EUROPEAN COMMISSION DIRECTORATE GENERAL ECONOMIC AND FINANCIAL AFFAIRS Economies of the Member States How Reliable are Statistics for the Stability and Growth Pact? European Economy-Economic Papers 273 Notas Estadísticas (Banco de España) 4 João Nogueira Martins co-author with Luis Gordo (Bank of Spain) DG ECFIN European Commission
Motivation Inconsistency between –the rhythm at which statistics are compiled and revised –the deadlines for political decisions under the SGP Statistics are compiled three months after the end of the year… and revised for four years until they become final Policy decisions are taken in the space of a few months… –Policy decisions are taken on the basis of preliminary data which are subject to large revisions. –Revisions often appear too late to be taken into account Compare with other administrative uses (e.g. GNI,…)
Revisions that revealed policy (or procedure) errors Finland: –Abrogation of excessive deficit procedure in spring 1997 –Deficit data revised from below to above 3% of GDP Portugal –Excessive deficit procedure in 2002 –Deficit data revised from below to above 3% of GDP –Procedure started 6 months later than usual Greece –Several major revisions. The deficit was never below 3% of GDP –Abrogation of excessive deficit procedure should have never taken place Italy –Excessive deficit procedure in 2005 –Deficit was above 3% for some time –Procedure started late Portugal, Spain and France: joined the euro area with deficits above 3% of GDP.
It is unrealistic to slow down the SGP-related procedure “The fundamental objective of the excessive deficit procedure is to identify, as quickly as possible, a situation of excessive deficit or debt and to put an end to it.” ECOFIN Council Conclusions on statistical governance, 8 November 2005 It is illusory to accelerate significantly the production of fiscal statistics. Is there a third way? To be explicitly aware of potential revisions. Countries are not equal in terms of reliability. Reliability increases with time.
Relevance Timeliness Punctuality Accessibility Coherence Completeness Comparability across time and space Accuracy Reliability It refers to the closeness of the initial estimated value to the subsequently revised values Transparency Interpretability Credibility Cost-efficiency Quality of statistics: A multi dimensional concept
Data reported in spring and autumn each year since Each reporting covers years t-1 to t-4 Available series start in 1990 Latest transmission covers years (now also 2006) Available sample for EU-15
Available sample
Each year is reported 8 times. (8 ages) [For 2006, we only have data of age 1] Data of age 2 for 2006 will arrive by end-September For 2005, we have data of age 1, 2 and 3 For 2004: data of ages 1 to 5 For 2003: data of ages 1 to 7 For 1993 to 2002: data of ages 1 to 8 For 1992: data of ages 3 to 8 For 1991: data of ages 5 to 8 For 1990: data of ages 7 and 8 Slightly shorter series for AT, SE, FI EU-10 are not covered for the time being. Greece also partially excluded from analysis Available sample for EU-15
Structural revisions and years with only a few ages reduce the available sample
Numerator (deficit / debt) Denominator (GDP) Our analysis concerns the reliability of the numerator GDP routine revisions rarely have any impact on the deficit ratio A routine revision of 1% in GDP level (a very and large revision) would reduce the deficit ratio from 3.00% to 2.97% Revisions
Routine revisions –New information –Errors or inconsistencies identified –Small steps to better comply with the rules Structural revisions –Change in accounting systems (ESA79 to ESA95) –Large structural errors corrected –Major changes in accounting systems of Member States An analysis of reliability is only concerned about routine revisions Revisions
Calculating average revisions and standard deviations
Routine revisions –New information –Errors or inconsistencies identified –Small steps to better comply with the rules Structural revisions –Change in accounting systems (ESA79 to ESA95) –Large structural errors corrected –Major changes in accounting systems of Member States An analysis of reliability is only concerned about routine revisions Revisions
Impact of structural breaks Are we underestimating reliability?
Deficit and debt Which are more reliable? Two approaches
Did data become less reliable because of ESA95?
Is there a link between deficit ratio is revisions?
Conclusions (1) Policy makers should be made aware that data are subject to revisions One cannot anticipate revisions, but countries are different… …useful to make explicit uncertainty ranges around the reported data Countercheck official deficit data with other indicators
Conclusions (2) FR, DE, UK: most reliable deficit (GR), DK, PT, LU, SE, IT: largest dispersion of revisions Large countries report better data? (but compare IT and IE) (GR), IT, PT: largest deficit-increasing revisions DK, SE: large revisions, offsetting each other (in line with other evidence, e.g. statistical discrepancies)
Conclusions (3) Debt not necessarily more reliable than deficit Debt better than deficit: DE, ES, LU, IT, PT Deficit better than debt: BE, FR, AT, FI Conjecture: Institutional issues (deficit and debt compilation; debt decentralisation, consolidation, financing of public enterprises) explain these differences
Conclusions (4) There is no evidence that shift from ESA79 to ESA95 has led to a deterioration of reliability Complex rules, more experience, better technology,…
Conclusions (5) Size of deficit may have an impact on revisions –Is 3% a focal point? –Biased sample? –Independence of data compilers?
Conclusions (6) Revisions are not an indicator of quality Reliability is only one dimension of quality Revision are inevitable; even desirable Regular analysis of revisions Improvement in processes.
Conclusions (7) It would be very useful to compare EU Member States with other OECD countries Does the SGP have an impact on the behaviour of data compilers?
EUROPEAN COMMISSION DIRECTORATE GENERAL ECONOMIC AND FINANCIAL AFFAIRS How Reliable are Statistics for the Stability and Growth Pact? Comments and questions: WELCOME Thanks for your attention!!