Heinisch, Ka ja e al.
A icle
The IWG Fo ecas ing Dashboa d: F om o ecas s o
e alua ion and compa ison
Jou nal o Economics and S a is ics
P o ided in Coope a ion wi h:
De G uy e B ill
Sugges ed Ci a ion: Heinisch, Ka ja e al. (2024) : The IWG Fo ecas ing Dashboa d: F om o ecas s
o e alua ion and compa ison, Jou nal o Economics and S a is ics, ISSN 2366-049X, De G uy e
Oldenbou g, Be lin, Vol. 244, Iss. 3, pp. 277-288,
h ps://doi.o g/10.1515/jbns -2023-0011
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/333284
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h ps://c ea i ecommons.o g/licenses/by/4.0/
Da a Obse e
Ka ja Heinisch*, Ch is oph Beh ens, Jö g Döpke, Alexande Fol as,
Ul ich F i sche, Tim Köhle , Ka s en Mülle , Johannes Puckelwald
and Hannes Reichmay
The IWH Fo ecas ing Dashboa d: F om
Fo ecas s o E alua ion and Compa ison
h ps://doi.o g/10.1515/jbns -2023-0011
Recei ed Feb ua y 3, 2023; accep ed Feb ua y 4, 2023
Abs ac : The pape desc ibes he “Halle Ins i u e o Economic Resea ch (IWH)
Fo ecas ing Dashboa d (Fo Das)”. This ool aims a p o iding, on a non-comme cial
basis, his o ical and ac ual mac oeconomic o ecas da a o he Ge many economy
o esea che s and in e es ed audiences. The da abase ende s i possible o di ec ly
compa e o ecas quali y ac oss selec ed ins i u ions and o e ime. I is pa ly based
on da a collec ed in he DFG- unded p ojec “Mac oeconomic o ecas s in g ea
c isis”.
Keywo ds: o ecas ing, mac oeconomic da a
We hank Ida Rockenbach, Ronny Ehlen, Au o a Li, Max Weinig, Sebas ian P ze ak, Ch is ophe Glu h, Tim
Be nu z, and Kay Felix Domke o da a suppo . Fu he , we hank Oli e Hol emölle o his use ul
commen s on implemen ing he dashboa d.
*Co esponding au ho : Ka ja Heinisch, Halle Ins i u e o Economic Resea ch (IWH), Kleine
Mae ke s aße 8, D-06108 Halle (Saale), Ge many, E-mail: [email p o ec ed]. h ps://o cid.o g/
0000-0002-2664-4950
Ch is oph Beh ens, F eie und Hanses ad Hambu g, Hambu g, Ge many,
E-mail: [email p o ec ed]
Jö g Döpke and Tim Köhle , Hochschule Me sebu g, Me sebu g, Ge many, E-mail: joe g.doepke@hs-
me sebu g.de (J. Döpke), [email p o ec ed] (T. Köhle )
Alexande Fol as, Helmu -Schmid -Uni e si ä Hambu g, Hambu g, Ge many, E-mail: ol asa@hsu-
hh.de. h ps://o cid.o g/0000-0003-2146-8158
Ul ich F i sche, Uni e si ä Hambu g, Hambu g, Ge many, E-mail: [email p o ec ed].
h ps://o cid.o g/0000-0003-0492-8300
Ka s en Mülle , Deu sches Zen um ü Lu - und Raum ah , Ins i u ü Ve ne z e Ene giesys eme,
S u ga , Ge many, E-mail: [email p o ec ed]
Johannes Puckelwald, Deu sches Ma i imes Zen um Hambu g, Hambu g, Ge many,
E-mail: Puckelwald@dmz-ma i im.de
Hannes Reichmay , Ma in-Lu he -Uni e si ¨
a Halle-Wi enbe g, Halle, Ge many,
E-mail: [email p o ec ed]
Jou nal o Economics and S a is ics 2024; 244(3): 277–288
Open Access. © 2023 he au ho (s), published by De G uy e . This wo k is licensed unde he
C ea i e Commons A ibu ion 4.0 In e na ional License.
