Economic Pape s a e w i en by he s a o he Di ec o a e Gene al o
Economic and Financial A ai s o by expe s wo king in associa ion wi h
hem. The se ies in ends o inc ease awa eness o he analy ical wo k done on
EU economic issues and o seek commen s and sugges ions o u he
analyses. Views exp essed ep esen exclusi ely he posi ions o he au ho and
do no necessa ily co espond o hose o he Eu opean Commission.
Commen s and enqui ies should be add essed o he:
Eu opean Commission
Di ec o a e-Gene al o Economic and Financial A ai s
Publica ions
BU1 - -1/13
B - 1049 B ussels, Belgium
ISBN 978-92-79-04626-1
KC-AI-07-273-EN-C
©Eu opean Communi ies, 2007
HOW RELIABLE ARE THE STATISTICS
FOR THE STABILITY AND GROWTH PACT? †
LUÍS GORDO MORA* AND JOÃO NOGUEIRA MARTINS°
Abs ac : The aim o his pape is o assess he eliabili y o he go e nmen de ici and deb igu es
epo ed o he Eu opean Commission by Membe S a es. Reliabili y is one o he se e al dimensions o
quali y in s a is ics; i e e s o he magni udes o da a e isions a e he publica ion o he i s ou comes.
The measu emen o he da a eliabili y and in e ence abou po en ial u u e e isions a e pa icula ly
ele an o iscal su eillance in he EU since s a is ical ins i u es ake a long ime – usually ou yea s –
o p o ide inal da a, while he decisions on he S abili y and G ow h Pac con ex a e aken on he basis o
he i s es ima es a ailable sho ly a e he end o each yea . The pape shows ha he e a e e y
signi ican di e ences in eliabili y among Membe S a es and indica es he ma gins o unce ain y in
ela ion o he mos ecen yea s’ da a. I also compa es he eliabili y o de ici and deb igu es; checks
ha he shi om ESA79 o ESA95 did no gene ally ha m he eliabili y o da a and sugges s ha he size
o de ici s may ha e an impac on he way s a is ical o ices e ise da a.
JEL classi ica ion codes: C80, H62
Keywo ds: eliabili y, quali y o s a is ics, S abili y and G ow h Pac , go e nmen de ici
† The iews exp essed a e hose o he au ho s and should no be a ibu ed o he Banco de España o o
he Eu opean Commission. The au ho s acknowledge help ul commen s by F. Balassone, M. Bu i,
A. Dé uennes, G. Giudice, F. Lequille , J. K öge , A. Pina, A. Tu ini, S. Zo e i and se e al o he
colleagues in he Banco de España and he Commission, as well as edi o ial assis ance by S. Bland.
The pape was p epa ed du ing an in e nship o L. Go do in he Commission, om Oc obe o
Decembe 2006.
* Banco de España – Di ec o a e-Gene al Economics, S a is ics and Resea ch. S a is ics and Cen al
Balance Shee Da a O ice ([email p o ec ed]).
° Eu opean Commission – Di ec o a e-Gene al o Economic and Financial A ai s
(joao.noguei ama [email p o ec ed]).
2
HOW RELIABLE ARE THE STATISTICS
FOR THE STABILITY AND GROWTH PACT?
“The undamen al objec i e o he excessi e de ici p ocedu e is
o iden i y, as quickly as possible, a si ua ion o excessi e de ici
o deb and o pu an end o i .”
ECOFIN Council Conclusions on s a is ical go e nance,
8 No embe 2005.
1. INTRODUCTION
The undamen al objec i e o he excessi e de ici p ocedu e (EDP) – ha is, he
co ec i e a m o he S abili y and G ow h Pac (SGP) – is o iden i y, and pu o an end
o, si ua ions o excessi e de ici , as quickly as possible.1 This aim is e lec ed in he
apidi y wi h which se e al s eps in he p ocedu e ake place. In e es ingly, he i le o he
applicable egula ion is “on speeding up and cla i ying he implemen a ion o he
excessi e de ici p ocedu e.”2 Al hough he 2005 e o m o he SGP has ex ended he
deadlines o se e al s eps in he p ocedu e, made hose deadlines less s ic and he
whole p ocedu e less igh , he whole p ocedu e s ill akes place a a an ic pace.3
This has e con as s wi h he ime ame o compila ion and publica ion o budge a y
s a is ics. The i s ou comes o a gi en yea a e published a he beginning o he
subsequen yea – bu hese igu es a e no inal; hey include pa ial in o ma ion o
es ima es ha need o be e ised a a la e s age. Da a a e e ised wice a yea o a leas
ou yea s; some imes mo e equen ly and o longe .
The di e ence in he ime ames o policy decisions and o a ailabili y and e ision o
da a means ha he i s ou comes a e hose ha ma e o he p ocedu e and hose on
he basis o which c ucial policy decisions a e aken. The e o e, he s a egic incen i es
linked o nume ical ules ela e i s and o emos o he i s da a ansmissions (see o
example Bu i e al., 2006). The e ised da a a e o be e quali y han he i s ou u ns
bu a e a ely ele an o he p ocedu e, as hey appea oo la e.
1 Fo he economic ounda ions o iscal su eillance in he EU, see B unila, e al. (eds.) (2001) and
Bu i and F anco (2005) and e e ences he ein.
2 Council Regula ion (EC) N° 1467/97 (OJ L 209, 2.8.1997, p.6). Regula ion as amended by Regula ion
(EC) N° 1056/2005 (OJ L 174, 7.7.2005, p. 5).
3 Da a a e epo ed by Membe S a es in sp ing and au umn each yea . Eu os a has h ee weeks o
sc u inise, alida e and publish he da a. I he e e ence alues a e exceeded, he Commission
p epa es a epo . Wi hin wo weeks o he adop ion o he epo , he Economic and Financial
Commi ee o mula es an opinion; wi hin ou mon hs o he epo ing da es, he Council decides
whe he he de ici is excessi e and adop s ecommenda ions; he coun y conce ned has six mon hs
o e ec i e ac ion; immedia ely a e his pe iod, he Council checks whe he e ec i e ac ion has
been aken; i no , i has wo mon hs o adop a no ice. A decision o impose inancial sanc ions can be
aken ou mon hs a e he no ice and 16 mon hs a e da a ha e been epo ed. Fo mo e on he s eps
o he p ocedu e and on he SGP e o m, see Eu opean Commission (2005, in pa icula sec ion II.1
he eo ).
3
In a numbe o ins ances, decisions would ha e been di e en i inal da a had been
a ailable ea lie .4 The la ge e isions in he da a o a ew coun ies ha e ha med no only
hose coun ies’ epu a ion, bu also he c edibili y o he whole p ocedu e, and ha e
aised issues o equi y among Membe S a es.
A e he e ways o imp o ing his si ua ion? Clea ly, he solu ion is no o s op he SGP-
ela ed p ocedu es and wai un il inal da a a e a ailable. In many cases, when inal
igu es a e published, he budge a y di icul ies ha mo i a ed a p ocedu e ha e al eady
been sol ed o , in a less benign con igu a ion – which would be likely in case i no
policy ac ion has been aken, he si ua ion has kep de e io a ing, and he co ec ion will
be mo e pain ul.
O e he las yea s, he e ha e been some measu es a he EU le el and in speci ic
Membe S a es in ended o upg ade he quali y o he budge a y s a is ics.5 While he
e y la ge e isions ha ook place in a ew coun ies in he pas we e excep ional and
will hope ully no be epea ed, i mus be acknowledged ha da a e isions a e
ine i able. Al hough echnological p og ess in in o ma ion echnology speeds up da a
collec ion, he complexi y o he accoun ing ules and he numbe o ansac ions o
eco d (millions o ansac ions by housands o go e nmen uni s in each coun y) mean
ha he p oduc ion o inal s a is ics emains a long and pains aking ask.
