0.82

Q Jan99 - Dec01 65.71%

Jan00 - Dec02 68.57%

Jan01 - Dec03 77.14%

Errors and omissions

RESEARCH NETWORK ON CAP ITAL MARKETSAND F INANC IAL INTEGRAT ION IN EUROPE

DECEMBER 2004

CENTER FORFINANCIALSTUDIES

RESULTS ANDEXPERIENCE AFTER TWO YEARS

EURO AREA BALANCE OF PAYMENTS AND INTERNATIONAL INVESTMENT POSITION STATISTICS, JANUARY 2005CONTENTSEXECUTIVE SUMMARYINTRODUCTION1 METHODOLOGICAL SOUNDNESS2 PERIODICITY AND TIMELINESS3 REVISION PRACTICE AND POLICY4 STABILITY4.1 THE DIRECTIONAL RELIABILITY SHOWS A RELATIVE WEAKNESS IN DIRECT INVESTMENT4.2 THE MEAN ABSOLUTE PERCENTAGE ERROR SHOWS AN IMPROVEMENT FOR INCOME4.3 THE IMPROVEMENT IN THE CURRENT ACCOUNT IS CONFIRMED BY THE ROOT MEAN SQUARE RELATIVE ERROR4.4 STABILITY OF THE INTERNATIONAL INVESTMENT POSITION

5 CONSISTENCY5.1 INTERNAL CONSISTENCY5.2 EXTERNAL CONSISTENCY

6 CONCLUSION

ANNEX 1 METHODOLOGICAL DOCUMENTATION FOR QUALITY INDICATORS1 RELIABILITY/STABILITY1.1 SIMPLE MEASURES OF REVISIONS1.2 RELATIVE MEASURES OF REVISIONS1.3 DECOMPOSABLE INDICATORS

2 SERVICEABILITY/CONSISTENCY2.1 INTERNAL CONSISTENCY2.2 EXTERNAL CONSISTENCY

ANNEX 2 TABLES AND CHARTS

Euro area balance of payments and international investment ...January 2005 Some aspects of quality...

Documents

Transcript of Euro area balance of payments and international investment ...January 2005 Some aspects of quality...