JEL Classifica ion: E32, E37, G11
1 In oduc ion
Mac oeconomic o ecas ing is one o he a eas o economics ha ecei es he mos
a en ion om he media and public discou se. Fo example, o ecas compe i ions
a e qui e popula : Se e al newspape s and da abase p o ide s e alua e o ecas e s
mo e o less egula ly (see Döh n 2015, o an o e iew o Ge man ankings). Also,
poin o ecas s by ins i u ions gain a lo o a en ion om he media and poli ics. Da a
p o ide s such as Consensus Economics o Focus Economics use he ins i u ions’
o ecas s o p oduce a mean o ecas ac oss se e al o ecas e s. As a esul , a
significan pa o he scien ific li e a u e is ocused on e alua ing hese o ecas s.
Howe e , ypically o ecas s a e e alua ed agains a heo e ical benchma k like
a nai e o ecas bu a ely ac oss o ecas ing ins i u ions. One eason o he ela i e
sca ci y o such analyses is he lack o non-comme cial da abases ha include se e al
ins i u ions, mac oeconomic indica o s, and o ecas pe iods. The IWH Fo ecas ing
Dashboa d (Fo Das) by he Halle Ins i u e o Economic Resea ch aims o fill he da a
gap by p o iding his o ical and ac ual da a o his pu pose.
1
Fu he mo e, on can
di ec ly compa e o ecas ing quali y o ins i u ions. The dashboa d elies pa ly on
da a collec ed wi hin he DFG p ojec “Mac oeconomic o ecas s in g ea c isis”.
2
The
IWH ook o e he da a collec ion p ocess in 2020, con inues wi h i , and has
implemen ed a ool o isualize he da a. Due o da a gaps and diffe en dimensions
( o example, g ow h s. le el p edic ions) p o ided by se e al ins i u ions, he
dashboa d only shows a selec ion o a iables. In he ollowing, we desc ibe he main
con en s
3
o he da abase and he IWH Fo ecas ing Dashboa d.
Since he da abase was ini ia ed fi s wi hin he con ex o economic his o y, we
aimed a collec ing all quan i a i e o ecas s made o he Ge man economy by
ins i u ions ele an o economic policy. As (An holz 2006) poin s ou , hese o ecas s
s a ed in he ea ly 1960s. While i was possible o find ex s desc ibing he p ospec s
o he Ge man economy be o e, hey a ely included any conc e e numbe s o GDP
g ow h o infla ion. Hence, analyses and compa isons based on Fo Das da a can s a
in 1965 a he ea lies o selec ed o ecas e s.
4
1h ps://www.iwh-halle.de/ o das.
2This p ojec has been pa o he P io i y P og amme “Expe iences and Expec a ions: His o ical
Founda ions o Economic Beha iou ”(SPP 1859) unded by he Deu sche Fo schungsgemeinscha (DFG).
3Please no e, ha Fo Das does no ye con ain he ull o ecas ex epo s. This is howe e planned
o he u u e.
4In he 1970s, figu es we e o en p o ided as a ac ional numbe and no decimal numbe , which is
line wi h a so called o ecas confidence in e al. Fo ins ance, a alue o 3 1
2is ew i en o 3.5 du ing he
da a collec ion, howe e , policy make s should ha e in mind a ange be ween 3.35 and 3.65.
278 K. Heinisch e al.
2 Con en s o he Da abase
The main objec i e o he DFG p ojec was o build a mac oeconomic o ecas
da abase, including a b oad se o o ecas s o Ge many o e he longes possible
pe iod, achie ing consis ency wi hin he da ase as a as possible. In combina ion
wi h use - iendliness and ee access o he da a, he da abase supplies a new
ins umen o o ecas e alua ion, made o he scien ific communi y, media, o he
public in gene al.