One possible way o imp o ing hings would be o explici ly acknowledge, in each s ep
o he EDP, ha he budge a y da a a e subjec o e isions, and ha he eliabili y o da a
inc eases wi h ime and di e s om coun y o coun y. Though we a e unable o say
how he i s ou come published by a gi en coun y will be e ised, expe ience can help
us in e he unce ain y su ounding he de ici and deb igu es epo ed by Membe
S a es. This may help in iden i ying si ua ions o isk o excessi e de ici . In some cases,
i may make sense o pos pone he ab oga ion o a p ocedu e when he de ici a io has
allen below, bu emains close o, 3% o GDP, and he e is a signi ican isk ha he
de ici igu es will ha e o be e ised o abo e he e e ence alue a a la e s age, in
pa icula i some o he indica o s sugges ha an upwa d e ision is likely.6 A leas he
policy-make s should be made awa e o , and he ele an ac s should explici ly e e o,
he ac ha he da a on which c ucial decisions a e aken a e subjec o e ision.7
4 A ew examples: he excessi e de ici p ocedu e o Finland was ab oga ed in June 1997 on he basis
o a de ici a io o 1996, epo ed on 1 Ma ch 1997, o 2.6% o GDP; ha a io was hen
successi ely e ised o 3.1%, 3.3% and e en 3.5% o GDP. In Ma ch 2002, Po ugal epo ed a de ici
a io o 2.2% o GDP o 2001; a e six mon hs he de ici a io was e ised o 4.1% o GDP; as a
esul , he ensuing p ocedu e s a ed hal a yea la e han should ha e been he case; in Ma ch 2004,
wo yea s a e he i s ansmission, he de ici a io was u he e ised upwa ds o 4.4% o GDP. In
G eece, he e we e se e al e y la ge e isions in he de ici and deb a ios; he la ges e ision
conce ned 2001: G eece epo ed a small go e nmen su plus o 0.1% o GDP in Ma ch 2002 which
was hen successi ely e ised o de ici s o 1.4%, 1.5%, 3.7% and 6.1% o GDP. In sp ing 2005, he
Commission ini ia ed a p ocedu e agains I aly, as i appea ed ha he 2004 de ici was a 3.0%; in ac
he 2004 de ici a io was e ised o 3.4% and i appea ed ha he 2001 and 2003 a ios we e also
abo e 3%.
5 See Eu opean Commission (2005, in pa icula box II.2; and 2006, in pa icula sec ion II.4.4 he eo ).
6 On he usage o mul iple indica o s o coun e check he o icial de ici and deb igu es, see Balassone
e al. (2006a).
7 Ano he way o accele a e he compila ion o s a is ics could be o simpli y he accoun ing ules.
Howe e , his op ion does no seem o be on he ca ds and his pape does no dwell on his speci ic
issue. No e, none heless, ha Tanzi (2006) a gues ha “ he concep ha he s a is icians ha e chosen
( he acc ual concep ), hough supe io o [a simple ] cash concep in some ways, su e s om he
sho coming ha i is no easy o calcula e and does no p o ide policy-make s wi h a gauge o he

4
The aim o his pape is o assess he eliabili y o he go e nmen de ici and deb
igu es epo ed o he Eu opean Commission by Membe S a es.8 As a as we know, no
o he pape has a emp ed o do so in a sys ema ic manne . The pape is o ganised as
ollows. In he nex sec ion, we discuss some concep ual issues abou quali y in s a is ics
and be e delimi he scope o he analysis; in pa icula , we iden i y eliabili y as one o
he se e al dimensions o quali y in s a is ics. Sec ion 3 de ails he a ailable da a and he
p ocedu e o calcula ing he ele an indica o s. Sec ions 4 and 5 compa e he eliabili y
o igu es epo ed by he di e en EU Membe S a es. Sec ion 6 compa es he e isions
in de ici and deb igu es. Sec ion 7 elabo a es on he e olu ion o eliabili y o e ime,
while sec ion 8 p o ides a p elimina y examina ion on he ela ionship be ween size o
e isions and size o de ici s. Sec ion 9 summa ises he main indings and concludes.
2. QUALITY OF STATISTICS: A MULTIDIMENSIONAL CONCEPT
The quali y o economic s a is ics is c ucial o an adequa e unde s anding o he
economic si ua ion and o e ec i e policymaking. Poo -quali y s a is ics may lead o
poo economic analysis, mis aken conclusions abou he beha iou o economic agen s
and inapp op ia e policy decisions. The quali y o he budge a y s a is ics o Membe
S a es is pa icula ly impo an gi en ha hese s a is ics a e he ounda ion o a
budge a y su eillance amewo k in a mul i-coun y con ex (Eu opean Commission,
2003, in pa icula sec ion II.4 he eo ).
The quali y o s a is ics can be gene ically de ined as i ness o use (OECD, 2003). This
encompasses all aspec s o how well s a is ics mee use s’ needs and hei expec a ions
abou he in o ma ion con en o he dissemina ed da a (Damia and Picón Aguila (2006),
quo ing he Encyclopaedia o S a is ical Sciences). The quali y o s a is ics is a
mul idimensional concep . The da a quali y assessmen amewo k (DQAF) o
go e nmen inance s a is ics o he IMF (2003) iden i ies he ollowing dimensions o
quali y: (i) in eg i y, (ii) me hodological soundness, (iii) accu acy and eliabili y, (i )
se iceabili y and ( ) accessibili y, as well as a numbe o ( i) p e equisi es o quali y.
These dimensions o quali y can be b oken down in o a numbe o quali y elemen s,
no ably ele ance, anspa ency, imeliness, punc uali y, consis ency, comple eness,
compa abili y ac oss ime and space, in e p e abili y, c edibili y, plausibili y and cos -
e iciency. The UN, OECD and Eu os a iden i y sligh ly di e en dimensions o
s a is ical quali y, o de ine hem in sligh ly di e en ways, bu di e ences a e mo e
seman ic han subs ance.9
S a is ics a e accu a e i hey su icien ly po ay eali y. Accu acy is hus ela ed o he
disc epancy be ween s a is ics and he unknown “ ue” igu es hey a e in ended o
measu e. This disc epancy can be sepa a ed in o wo componen s. The i s is he
di e ence be ween inal igu es and eali y. The measu emen o his disc epancy is e y
iscal si ua ion a he ime when i is mos needed – du ing and no a e , o much a e , he iscal
yea ”. Balassone e al. (2003) ecall ha he o e all s a is ical amewo k and i s complexi y was no
designed o suppo he en o cemen o ules and e en sugges ede ining he e e ence de ici as he
change in nominal deb , wi h bene i s in imeliness and anspa ency. Bu i (2006) also pleads o
simple ules as a way o inc easing he eliabili y o da a.
8 We a e no awa e o any simila pape o he US. The compa ison be ween e isions in he US
go e nmen inance s a is ics and he EU Membe S a es is a opic o u he esea ch.
9 See e.g. De V ies (2002), Eu os a (2003), IMF (2003), OECD (2003), Saeboe (2003), Gio annini
(2004). Fo a compa ison be ween he IMF’s da a quali y assessmen amewo k and Eu os a ’s
quali y de ini ion, see Lalibe é e al. (2004) and Lalibe é and De ays (2006).
5
complex, since he “ ue” igu es a e no obse able. The second is eliabili y. This is he
di e ence be ween he i s published da a and he inal igu es. (Bie and Ahne , 2001,
and Ca son and Lalibe é, 2002). The e o e, eliabili y (o s abili y) o da a e e s o he
closeness o he ini ial es ima ed alues – which a e hose which a e mo e ele an o
policymaking – o he subsequen ly e ised alues.
The dimension o s a is ical quali y ha is discussed in he pape is eliabili y. Howe e ,
se e al dimensions o s a is ical quali y, o he han eliabili y, a e o he u mos ele ance
o iscal su eillance. The e a e in e es ing pa allels be ween he dimensions o
s a is ical quali y and he ideal quali ies o iscal ules as p esen ed in he academic
li e a u e: namely ha hey should be well-de ined, anspa en , simple, lexible,
adequa e o inal goal, en o ceable, consis en , and linked wi h s uc u al e o m (Kopi s
and Symansky, 1998). Fiscal ules will no be en o ceable i iscal s a is ics a e no
imely and eliable; he anspa ency o he iscal nume ical ules is di ec ly connec ed
wi h he anspa ency o he a ailable da a; simple ules (e.g. hose ha e e o a single
nume ical indica o wi h a small o no ma gin o judgemen ) will be adequa e o he
inal goal only i he nume ical indica o is e y accu a e and ele an ; mo e lexible
iscal ules which allow a wide ma gin o judgemen will be less dependen on he
accu acy o he nume ical indica o s (Bu i, 2006).
Mo eo e , he dimensions o s a is ical quali y a e no independen om each o he .
The e a e ade-o s be ween se e al dimensions, o example be ween imeliness and
eliabili y (Bie and Ahne , 2001); in eg i y, eliabili y and accu acy; comple eness and
compa abili y; me hodological soundness and compa abili y ac oss ime; e c.
The e o e, any conclusion on he eliabili y o da a o any coun y canno be unde s ood
as an o e all assessmen o he quali y o hei da a, no on he i ness o hese da a o
use in iscal su eillance. The eliabili y o go e nmen da a is a necessa y bu by no
means su icien condi ion o he i ness o da a o use in iscal su eillance.
3. DATA AVAILABLE AND RELIABILITY INDICATORS
The EU Membe S a es compile and publish go e nmen de ici and deb igu es a hei
own pace. Fo example, in Ge many, he de ici and deb igu es a e usually published
e y sho ly a e he end o he espec i e yea . De ici da a o 2005 we e published by
he Ge man S a is ical O ice on 12 Janua y 2006; a i s e ision, oge he wi h de ailed
e enue and expendi u e accoun s, was eleased on 22 Feb ua y 2006.10 O he EU
coun ies publish hei da a somewha la e .