To his end, he o ecas da abase co e s he key mac oeconomic a iables a
each poin in ime, including a iables ela ed o na ional accoun s, financial and
mone a y a iables, and also a iables ela ed o labo and (un-)employmen . We
choose he ins i u ions acco ding o hei long- e m expe ience in o ecas ing and
hei ele ance o economic policy suppo in Ge many. The field o mac oeconomic
o ecas ing has ma kedly g own o e ime. The e o e, he quan i y o o ecas s and
o ecas e s has also inc eased subs an ially. To gi e an imp ession o he g ow h o
he “ o ecas ing indus y”du ing he pe iod co e ed, Figu e 1 shows he shee
numbe o o ecas s ega ding he headline measu e o economic g ow h pe yea
included in he da abase. Fo he la ge pa o he pe iod, his headline measu e was
he a e o change o eal GDP. In a smalle pa o he sample, howe e , eal GNP
g ow h se ed as he mos p ominen figu e in his con ex and is, hus, also aken
in o accoun . The numbe o o ecas s is bo h d i en by he numbe o o ecas ing
ins i u ions as well as he o ecas equency pe yea . While a he beginning o he
sample, o ecas s a e published once pe yea , while up o ou o ecas s a e
Figu e 1: Numbe o o ecas s o he headline figu e o economic g ow h in he da abase, 1965 o
2019. Sou ce: Own compila ion.
The IWH Fo ecas ing Dashboa d 279
published ecen ly by o ecas e s. The numbe o a iables o in e es o he o e-
cas e s also g ew o e ime. Defini ions and names o a iables a ied o e ime as
well, which o some ex en eflec ed changes in he sys em o na ional accoun s.
2.1 Fo ecas e s Co e ed
The da abase includes o ecas da a on he Ge man economy co e ed by 15 na ional
and in e na ional ins i u ions wi h diffe en ins i u ional backg ounds:
–The six la ges Ge man economic esea ch ins i u es: Ge man Ins i u e o
Economic Resea ch (DIW), i o Ins i u e o Economic Resea ch (i o), RWI –Leibniz
Ins i u e o Economic Resea ch (RWI), Halle Ins i u e o Economic Resea ch
(IWH), Hambu g Ins i u e o In e na ional Economics (HWWA, since 2007 HWWI),
and Kiel Ins i u e o he Wo ld Economy (I W).
–Resea ch ins i u es ela ed o ade unions o employe s’associa ions: Mac o-
economic Policy Ins i u e (IMK) and Ge man Economic Ins i u e (IW).
5
–Ins i u ion ela ed o policy o policy ad ice: he Deu sche Bundesbank (Ge man
cen al bank BBK), he join o ecas (GD) o he leading esea ch ins i u es, and
he Ge man Council o Economic Expe s (SVR).
–In e na ional o ganisa ions, namely: he O ganisa ion o Economic Coope a-
ion and De elopmen (OECD), he In e na ional Mone a y Fund (IMF), and he
Eu opean Commission (EC).
6
2.2 Va iables Included
The da abase co e s key mac oeconomic a iables (Table 1). In addi ion, financial
and mone a y a iables as well as da a ela ed o ade, labo , employmen , and
unemploymen a e included.
7
Howe e , no e ha no all o ecas ing ins i u ions
5Ins i u ions o me ly in ol ed in o ecas ing a e also co e ed, i.e. he Wi scha s-und sozialwis-
senscha liches Ins i u in de Hans-Boeckle -S i ung (WSI) up o 2004. Al hough s ill exis ing, he
WSI ins i u e has no p o ided business cycle o ecas s since he IMK came in o exis ence. Second, he
Hambu ge Wel wi scha sa chi (HWWA) un il 2006. This ins i u e was mainly unded by public
money. F om 2007 onwa ds, he ins i u e was enamed o HWWI and unc ions as a p i a ely unded
ins i u e.
6No e, ha he o ecas s o he EC a e pa o he IWH Fo ecas ing Dashboa d, bu no o he da a se
used in he DFG p ojec .