Howe e , all EU Membe S a es a e unde an obliga ion o epo hei de ici and deb
igu es o he Commission a p ecise da es in sp ing and au umn each yea . Such igu es
– oge he wi h he Commission se ices’ o ecas s, which a e published sho ly a e he
iscal epo ing – cons i u e he aw ma e ial on he basis o which iscal su eillance
p ocedu es a e pu in mo ion. Each da a ansmission co e s he p e ious ou yea s.
A e ou yea s, o eigh no i ica ions, i is assumed ha da a a e inal; hey hen d op
om he da a ansmission. In o he wo ds, da a o yea a e epo ed o he i s ime
in sp ing +1 and e ised un il au umn +4: 42 mon hs a e he ansmission o he i s
ou come, o 45 mon hs a e he end o he yea o which hey e e . These
ex ao dina ily long lags o he publica ion o inal da a illus a e jus how demanding
10 See Des a is P ess Release o 22 Feb ua y 2006.
6
he ask o compiling go e nmen de ici and deb igu es is. Some coun ies keep
e ising hei da a a e ou yea s; howe e , o he pu pose o his pape , we will
assume ha da a o all coun ies a e inal a e ou yea s
The a ailable sample consis s o da a epo ed o he Commission by Membe S a es
be ween Ma ch 1994 and Oc obe 2006.11 Each da a ansmission includes da a wi h
di e en e ision his o ies. In sp ing , da a o -1 a e ansmi ed o he i s ime, da a
o -2 a e ansmi ed o he hi d ime, da a o -3 a e ansmi ed o he i h ime and
da a o -4 a e ansmi ed o he se en h ime. The second, ou h, six h and eigh
ansmissions ake place in au umn. To simpli y, we will say ha each da a ansmission
(also called a da a in age) con ains da a o di e en ages: ages 1, 3, 5 and 7 in sp ing,
and 2, 4, 6 and 8 in au umn.
Fo each coun y, he e a e 104 obse a ions, ha is, 13 obse a ions o each age.12 The
analysis conce ns 14 coun ies: Belgium, Denma k, Ge many, Spain, F ance, I eland,
I aly, Luxembou g, Ne he lands, Aus ia, Po ugal, Finland, Sweden and he Uni ed
Kingdom. We ha e excluded G eece om mos o he analysis, gi en ha he e y la ge
e isions o a s uc u al na u e, which ook place be ween au umn 2002 and au umn
2006, canno be used o in e any hing abou po en ial u u e e isions. Because o da a
a ailabili y issues, he assessmen o he eliabili y o epo ed de ici igu es does no
conce n he en new Membe S a es, which ha e epo ed hei da a unde he same
condi ions as o he Membe S a es since sp ing 2004 only.
Da a o age 2 a e expec ed o be mo e eliable – i.e. close o inal da a – han da a o age
1, da a o age 3 a e expec ed o be mo e eliable han da a o age 2, and so on. As a
esul , o each o he 14 coun ies conside ed, he e a e 8 di e en samples o 13
obse a ions each. These a e sho samples o s a is ical in e ence pu poses; as a esul ,
eliabili y indica o s need o be in e p e ed ca e ully and p udence is necessa y in he
conclusions.
The mos ele an s a is ics o he implemen a ion o he S abili y and G ow h Pac a e
he a ios- o-GDP o he go e nmen de ici and deb . Re isions in he de ici and deb
a ios can occu because o e isions in he nume a o s and denomina o s, ha is,
because o e isions in he de ici and deb igu es, o e isions in he GDP le els. This
pape is conce ned abou he e isions in he de ici and deb le els only. We dis ega d
he ac ha he nominal GDP igu es a e also subjec o e isions.13 Howe e , since
he e is a need o s anda dise da a, o make hem compa able ac oss ime and ac oss
coun ies, ou calcula ions a e as a pe cen age o he la es a ailable GDP se ies.
Table 1 summa ises he a ailable sample o de ici e isions. The numbe o e isions
alling in each o i e b acke s (la ge and medium-size e isions ha educe he de ici ,
small e isions, and la ge and medium-size e isions ha inc ease he de ici ) is
indica ed. In ela ion o he cen al b acke co esponding o small e isions, he able
also epo s on he de ici -inc easing e isions and on e y small e isions. As expec ed,
he numbe o small e isions ( e isions in he de ici a ios be ween -0.2 and 0.2% o
GDP), and e y small e isions ( e isions be ween -0.05% and 0.05% o GDP) ends o
11 De ici and deb da a epo ed by Membe S a es since 1994 a e a ailable om he au ho s upon
eques .
12 The numbe o obse a ions o Aus ia, Sweden and Finland, which ha e epo ed hei da a o he
Commission since hei accession o he EU in 1995, is 96.
13 See Annex I. On GDP e isions, see also A uoba (2005) and Di Fonzo (2005).
7
inc ease wi h he da a ages, while he numbe o e y la ge e isions ( he i s and i h
column co esponding o e isions in he de ici a ios below -0.5% and abo e 0.5% o
GDP) alls. Mo eo e , he able sugges s no mali y in he e isions (see Annex I.3).
Table 1: Sample a ailable – Re isions in go e nmen de ici be ween wo consecu i e da a
ansmissions
x<-0.5% -0.5%<x<-0.2% -0.2%<x<0.2% 0.2%<x<0.5% 0.5%<x TOTAL
To al O which: O which:
0.0%<x<0.2% -0.05%<x<0.05%
1s 7 16 123 62 58 25 11 182
2nd 4 20 125 55 59 15 4 168
3 d 6 13 147 55 85 13 3 182
4 h 6 17 131 53 82 11 3 168
5 h 2 7 160 51 110 11 2 182
6 h 2 16 144 54 104 2 4 168
7 h 251674812962182
TOTAL 29 94 997 378 627 83 29 1232
Size o e isions in % o GDP
No e: The able summa ises he e isions in go e nmen de ici da a in he sample. Re isions a e classi ied
acco ding o he numbe o e isions (lines) and hei size (columns). A i s e ision o a gi en yea
co esponds o he di e ence be ween he i s and he second da a ansmissions, i.e., da a o yea
epo ed in sp ing +1 and au umn +1. Second e isions co espond o he di e ences be ween he second
and he hi d da a ansmissions, i.e. da a o yea epo ed in au umn +1 and sp ing +2, and so on.
Posi i e e isions (+) mean inc eases in de ici s o educ ions in su pluses; nega i e e isions (–) mean
educ ions in de ici s o inc eases in su pluses. As expec ed, he numbe o small e isions (be ween -0.2
and 0.2% o GDP) and e y small e isions (be ween -0.05% and 0.05% o GDP) inc eases wi h he da a
age, while he numbe o e y la ge e isions ( e isions in he de ici a ios below -0.5% and abo e 0.5%
o GDP) alls.
No all da a e isions ha e he same na u e. Da a a e ypically e ised because new
in o ma ion on go e nmen ansac ions becomes a ailable, e o s o inconsis encies a e
iden i ied and co ec ed, o as a esul o he con inuous e o s o be e comply wi h he
in icacies o he accoun ing ules: hese a e ou ine e isions. We may conjec u e ha
he size o he ou ine e isions di e s among coun ies because o he size o he
coun y, he ins i u ional a angemen s in place, including he deg ee o decen alisa ion
o go e nmen , he echnical a angemen s pu in place o collec da a om each
go e nmen uni , he habi s and obliga ions in he p oduc ion o sou ce da a by each o
he go e nmen uni s, he simila i y be ween he accoun ing s anda ds in use in each o
go e nmen uni and he na ional accoun ing ules, he a ailable esou ces and
echnologies o compile he i s ou come and o e ise da a, as well as he complexi y o
he ansac ions ca ied ou by each coun y, e c. The e a e also s uc u al e isions o
da a se ies, when he accoun ing ules a e signi ican ly e ised ( o example he shi o
ESA79 o ESA95 a he beginning o 2000, o cla i ica ions o accoun ing ules
published by Eu os a 14), majo sys emic changes in he da a compila ion, o when
excep ionally la ge e o s in he accoun s a e iden i ied and co ec ed.15
14 Since 1996, Eu os a has egula ly published cla i ica ions on he implemen a ion o ESA. These
cla i ica ions a e compiled in a “Manual on go e nmen de ici and deb ”. The manual, oge he wi h
upda es, is a ailable a : h p://epp.eu os a .ec.eu opa.eu/po al/page?_pageid=2373,47631312,2373_
58674363&_dad=po al&_schema=PORTAL.
15 A uoba (2005) uses he e ms ‘in o ma i e’ and ‘unin o ma i e’ e isions o e e o wha we call
‘ ou ine’ and ‘s uc u al’ e isions.