7Fo he sake o b e i y, in his pape we desc ibe he mos ecen e sion o he da abase only,
namely he one accessible o e he IWH homepage as explained below. The e o e, in he IWH
Fo ecas ing Dashboa d, only a selec ion o a iables is chosen ha is co e ed by mos o he o e-
cas e s. A e sion wi h a b oade da ase , in pa icula olde da a, collec ed o he DFG p ojec , can
be ound in he p ojec ’s da a eposi o y “Empo ion”, see: h ps://empo ion.gswg.in o.
280 K. Heinisch e al.
men ioned abo e p o ide o ecas s o all a iables in all pe iods.
8
Fu he mo e,
ins i u ions migh p o ide o ecas s ei he in le els o g ow h a es.
2.3 Realisa ions (Ac ual Da a) and Fo ecas Ho izons
Fo he compa ison o he o ecas da a wi h he ac ual economic de elopmen , he
so-called “ eal- ime”da a p oblem is an impo an issue (S a k and C ousho e 2002)
in economic o ecas ing and o ecas e alua ion. Since ime se ies a e p one o
e isions, he da abase p o ides ealisa ions o selec ed impo an se ies in wo
Table :O e iew o he a iables in Fo Das.
Componen s o Real GDP
G oss domes ic p oduc (cons an p ices)
Final consump ion expendi u e o households and NPISHs (cons an p ices)
Go e nmen final consump ion expendi u e (cons an p ices)
G oss fixed capi al o ma ion (cons an p ices)
G oss fixed capi al o ma ion in machine y and equipmen (cons an p ices)a
Cons uc ion (cons an p ices)
G oss fixed capi al o ma ion in o he fixed asse s (cons an p ices) (since )
Expo s o goods and se ices (cons an p ices)
Impo s o goods and se ices (cons an p ices)
Labo ma ke
Pe sons in employmen
Unemploymen egis e ed (pe sons) acco ding o “Bundesagen u ü A bei ”
Unemploymen egis e ed ( a io)
Unemploymen acco ding o in e na ional labo o ganisa ion
Unemploymen a e
Memo andum I ems
Consume p ice index
GDP defla o
Uni labo cos s
Ne lending/ne bo owing as a pe cen age o GDP
Cu en accoun balance as a pe cen age o GDP
Technical Assump ions
Oil p ice
Exchange a e
Policy in e es a e
Wo ld ade g ow h
Own compila ion. aNo e ha o ecas e s ha e o en summa ised G oss fixed capi al o ma ion
in machine y and equipmen and g oss fixed capi al o ma ion in o he p oduc s un il .
8See he da a a ailabili y sec iono he IWH Fo ecas ing Dashboa d o see he ime span andmissing
obse a ions o selec ed a iables and ins i u ions.
The IWH Fo ecas ing Dashboa d 281
a ian s: fi s , he ini ial publica ion (fi s elease) and, second, he e ised da a
(cu en da a in age) by he Ge man s a is ical office, i a ailable.
The da abase includes in o ma ion on h ee possible o ecas ho izons: he
cu en yea , i.e. he yea in which he o ecas is eleased (labeled
0
in he da abase),
a o ecas o he nex yea (
1
), and he yea a e he nex yea (
2
). Hence, ollowing
he p ac ice o he ins i u ions co e ed, he o ecas s in he da abase a e “fixed e en ”
a he han “fixed ho izon”p edic ions (see, e.g. Knüppel and Vladu 2016).
Howe e , mos o he ins i u ions publish mul iple o ecas s pe yea ( o ecas
ounds). The e o e, he exac da e o o ecas publica ion is s o ed, o dis inguish
diffe en o ecas ounds (e.g. qua e s, mon hs). In addi ion, i a ailable, he da e on
which he o ecas was comple ed is epo ed. Gene ally, he da abase e e s o he
name o a a iable as i was a he da e o he p oduc ion o he o ecas .
9
In a simila
ein, all dimensions e e o he da e o he o ecas s. Thus, a iables a e exp essed in
Deu sche Ma k up o 2000, and in Eu os a e . Real a iables usually e e o he
espec i e base yea . A ound Ge man eunifica ion, he swi ch om o ecas s
e e ing o Wes Ge many o p edic ions o Ge many as a whole diffe s by se ies
and by he ins i u ion and is, hence, no ed simila ly o each se ies.
10
Figu e 2 shows he eal GDP g ow h o ecas s o Ge many, p o ided by
he economic esea ch ins i u es as well as he ealised alues om 2001 o 2022. The
Figu e 2: Realised g ow h a e and o ecas s, 2000 o 2022. Sou ce: IWH Fo ecas ing Dashboa d, 2023.
9This is impo an o cases in which he official name in he na ional accoun s has changed. Fo
example, un il a ce ain da e, he da abase e e s o “G oss na ional p oduc ”, ollowed by “G oss
Na ional Income”in la e yea s.
10 The disen anglemen om Wes Ge man o Ge man o ecas s is no uni o m in he pe iod o
unifica ion ac oss o ecas e s. The e o e, we excluded he yea 1991 o he calcula ion o o ecas
e o s in he IWH Fo ecas ing Dashboa d.
282 K. Heinisch e al.
o ecas s ha e been conduc ed in au umn (mon hs 9, 10) o he p e ious yea . The
black line ep esen s he ealised g ow h a e. The figu e illus a es, a fi s glance,
some key insigh s ega ding economic o ecas e alua ion. Fi s , he o ecas s a e
ela i ely close o each o he , and i seems ha hey do no diffe significan ly o e
ime. Second, he p edic ion o economic u ning poin s (and/o ecessions) is s ill a
big challenge in economic o ecas ing. Diffe en o ecas ing da es and di e ging
in o ma ion se s seem o be impo an in de e mining o ecas accu acy.
3 Recen and Possible Applica ions
The IWH Fo ecas ing Dashboa d p o ides he basis o a ious po en ial esea ch
ques ions, mos ob iously e alua ing Ge man business cycle o ecas s based on hei
accu acy o efficiency. Po en ial o he uses could be he in es iga ion o he ins i u es
hemsel es, hei beha io , and change a e economically significan e en s o
due o pa adigma ic shi s. An addi ional esea ch field conce ns assessing business
cycle o ecas s’benefi s o economic agen s. Se e al s udies ha e al eady p oduced
scien ificfindings based on he da a a ailable on he IWH Fo ecas ing Dashboa d.
(Köhle and Döpke 2023) use he IWH Fo ecas ing Dashboa d o conduc an
o e all anking o 14 ins i u ions om 1993 o 2019 acco ding o hei o ecas
accu acy. They epo subs an ial long- un diffe ences in o ecas ing quali y, which
hey mos ly a ibu e o dis inc a e age o ecas ho izons. The e o e, hey canno
single ou ins i u ions as being supe io a p edic ing he Ge man economy. (Engelke
e al. 2019) examine he ex en o which ini ial assump ions ha p o e inco ec
ex-pos d i e economic o ecas e o s. Based on an unbalanced panel o annual
o ecas s om diffe en ins i u ions o ecas ing Ge man GDP and he unde lying
assump ions, hey ound ha o e 75% o squa ed e o s o he GDP o ecas co-mo e
wi h he squa ed e o s in hei unde lying assump ions. This finding implies ha
he accu acy o he assump ions is o g ea impo ance and ha o ecas e s should
e eal he amewo k o hei assump ions in o de o ob ain use ul policy ecom-
menda ions based on economic o ecas s. The impac o he G ea Recession on
o ecas accu acy o g ow h and infla ion and o ecas e beha iou a e in es iga ed
by (Döpke e al. 2019) using a da a panel om 1971 o 2017. The au ho s epo s able
accu acy o g ow h o ecas s, bu sligh ly lowe p ecision o infla ion o ecas s.