14
I aly: o al e ision
-7%
-6%
-5%
-4%
-3%
-2%
-1%
0%
2001 2002 2003 2004 2005
% o GD
P
I aly: e ision o he nex ansmission
-7%
-6%
-5%
-4%
-3%
-2%
-1%
0%
2001 2002 2003 2004 2005
% o GD
P
Ne he lands: o al e ision
-4%
-3%
-2%
-1%
0%
1%
2%
3%
2001 2002 2003 2004 2005
% o GD
P
Ne he lands: e ision o he nex ansmission
-4%
-3%
-2%
-1%
0%
1%
2%
3%
2001 2002 2003 2004 2005
% o GD
P
Aus ia: o al e ision
-5%
-4%
-3%
-2%
-1%
0%
1%
2%
2001 2002 2003 2004 2005
% o GD
P
Aus ia: e ision o he nex ansmission
-5%
-4%
-3%
-2%
-1%
0%
1%
2%
2001 2002 2003 2004 2005
% o GD
P
Luxebou g: o al e ision
-3%
-2%
-1%
0%
1%
2%
3%
4%
5%
6%
7%
2001 2002 2003 2004 2005
% o GD
P
Luxembou g: e ision o he nex ansmission
-2%
-1%
0%
1%
2%
3%
4%
5%
6%
7%
2001 2002 2003 2004 2005
% o GD
P

15
Po ugal: o al e ision
-8%
-7%
-6%
-5%
-4%
-3%
-2%
-1%
2001 2002 2003 2004 2005
% o GD
P
Po ugal: e ision o he nex ansmission
-7%
-6%
-5%
-4%
-3%
-2%
-1%
0%
2001 2002 2003 2004 2005
% o GD
P
Sweden: o al e ision
-2%
-1%
0%
1%
2%
3%
4%
5%
2001 2002 2003 2004 2005
% o GD
P
Sweden: e ision o he nex ansmission
-2%
-1%
0%
1%
2%
3%
4%
5%
2001 2002 2003 2004 2005
% o GD
P
Finland: o al e ision
0%
1%
2%
3%
4%
5%
6%
7%
2001 2002 2003 2004 2005
% o GD
P
Finland: e ision o he nex ansmission
0%
1%
2%
3%
4%
5%
6%
7%
2001 2002 2003 2004 2005
% o GD
P
16
No e: This se ies o cha s show he la es a ailable de ici se ies ( he hick line) and con idence in e als
su ounding hose igu es. The do ed and he dashed lines show how he de ici igu es a e suscep ible o
e ision un il hey become inal (le -hand side cha s). The e is a p obabili y o 60% ha da a will be
comp ised be ween lines B and C, and o 90% be ween lines A and D (see he i s cha o Belgium). In
e ms o unila e al con idence le els, he e is a p obabili y o 80% ha he de ici /su plus will no go below
line C and o 95% ha i will no go below line D. Fo he nex da a ansmission ( igh -had side cha s),
he bila e al con idence le el is 90%; he unila e al con idence le el is 95%.
6. COMPARING DEFICIT AND DEBT REVISIONS
Deb igu es a e o en assumed o be mo e eliable han de ici igu es.25 This is because
accoun ing ules o deb a e much simple han hose o de ici s (deb is a cash-based
concep which is di ec ly obse able and o which da a a e easie o compile han
de ici s which a e calcula ed on an acc ual basis), and because he de ici is a balancing
low ha is in luenced by all changes in o he i ems o go e nmen accoun s, while deb
is a s ock, which by na u e is mo e ine han a low. The idea ha he de ici igu es a e
less eliable also comes om he ac ha he de ici has been a mo e poli ically sensi i e
a iable.26 The e o e, he go e nmen s ha e mo e in e es in unde es ima ing he de ici
igu es in he i s da a ansmission, e en i he da a will be e ised a a la e s age. In
his sec ion, we es whe he his iew is well ounded.
A di ec compa ison be ween he eliabili y indica o s o he de ici and deb igu es as
hose shown in sec ions 4 and 5 would be meaningless, because o di e ences in he size
and na u e o bo h concep s.27 A meaning ul compa ison o he eliabili y o de ici and
deb mus ake in o accoun hese di e ences. We sugges compa ing he eliabili y o
de ici and deb igu es h ough wo complemen a y app oaches.
25 Al hough he go e nmen de ici /su plus is he main d i e o he go e nmen deb , he yea ly change
in he go e nmen deb does no exac ly co espond o he go e nmen de ici . The di e ence is
known as s ock- low adjus men (o SFA). Fo u he de ails on he componen s o he s ock- low
adjus men , and a heo e ical model on he au ho i ies’ beha iou in ela ion o de ici s, deb and SFA
in he p esence o iscal ules, see Bu i e al (2006).
26 Balassone e al. (2003, 2006a and b), Bu i e al. (2006) and Koen and an den Noo d (2006).
27 Annex III p o ides a se ies o indica o s on he e isions o deb da a, simila o hose shown abo e
o he de ici .
Uni ed Kingdom: o al e ision
-5%
-4%
-3%
-2%
-1%
0%
1%
2%
2001 2002 2003 2004 2005
% o GD
P
Uni ed Kingdom: e ision o he nex ansmission
-5%
-4%
-3%
-2%
-1%
0%
1%
2%
2001 2002 2003 2004 2005
% o GD
P
17
The i s app oach is o calcula e a coe icien o eliabili y s anda dised by ola ili y
(CRSV). We ha e de ined he CRSV as he s anda d de ia ion o he e isions o e e y
da a age di ided by he ola ili y o he unde lying se ies. The ola ili y o each se ies
(de ici and deb ) is measu ed by he s anda d de ia ion o he yea ly changes. The
CRSV is a noise- o-signal a io. The a ionale o he CRSV is ha ola ile se ies a e
mo e di icul o compile in a imely way and he e o e equi e la ge e isions. A highe
alue in CRSV means less ela i e eliabili y.
Cha s in Figu e 3 show he e olu ion o CRSV by age and coun y o de ici and deb
igu es. Since da a o age 8 a e de ined as inal, he CRSV o age 8 o all se ies is ze o,
by cons uc ion. Al hough he CRSV o de ici and deb can be compa ed o each age,
he mos use ul compa ison is o age 1, ha is, o he i s da a ansmissions. Deb
igu es ha e been mo e ela i ely eliable han de ici igu es o Denma k, Ge many,
Spain, I eland, I aly, Ne he lands, Finland, Sweden and he Uni ed Kingdom. Howe e ,
he e e se is ue o Belgium, F ance, Luxembou g, Aus ia and Po ugal.
Figu e 4: Coe icien s o eliabili y o de ici and deb s anda dised by ola ili y
Belgium
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
12345678
Ages
CRSV de ici CRSV deb
Denma k
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
12345678
Ages
CRSV de ici CRSV deb
Ge many
0.00
0.05
0.10
0.15
0.20
0.25
0.30
12345678
Ages
CRSV de ici CRSV deb
Spain
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
12345678
Ages
CRSV de ici CRSV deb
18
F ance
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
12345678
Ages
CRSV de ici CRSV deb
I eland
0.00
0.05
0.10
0.15
0.20
0.25
0.30
12345678
Ages
CRSV de ici CRSV deb
I aly
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
12345678
Ages
CRSV de ici CRSV deb
Luxembou g
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
12345678
Ages
CRSV de ici CRSV deb
Ne he lands
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
12345678
Ages
CRSV de ici CRSV deb
Aus ia
0.00
0.20
0.40
0.60
0.80
1.00
1.20
12345678
Ages
CRSV de ici CRSV deb
Po ugal
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
12345678
Ages
CRSV de ici CRSV deb
Finland
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
0.18
12345678
Ages
CRSV de ici CRSV deb
19
No e: This se ies o cha s compa e he coe icien s o eliabili y s anda dised by ola ili y (CRSV) o
de ici and deb a ios o di e en ages. The CRSV is calcula ed as he s anda d de ia ions o e isions
di ided by he s anda d de ia ion o yea -on-yea changes o he unde lying se ies. A smalle coe icien
o de ici (deb ) indica es ha de ici (deb ) igu es a e ela i ely mo e eliable.
A second app oach o compa e he eliabili y o de ici and deb igu es is o compa e he
e isions in de ici igu es and he e isions in he yea ly changes in deb , whe e he
compa ison is be ween wo lows o simila sizes. This app oach is he mo e ele an i
one is in e es ed in he deb de elopmen s o coun e check he plausibili y o de ici
igu es. The cha s in Figu e 4 show he e olu ion o he s anda d de ia ions in he
e isions o de ici and yea ly change in deb igu es. Since assessing changes in deb
implies a compa ison o wo yea s’ da a epo ed a he same ime, he analysis only
conce ns he i s six da a ansmissions.