Mo e significan ly, hey epo ha he loss unc ion has changed a e he G ea
Recession, leading o mo e pessimis ic o ecas s om Ge man p o essional
o ecas e s.
(Beh ens e al. 2018a) e alua e whe he g ow h and infla ion o ecas s a e
efficien o op imal, which equi es ha he in o ma ion a ailable a he ime o
o ecas c ea ion has no explana o y powe o he co esponding o ecas e o . The
The IWH Fo ecas ing Dashboa d 283
join o ecas efficiency e alua ion shows he e ogenei y ac oss he ins i u es, wi h
diffe en ins i u es conduc ing inefficien o ecas s o diffe en p ognosis ho izons.
(Beh ens e al. 2020) confi m his esul , which ex ends he p e ious s udy wi h
a ious scena ios and obus ness checks, ejec ing s ong and weak o ecas effi-
ciency o g ow h and infla ion o ecas s in mul iple cases. Addi ionally, he au ho s
show in an ou -o -sample expe imen ha a Bayesian addi i e eg ession ees
(BART) model p oduces significan ly mo e accu a e o ecas s. (Beh ens 2020) finds
ade o ecas s simila ly he e ogeneously inefficien . Rema kably, he o ecas e s
include ypical ade p edic o s mo e efficien ly han mac oeconomic a iables in
hei expo and impo o ecas s.
Fu he esea ch examines o ecas efficiency while assuming a flexible ins ead
o a symme ic (quad a ic) loss unc ion. Fo his pu pose, he esea che s es
whe he he se o p edic o s has p edic i e alue o he sign o he o ecas e o
using andom decision o es s. Re-e alua ing infla ion o ecas op imali y, (Beh ens
e al. 2018b) sugges ha sho - e m infla ion o ecas s a e subop imal o some
ins i u es while ailing o ejec he null hypo hesis o long- e m o ecas s. Recon-
side ing ade o ecas s, (Beh ens 2019) ejec s op imali y only in one case, hus
suppo ing a mo e a o able assessmen o o ecas s i flexible loss unc ions a e
assumed. Se e al s udies ha e implemen ed ex ual business cycle epo s using
na u al language p ocessing (NLP) in hei o ecas efficiency analyses. While he
w i en epo s a e no pa o he IWH Fo Das i sel , hey a e sou ce and a ionale o
he o ecas s, making a combined analysis easonable. (Mülle 2022) ans o ms he
w i en accoun s o o ecas e s’expec a ions in o sen imen indices using nine
diffe en me hods. The au ho demons a es ha se e al indices can imp o e he
accu acy o Ge man business cycle o ecas s p o ing ha o ecas e s do no ully
exploi he in o ma ion con en o hei business cycle epo s o hei nume ical
poin o ecas s. (Fol as 2022) uses he Wo d2Sense-LDA opic model de eloped
specifically o his ask o measu e he p opo ions o diffe en economic opics in
each business cycle epo and uses hei shi o es in es men o ecas efficiency.
In some cases, he au ho ejec s o ecas efficiency wi h opics as he mos impo an
p edic o s suppo ing he hesis ha ins i u es inefficien ly inco po a e quali a i e
in o ma ion discussed in hei business cycle epo s in o poin o ecas s. Wi h an
app oach using opics as he sole p edic o s o he o ecas e o , (Fol as and
Pie dzioch 2022a) affi m he use ulness o opic modeling o o ecas efficiency
analysis. The au ho s find se e al in e p e able opics ela ed o he o ecas e o
unde symme ic and flexible loss unc ions. Las ly, (Fol as and Pie dzioch 2022b)
u ilise a mixed sample o indica o s and opics o p edic g ow h o ecas e o s
using quan ile andom o es s and ou -o -sample densi y o ecas s. E en hough
none o he opics a e among he op p edic o s, hei agg ega ed ela i e impo ance
a ied be ween oughly 30–50%.
284 K. Heinisch e al.