These cha s ha e been complemen ed by F es s o es ablish whe he he di e ences
be ween he s anda d de ia ions o he de ici and change-in-deb e isions a e
s a is ically signi ican . The conclusions de i ed om he cha s and he F es s a e ha ,
as a as da a o age 1 a e conce ned, he igu es on he yea ly changes in deb a e
e ec i ely mo e eliable han he de ici o Ge many, Spain, Luxembou g, I aly and
Po ugal. The e o e, o hese coun ies, he yea ly change in deb may be used o
coun e check he plausibili y o de ici igu es, and e en o an icipa e u u e de ici
e isions. Howe e , he esul s o Belgium, F ance, Aus ia and Finland go agains he
ecei ed wisdom, wi h hei de ici igu es mo e eliable han he deb igu es. Fo ,
Denma k, I eland, Ne he lands and Sweden, he di e ences in he s anda d de ia ions
a e no s a is ically signi ican .
We conjec u e ha di e ences in he ela i e eliabili y o de ici and deb da a come
om he ins i u ional a angemen s o he compila ion o de ici and deb igu es, which
in se e al coun ies a e no compiled by he same ins i u ion, deb decen alisa ion and
he size o go e nmen liabili ies held by o he go e nmen uni s ( ha is, liabili ies which
a e consolida ed when calcula ing he o e all go e nmen deb ).
Sweden
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
12345678
Ages
CRSV de ici CRSV deb
Uni ed Kingdom
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
0.18
12345678
Ages
CRSV de ici CRSV deb

20
Figu e 4: Compa ing he eliabili y o de ici and change in deb da a: e olu ion o he s anda d
de ia ions o e isions
Belgium
0.00%
0.10%
0.20%
0.30%
0.40%
0.50%
0.60%
0.70%
0.80%
123456
Change in deb De ici
Denma k
0.00%
0.10%
0.20%
0.30%
0.40%
0.50%
0.60%
0.70%
0.80%
123456
Change in deb De ici
Ge many
0.00%
0.05%
0.10%
0.15%
0.20%
0.25%
0.30%
123456
Change in deb De ici
Spain
0.00%
0.05%
0.10%
0.15%
0.20%
0.25%
0.30%
0.35%
0.40%
0.45%
123456
Change in deb De ici
F ance
0.00%
0.10%
0.20%
0.30%
0.40%
0.50%
0.60%
123456
Change in deb De ici
I eland
0.00%
0.05%
0.10%
0.15%
0.20%
0.25%
0.30%
0.35%
0.40%
123456
Change in deb De ici
I aly
0.00%
0.10%
0.20%
0.30%
0.40%
0.50%
0.60%
123456
Change in deb De ici
Luxembou g
0.00%
0.10%
0.20%
0.30%
0.40%
0.50%
0.60%
0.70%
0.80%
0.90%
1.00%
123456
Change in deb De ici
21
No e: This se ies o cha s compa es he e olu ion o he s anda d de ia ions o de ici s and yea ly changes
in deb . When he yea ly change in deb is mo e eliable (i.e. smalle s anda d de ia ion) han he de ici ,
deb igu es can be used o assess he plausibili y o , and an icipa e e isions in, de ici igu es.
7. EVOLUTION OF RELIABILITY OVER TIME
The p ocess o compiling go e nmen accoun s has changed qui e conside ably o e he
las 15 yea s. Technological ad ances in he au oma ed collec ion and p ocessing o da a,
he de elopmen o qua e ly accoun s and an inc ease in he awa eness o he ele ance
o eliable go e nmen accoun s may ha e led o an inc ease in he eliabili y o da a.
The inc easing ele ance o go e nmen accoun s has also led o an upg ade in he
Ne he lands
0.00%
0.10%
0.20%
0.30%
0.40%
0.50%
0.60%
123456
Change in deb De ici
Aus ia
0.00%
0.10%
0.20%
0.30%
0.40%
0.50%
0.60%
0.70%
0.80%
0.90%
1.00%
123456
Change in deb De ici
Po ugal
0.00%
0.10%
0.20%
0.30%
0.40%
0.50%
0.60%
0.70%
0.80%
123456
Change in deb De ici
Finland
0.00%
0.10%
0.20%
0.30%
0.40%
0.50%
0.60%
0.70%
123456
Change in deb De ici
Uni ed Kingdom
0.00%
0.05%
0.10%
0.15%
0.20%
0.25%
0.30%
0.35%
123456
Change in deb De ici
Sweden
0.00%
0.10%
0.20%
0.30%
0.40%
0.50%
0.60%
0.70%
0.80%
0.90%
123456
Change in deb De ici
22
quan i y and quali y o esou ces dedica ed o compiling go e nmen accoun s in
Membe S a es. The e o e, i s ou u ns may be close o inal da a han hey used o be.
A he same ime, he shi om ESA79 o ESA95 may ha e led o an inc ease in he
complexi y o accoun ing ules – o example a wide applica ion o he acc uals
p inciple – which would imply la ge e isions. Mo eo e , he EU go e nmen s ha e
pa icipa ed mo e and mo e in complex ansac ions – such as secu i isa ion o p i a e-
public pa ne ships – ha aise accoun ing di icul ies and o en en ail la ge e isions.
The ac ha Eu os a now sc u inises he accoun s epo ed by Membe S a es in much
mo e de ail han i used o do in he pas may also ha e led o changes in he way da a a e
compiled and e ised.
In o de o each conclusions on he e olu ion o eliabili y o e ime o he de ici da a,
we ha e pooled da a o he ou een coun ies, and spli he sample in o wo pe iods. The
i s pe iod includes da a co esponding o da a ansmissions unde ESA79, ha is om
sp ing 1994 un il au umn 1999; he second pe iod includes da a compiled unde ESA95
since sp ing 2000.
Figu e 5 shows he e olu ion in he s anda d de ia ion in he e isions o he pooled
sample and o he sub-samples ESA79 and ESA95. The s anda d de ia ions o he
cumula ed e isions om he i s ansmission o inal da a a e p ac ically he same
unde ESA79 and ESA95 while he da a ansmi ed om he second o he six h
ansmission appea mo e eliable unde ESA95 han unde ESA79. F es s con i m his
inding. This indica es ha igu es compiled since 2000 unde ESA95 con e ge quicke
o inal han da a compiled p io o 2000. This as e con e gence o inal da a may be
connec ed wi h he ac ha un il 2000, se e al Membe S a es used o e ise he
comple e go e nmen expendi u e and e enue accoun only once a yea . The e o e,
while one canno conclude ha ESA95 da a a e mo e eliable ha ESA79 da a, he
conjec u e ha he shi om ESA79 o ESA95 has ha med he eliabili y o da a is no
con i med by he e idence.28,29
28 To check he obus ness o ou inding, we pe o med he same exe cise o he g oup o coun ies ha
ha e anked lowe in e ms o eliabili y (Ne he lands, I aly, Denma k, Po ugal, Luxembou g and
Sweden; see he anking in able 2). Fo his g oup, he s anda d de ia ion in e isions om age 1 o
inal da a o ESA95 is sligh ly highe han o ESA79, while he e e se is ue o o he ages (2 o 7).
Howe e , he di e ences in he s anda d de ia ions a e no signi ican o any age.
29 We ha e also compa ed he e olu ion o he a e age o e isions in absolu e alue unde ESA79 and
ESA95. The conclusions a e he same as epo ed abo e o he s anda d de ia ion. Ou inding ha
he e has no been a gene al de e io a ion in he eliabili y in da a epo ed p io and a e he shi o
ESA95 is no necessa ily alid o each speci ic coun y. Gi en he small samples a ailable, i is no
possible o check whe he he gene al inding is applicable o each indi idual coun y.
23
Figu e 5: E olu ion o eliabili y (s anda d de ia ion o e isions) o de ici unde ESA79 and
ESA95
0.0%
0.1%
0.2%
0.3%
0.4%
0.5%
0.6%
0.7%
12345678
Ages
% o GDP
To al sample
ESA79
ESA95
No e: This cha compa es he e olu ion o he s anda d de ia ions o he e isions in de ici s be ween each
gi en age and inal da a, when compiled unde ESA79 and unde ESA95. The conjec u e acco ding o
which he shi om ESA79 o ESA95 has ha med he eliabili y o da a is no con i med by he e idence.
Ac ually, da a compiled since 2000 unde ESA95 con e ge as e o inal da a.
8. COMPARING SIZE OF REVISIONS AND SIZE OF DEFICIT
We ha e also in es iga ed whe he he size o he de ici has an impac on he e isions.
F om he da a compile iewpoin , he e is no ob ious eason o he size o he de ici o
ha e an impac on he pa e n o e isions. Tha is, he e is no eason o expec ha he
e o s ha equi e co ec ion and unde lie u u e e isions a e signi ican ly di e en
when he de ici is high o low, when one is in de ici o when one is in su plus.
Howe e , in he con ex o SGP and he ela ed p ocedu es, he e e ence alue o 3% o
GDP o he de ici plays a c ucial ole. S a is ical au ho i ies a e ce ainly no insensi i e
o he poli ical consequences o he e isions.30 A e ision which inc eases o educes a
su plus o ha inc eases o educes a la ge de ici does no ha e he same consequences
as a e ision ha leads he de ici a io o c oss he ba o 3% o GDP.
To check whe he he size o he de ici a io has an impac on he pa e n o e isions,
we ha e pooled all e isions (i.e. o all coun ies and o all in ages) in a o al o 1232
obse a ions. This sample has been di ided in o i e g oups. The i s g oup includes
e isions a ec ing de ici s which p io o he e isions a e abo e 3.5% o GDP, while he
second g oup co e s e isions o de ici s be ween 3.0% and 3.5% o GDP. The hi d
g oup encloses he mos sensi i e e isions in he sense ha hey a ec de ici s be ween
30 Fo a synopsis o he ecen deba e on he ins i u ional independence o da a compile s, see Eu opean
Commission (2006a, in pa icula sec ion II.4.4.4).
30
Annexes
I- Da a managemen
I.1 A ailable sample
The a ailable sample consis s o da a epo ed o he Commission by Membe S a es in
he con ex o he excessi e de ici p ocedu e. Each da a ansmission (also called a
in age) co e s he p e ious ou yea s. The da a ansmissions also include o ecas s o
he cu en yea , which a e no aken in o conside a ion in his pape . The sample
con ains da a o 16 yea s: 1990 o 2005, epo ed be ween Ma ch 1994 and Oc obe
2006. Un il he end o 2005, da a used o be epo ed by 1 Ma ch and 1 Sep embe each
yea . F om he beginning o 2006, he epo ing da es mo ed o 1 Ap il and 1 Oc obe .31
The e o e, de ici and deb igu es o yea a e epo ed o he i s ime in sp ing +1,
and hen successi ely e ised e e y 6 mon hs un il au umn +4.
Table I.1 shows he a ailable sample o he Uni ed Kingdom; o he Membe S a es’ da a
ha e a simila shape. No e ha he numbe o a ailable ages is no he same o each
yea , since each yea has no been subjec o he same numbe o da a ansmissions. Da a
o yea s 1993 o 2002 we e epo ed eigh imes, ha is, om he i s ou come un il
inal da a; o 2003 six imes: he se en h ansmission will ake place in sp ing 2007 and
he eigh h in he au umn 2007; o 2004 he e ha e been ou ansmissions o da e, while
da a o 2005 ha e so a been ansmi ed only wice. Da a o 1990 we e subjec o wo
ansmissions only ( hey d opped ou o he da a ansmission a e being epo ing wice,
in sp ing and au umn 1994), da a o 1991 we e ansmi ed ou imes, and da a o 1992
six imes.
Table I.1: Example o basic da a sample – de ici da a epo ed by he Uni ed Kingdom
De ici (-) igu es 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
Ma -94 -8300 -15900 -37800 -48300
Sep-94 -8165 -15127 -36601 -48664
Ma -95 -15127 -36601 -49365 -46063
Sep-95 -14995 -37575 -49093 -45586
Ma -96 -37575 -49093 -45382 -41742
Sep-96 -37540 -49310 -45380 -40358
Ma -97 -49310 -45380 -39052 -32311
Sep-97 -49660 -45532 -38854 -35971
Ma -98 -45613 -38592 -34805 -14377
Sep-98 -45613 -38592 -34805 -16142
Ma -99 -40032 -32970 -14823 5302
Sep-99 -40032 -32970 -15292 3940
Ma -00 ESA 95 s a s -32909 -16294 2549 10728
Sep-00 -32909 -16294 3810 11835
Ma -01 -16294 3810 11420 41453
Sep-01 -16924 3810 11420 40439
Ma -02 3458 10333 39086 8803
Sep-02 1992 10322 37477 7622
Ma -03 10322 37477 7881 -14058
Sep-03 9092 36136 7154 -15401
Ma -04 36106 6987 -16277 -34945
Sep-04 36351 6716 -18074 -36049
Ma -05 6716 -18074 -36977 -36893
Sep-05 7234 -16658 -36640 -36139
Ma -06 -16817 -36715 -38660 -44727
Sep-06 -17039 -36257 -37641 -39732
As he case o he UK in Table I.1 illus a es, each coun y sample con ains 104
obse a ions. Howe e , he e a e in ac 8 di e en sub-samples o each coun y
co esponding o he 8 ages o da a wi h 13 obse a ions o e e y age.
31 In he pape , we dis ega d his change in he ansmission deadlines and will simply e e o he sp ing
and au umn da a ansmissions.

31
I.2 Le els and a ios
Re isions in he de ici and deb a ios can occu because o e isions in he nume a o s
(de ici /deb ) and denomina o s (GDP). In his pape , howe e , we s udy only he
e ision in he nume a o s. We ha e dis ega ded he ac ha he nominal GDP igu es
a e also subjec o e ision and ha such e isions ha e an impac on he de ici /deb
a ios. This is o wo easons, bo h p agma ic. Fi s , ou ine e isions in he GDP igu es
a ely ha e a signi ican impac on he de ici a ios. This can be illus a ed wi h a simple
nume ical example. Assume a de ici o 3% o GDP; i he nominal GDP o a gi en yea
is e ised upwa ds by 1% – which, o a ou ine e ision, would be a e y la ge, and
ex emely a e, e ision – he de ici a io would all by 0.03% poin s only. O cou se,
he e a e also la ge e isions in GDP o a s uc u al na u e. Fo example in au umn
2005, he GDP le els o he Ne he lands, Spain and Po ugal we e e ised upwa ds by
a ound 4%.32 Howe e , by hei e y na u e, s uc u al e isions a e no ele an o his
analysis, as hey do no help us in e he size o u u e e isions.
A second eason o no conside ing e isions in he denomina o is connec ed wi h he
di e ences in he de ici and GDP compila ion. The de ici is mainly compiled by an
addi i e p ocedu e. This means ha he iden i ica ion and co ec ion o a ou ine e o in
he de ici o a gi en yea does no necessa ily ha e any implica ions o he de ici o
yea +1. The compila ion o GDP is mainly a mul iplica i e p ocess; he iden i ica ion
and co ec ion o a ou ine e o in he GDP le el o a gi en yea usually leads o
simila e isions o he subsequen yea s. As a esul , he eliabili y o GDP le els is
much mo e complex o measu e, and analyses o GDP e isions o en ocus on he
g ow h a es.33
Al hough we a e only in e es ed in he e ision in he nume a o s, he e is a need o
s anda dise he de ici /deb le els and he espec i e e isions ac oss ime and coun ies.
The e o e, all calcula ions a e as a pe cen age o GDP; o simpli y, we use he GDP
se ies a ailable in he Eu opean Commission’s AMECO da abank.
I.3 Syn he ic indica o s o eliabili y
Fo each da a ansmission, he e a e da a o di e en ages and wi h di e en e ision
his o ies. In sp ing , da a o -1 a e ansmi ed o he i s ime, da a o -2 a e
ansmi ed o he hi d ime, da a o -3 a e ansmi ed o he i h ime and da a o
-4 a e ansmi ed o he se en h ime. Each da a ansmission hus con ains da a o
di e en ages: ages 1, 3, 5 and 7 in sp ing, and 2, 4, 6 and 8 in au umn. We aim a
calcula ing eliabili y indica o s o each age, ha is, o es ima e how hese igu es a e
suscep ible o e ision be o e eaching inal s a us in age 8. Da a o age 2 a e expec ed o
be mo e eliable han da e o age 1, da a o age 3 a e expec ed o be mo e eliable han
da a o age 2, e c. Da a o age 8 a e inal.
Fo yea s ha ha e been subjec o eigh ansmissions, he e a e se en e isions be ween
wo consecu i e in ages. These e isions can be ea ed as andom a iables. The e o e,
he e a e se en a iables o he de ici and deb igu es o each coun y o which a
small numbe (10 o 13) o obse a ions a e a ailable. This is ob iously a e y small
32 We should also no e a e ision in he G eek GDP se ies by 26% as announced in Sep embe 2006. A
he ime o w i ing, he new se ies has no ye been alida ed by Eu os a (see Eu os a News Release
N° 139/2006 o 23 Oc obe 2006). In any case, G eece is no sys ema ically conside ed in his pape .
33 On e isions in GDP igu es see, A uoba (2005) and Di Fonzo. (2005).
32
sample o in e ence pu poses; a mo e app op ia e sample which would allow be e
in e ence would need o con ain a ound 30 e isions o each age o da a. Howe e , a
sample o such size will only be a ailable beyond 2020.
Ou aim is o in e he beha iou o hese a iables ep esen ing he e isions in o de o
calcula e syn he ic indica o s o eliabili y. These a iables, co esponding o he
e isions o he di e en ages, can be cumula ed o calcula e o al e isions om any
gi en age o inal da a. I is possible o apply he cen al limi heo em and he law o
la ge numbe s (Lyapuno heo em) assuming ha he gene al condi ions o his heo em
a e complied and ha he a iables a e no mally dis ibu ed. Gi en he small size o he
a ailable sample by coun y, we we e no able o pe o m he usual no mali y es s, such
as he Lillie o s es . The Shapi o-Wilks es , which is app op ia e o samples smalle
han 50 obse a ions, has been applied and i shows ha o 28 a iables o a o al o 196
(98 o de ici and 98 o deb ) he null hypo hesis o no mal dis ibu ion is no ejec ed.
The Lillie o s es shows ha he comple e sample (pooling da a o he 14 coun ies) is
no mally dis ibu ed, a esul which is a consequence o he law o la ge numbe s.
In o de o se a p ocedu e o calcula e a iances, and also as a p e ious condi ion o he
Lyapuno heo em, i is c ucial o es he independence o he a iables. I a iables a e
no mally dis ibu ed and a e independen , he cha ac e is ic unc ion o he sum is he
cha ac e is ic unc ion o a a iable no mally dis ibu ed wi h a mean equal o he sum o
he means and a iance equal o he sum o a iances. The Pea son p oduc momen
co ela ion coe icien shows ha 57% o he o al a iables on a coun y-by-coun y
basis a e independen ; he Spea man ank co ela ion coe icien allows 86% o a iables
o be conside ed independen . The e o e, we ha e accep ed he independence o he
se ies.
Fu he mo e, we ha e ca ied ou equali y es s o mean and a iance on a basis o an
ANOVA model wi h a single ac o . The es shows ha i is no possible o ejec he
equali y o means o he e isions o e he p e ious in ages, bu allows us o clea ly
ejec he null hypo hesis o equali y o a iance ac oss a iables.
To calcula e he a e age e ision and he s anda d de ia ion o hese e isions o each
age, one could simply compa e he se ies o da a o each age wi h inal da a. Howe e ,
because o he small size o he samples, he he e ogenei y o samples ( o se e al yea s,
we do no ha e da a o all ages) and he s uc u al b eaks ha need o be emo ed om
he sample – namely he b eak because o he shi om ESA79 o ESA95 – ano he
echnique mus be used. I we simply compa ed da a o age 1 wi h da a o age 8, we
could only wo k wi h da a o 1993 o 1995 and 1999 o 2002. The da a ansmissions o
1990 o 1992 would be los as he da a o ea lie ages a e no a ailable; da a o 1996 o
1998 would also be los because he e was a s uc u al b eak (shi o ESA79 o ESA95)
in he ansmission o sp ing 2000; and inally da a o 2003 o 2005 would also be los
as he inal ages a e s ill missing.
We ha e ci cum en ing he he e ogenei y o he sample by calcula ing he a e ages and
a iances o e isions in da a be ween each consecu i e ages o da a (da a o age 1 a e
compa ed o da a o age 2; da a o age 2 a e compa ed wi h da a o age 3, and
successi ely). The a e age e ision om each age un il age 8 is calcula ed by adding he
successi e a e age e isions oge he . Assuming ha he a iables a e independen , as
discussed abo e, ha is, assuming ha he e isions linked o an age a e independen o
e isions o o he ages, he s anda d de ia ions a e calcula ed by adding up he unbiased
33
es ima o s o a iances o each a iable, ha is, o each age e ision, and hen calcula ing
he squa e oo .
34
II- Reliabili y indica o s and s uc u al e isions
II.1. Denma k
Denma k used o classi y 2nd-pilla pension schemes in go e nmen . In Sep embe 2004,
pu suan o he Eu os a decision o 2 Ma ch 2004, Denma k eclassi ied hese pension
schemes in he inancial sec o ; his led o a all in he go e nmen su plus o a ound 1%
o GDP and an inc ease in go e nmen deb . In Ma ch 2005, Denma k e u ned o he
p e ious sec o al classi ica ion in line wi h a ansi o y pe iod g an ed by Eu os a on 23
Sep embe 2004. Denma k will once again ha e o change he sec o al classi ica ion o
pensions schemes in sp ing 2007.34
When calcula ing he eliabili y indica o s o Denma k, he de ici and deb e isions o
au umn 2004 and sp ing 2005 o each o he conce ned yea s we e pu oge he and he
o al e ision ( he wo e isions la gely o se each o he ) was e enly dis ibu ed among
he ele an da a ages.
II.2. Po ugal and Sweden
As discussed in sec ion 5, he in e ence on u u e e isions is meaning ul only i he
sample con ains only ou ine e isions and s uc u al e isions ha e been emo ed. This
was no sys ema ically a emp ed in his pape , gi en he small size o he a ailable
sample – he e is he isk o h owing he baby ou wi h he ba hwa e – and he di icul y
o iden i ying s uc u al e isions.
Howe e , o illus a e he impac ha s uc u al e isions may ha e on da a, Figu e II.1
shows an cha s (con idence le el o 90%) o Po ugal and Sweden, wo o he
coun ies mos a ec ed by s uc u al b eaks. The do ed lines a e as in he main ex o
he pape – ha is, he only s uc u al b eak iden i ied and emo ed co esponds o he
shi om ESA79 o ESA95. The long dashed lines de ine he an cha a e emo ing a
numbe o coun y-speci ic s uc u al b eaks.
The s uc u al b eaks emo ed om he sample o Po ugal a e he 5 h e ision o 1999,
he 3 d e ision o 2000 and he 1s e ision o 2001, which ook place a e he so-
called Cons âncio-I epo o summe 2002. Those de ici e isions conce ned mainly he
eco ding o capi al injec ions and axes. Fo Sweden, he e isions emo ed a e he 6 h
e ision o 2000, he 4 h e ision o 2001 and he 2nd e ision o 2002 which
conce ned changes in he eco ding o axes.
Al hough he numbe o e isions emo ed om he Po uguese and Swedish samples is
he same, he impac on he eliabili y indica o s as summa ised in he an cha s is qui e
di e en . Fo Po ugal, he impac is negligible; o Sweden, he an becomes much
na owe . These wo cases illus a e ha quan i a i e indica o s on he eliabili y o da a
mus be in e p e ed ca e ully, in pa icula o coun ies ha ha e unde gone s uc u al
changes in he compila ion o hei da a.
34 See Eu os a News Releases N°30/2004 o 2 Ma ch 2004 and N° 117/2004 o 23 Sep embe 2004.
35
Figu e II.1: E ec o s uc u al b eaks o Po ugal and Sweden on he con idence in e als o he
de ici /su plus
No e: The cha s illus a e how he emo al o s uc u al e isions a ec s he eliabili y indica o s o
Po ugal and Sweden. While o he o me , he impac is negligible; he con idence in e al becomes
much na owe o Sweden. These wo cases illus a e ha p udence is needed when in e p e ing he
eliabili y o da a, in pa icula o coun ies ha ha e unde gone majo changes in he compila ion o hei
da a.
Po ugal: o al e ision
-8%
-7%
-6%
-5%
-4%
-3%
-2%
-1%
2001 2002 2003 2004 2005
% o GD
P
Sweden: o al e ision
-1%
0%
1%
2%
3%
4%
5%
6%
7%
2001 2002 2003 2004 2005
% o GD
P

36
III- Reliabili y indica o s o deb
The analysis o de ici igu es shown in sec ions 4 and 5 can be eplica ed o deb
igu es. Tables III.1 and III.2 show Membe S a es anked acco ding o he s anda d
de ia ions o deb da a e isions and he a e age e ision, while able III.3 shows he
a e age o deb da a e isions in absolu e alue o all coun ies.35
In o ma ion on he eliabili y o he deb can also be summa ised in an cha s (Figu e
III.1), as we did o he de ici in sec ion 5. The bands delimi ed by he do ed lines show
he con idence in e als a a con idence le el o 90% (o 95% o unila e al con idence
le els).
Table III.1: Membe S a es anked by he s anda d de ia ion o e isions in deb (% o GDP)
1s
ansmission o
inal
2nd
ansmission o
inal
3 d
ansmission o
inal
4 h
ansmission o
inal
5 h
ansmission o
inal
6 h
ansmission o
inal
7 h
ansmission o
inal
1DE(0.33%) DE(0.27%) DE(0.26%) NL(0.16%) NL(0.12%) NL(0.03%) SE(0.01%)
2ES(0.53%) NL(0.38%) NL(0.31%) DE(0.17%) DE(0.17%) DE(0.12%) PT(0.01%)
3UK(0.54%) FI(0.39%) FI(0.33%) FI(0.29%) FI(0.23%) FR(0.21%) NL(0.03%)
4FI(0.56%) FR(0.41%) ES(0.38%) FR(0.30%) ES(0.23%) ES(0.22%) ES(0.05%)
5NL(0.62%) ES(0.50%) FR(0.39%) ES(0.36%) IT(0.28%) FI(0.23%) FI(0.05%)
6IT(0.68%) UK(0.53%) UK(0.41%) UK(0.40%) FR(0.28%) IT(0.26%) UK(0.07%)
7FR(0.72%) SE(0.64%) IT(0.48%) IT(0.45%) UK(0.31%) UK(0.29%) IT(0.09%)
8SE(0.79%) IT(0.65%) SE(0.62%) SE(0.60%) LU(0.58%) SE(0.45%) DE(0.12%)
9IE(1.03%) IE(0.99%) LU(0.81%) LU(0.79%) SE(0.60%) LU(0.54%) FR(0.14%)
10 LU(1.06%) LU(1.00%) IE(0.95%) IE(0.88%) IE(0.85%) IE(0.64%) LU(0.16%)
11 DK(1.82%) DK(1.76%) AT(1.54%) DK(1.49%) AT(0.92%) AT(0.90%) AT(0.23%)
12 BE(1.96%) BE(1.78%) DK(1.65%) AT(1.52%) DK(1.20%) BE(1.10%) DK(0.59%)
13 AT(2.24%) AT(2.11%) BE(1.66%) BE(1.53%) BE(1.28%) DK(1.13%) IE(0.60%)
14 PT(2.42%) PT(2.41%) PT(1.96%) PT(1.95%) PT(1.39%) PT(1.39%) BE(0.71%)
Re isions cumula ed om
No e: The able shows he s anda d de ia ions o e isions in deb igu es be ween any gi en da a
ansmission and inal da a. The i s column co esponds o he e isions be ween he i s ou u n o yea
epo ed in sp ing +1 and inal da a o he same yea as ansmi ed in au umn +4. A highe alue in
s anda d de ia ion o he e isions means mo e dispe sion in de ici da a o he same pe iod o e di e en
da a ansmissions and a wo se posi ion o he coun y in e ms o eliabili y.
35 G eece has no been sys ema ically conside ed in he ables and cha s. The s anda d de ia ion o
e isions in deb da a in G eece om he 1s ansmission o inal da a has been 5.78% o GDP and he
a e age e ision is +4.17% o GDP.
37
Table III.2: Membe S a es anked by he a e age o e isions in deb (% o GDP)
1s
ansmission o
inal
2nd
ansmission o
inal
3 d
ansmission o
inal
4 h
ansmission o
inal
5 h
ansmission o
inal
6 h
ansmission o
inal
7 h
ansmission o
inal
1PT(-1.22%) PT(-1.27%) PT(-0.78%) PT(-0.81%) PT(-0.39%) PT(-0.40%) FR(-0.06%)
2SE(-0.24%) SE(-0.20%) SE(-0.24%) SE(-0.20%) SE(-0.22%) SE(-0.15%) LU(-0.04%)
3BE(-0.23%) UK(-0.19%) UK(-0.12%) UK(-0.14%) UK(-0.07%) UK(-0.09%) NL(-0.02%)
4UK(-0.17%) FI(-0.04%) BE(0.01%) BE(-0.08%) NL(0.02%) BE(-0.03%) FI(-0.02%)
5FI(-0.16%) BE(-0.01%) FI(0.03%) NL(0.00%) FR(0.02%) NL(-0.01%) UK(-0.01%)
6NL(0.04%) FR(0.07%) FR(0.04%) FI(0.04%) FI(0.04%) FR(0.02%) SE(0.00%)
7DE(0.15%) NL(0.09%) NL(0.07%) DE(0.05%) DE(0.05%) DE(0.03%) PT(0.01%)
8FR(0.24%) DE(0.13%) DE(0.11%) FR(0.08%) LU(0.10%) FI(0.06%) IT(0.02%)
9LU(0.52%) LU(0.45%) LU(0.30%) LU(0.26%) BE(0.14%) ES(0.12%) ES(0.02%)
10 ES(0.57%) ES(0.47%) ES(0.37%) ES(0.28%) ES(0.15%) LU(0.13%) DE(0.03%)
11 IE(0.59%) IT(0.52%) IT(0.37%) IT(0.33%) IT(0.17%) IT(0.16%) AT(0.03%)
12 DK(0.67%) DK(0.54%) DK(0.50%) IE(0.46%) AT(0.33%) IE(0.24%) IE(0.15%)
13 IT(0.70%) IE(0.65%) IE(0.54%) DK(0.51%) IE(0.36%) DK(0.30%) DK(0.18%)
14 AT(1.20%) AT(1.19%) AT(0.72%) AT(0.71%) DK(0.44%) AT(0.33%) BE(0.19%)
Re isions cumula ed om
No e: The able shows he a e age e isions in deb igu es be ween any gi en da a ansmission and inal
da a. The i s column ( e isions om age 1 o age 8) co esponds o he e isions be ween he i s ou u n
o yea epo ed in sp ing +1 and inal da a o he same yea as ansmi ed in au umn +4. A nega i e
(posi i e) alue means an expec ed inc ease ( educ ion) in he deb .
Table III.3: Membe S a es anked by he a e age o e isions in deb in absolu e alues (% o
GDP)
1s
ansmission o
inal
2nd
ansmission o
inal
3 d
ansmission o
inal
4 h
ansmission o
inal
5 h
ansmission o
inal
6 h
ansmission o
inal
7 h
ansmission o
inal
1DE(0.28%) DE(0.17%) DE(0.15%) DE(0.08%) DE(0.08%) NL(0.02%) SE(0.00%)
2UK(0.54%) SE(0.44%) FI(0.29%) NL(0.16%) NL(0.10%) DE(0.04%) PT(0.01%)
3SE(0.73%) FI(0.44%) UK(0.33%) FI(0.18%) FI(0.10%) FI(0.09%) FI(0.02%)
4FI(0.74%) UK(0.45%) NL(0.35%) UK(0.27%) UK(0.18%) UK(0.13%) NL(0.02%)
5ES(0.75%) NL(0.50%) SE(0.37%) SE(0.31%) ES(0.19%) ES(0.15%) ES(0.02%)
6NL(0.86%) FR(0.55%) ES(0.45%) FR(0.33%) FR(0.26%) SE(0.15%) UK(0.04%)
7FR(0.90%) ES(0.65%) FR(0.47%) ES(0.36%) SE(0.29%) FR(0.16%) DE(0.04%)
8IT(1.05%) IT(0.88%) LU(0.61%) LU(0.50%) IT(0.30%) LU(0.23%) IT(0.05%)
9LU(1.11%) LU(0.90%) IT(0.63%) IT(0.54%) LU(0.30%) IT(0.23%) LU(0.05%)
10 IE(1.49%) IE(1.31%) PT(1.00%) AT(0.90%) PT(0.44%) AT(0.40%) FR(0.08%)
11 PT(1.67%) PT(1.53%) AT(1.06%) PT(0.91%) AT(0.49%) PT(0.44%) AT(0.09%)
12 DK(1.93%) DK(1.59%) IE(1.15%) IE(0.92%) DK(0.70%) IE(0.47%) DK(0.18%)
13 AT(2.15%) AT(1.63%) DK(1.34%) DK(1.06%) IE(0.80%) DK(0.56%) BE(0.23%)
14 BE(2.41%) BE(1.84%) BE(1.49%) BE(1.23%) BE(0.82%) BE(0.60%) IE(0.36%)
Re isions cumula ed om
No e: The able shows he a e age e isions in deb in absolu e alues be ween any gi en da a
ansmission and inal da a. The i s column co esponds o he absolu e e isions be ween he i s
ou u n o yea epo ed in sp ing +1 and inal da a o he same yea as ansmi ed in au umn +4. A
highe alue in his indica o means less eliable igu es.
38
Figu e III.1: Fan cha s illus a ing he eliabili y o go e nmen deb igu es.
Belgium
85%
90%
95%
100%
105%
110%
2001 2002 2003 2004 2005
% o GDP
Denma k
30%
35%
40%
45%
50%
55%
2001 2002 2003 2004 2005
% o GDP
Ge many
55%
60%
65%
70%
75%
80%
2001 2002 2003 2004 2005
% o GDP
Spain
40%
45%
50%
55%
60%
65%
2001 2002 2003 2004 2005
% o GDP
F ance
55%
60%
65%
70%
75%
80%
2001 2002 2003 2004 2005
% o GDP
I eland
25%
30%
35%
40%
45%
50%
2001 2002 2003 2004 2005
% o GDP
39
I aly
100%
105%
110%
115%
120%
125%
2001 2002 2003 2004 2005
% o GDP
Luxembou g
0%
5%
10%
15%
20%
25%
2001 2002 2003 2004 2005
% o GDP
Ne he lands
45%
50%
55%
60%
65%
70%
2001 2002 2003 2004 2005
% o GDP
Aus ia
50%
55%
60%
65%
70%
75%
2001 2002 2003 2004 2005
% o GDP
Po ugal
50%
55%
60%
65%
70%
75%
2001 2002 2003 2004 2005
% o GDP
Finland
30%
35%
40%
45%
50%
55%
2001 2002 2003 2004 2005
% o GDP