Geche , Sebas ian e al.
Wo king Pape
Con en ional wisdom, me a-analysis, and esea ch
e ision in economics
FMM Wo king Pape , No. 95
P o ided in Coope a ion wi h:
Mac oeconomic Policy Ins i u e (IMK) a he Hans Boeckle Founda ion
Sugges ed Ci a ion: Geche , Sebas ian e al. (2024) : Con en ional wisdom, me a-analysis,
and esea ch e ision in economics, FMM Wo king Pape , No. 95, Hans-Böckle -S i ung,
Mac oeconomic Policy Ins i u e (IMK), Fo um o Mac oeconomics and Mac oeconomic Policies
(FMM), Düsseldo
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/297858
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/legalcode
FMM WORKING PAPER
No. 95 • Janua y 2024 • Hans-Böckle -S i ung
CONVENTIONAL WISDOM,
META
-ANALYSIS, AND
RESEARCH REVISION IN
ECONOMICS
Sebas ian Geche
1, Bianka Mey2, Ma ej Opa ny3, Tomas Ha anek4, T. D. S anley5,
Ped o R.D. Bom
6
, H is os Doucouliagos
7
, Philipp Heimbe ge
8
, Zuzana I so a
9 ,
Heiko J. Rachinge
10
ABSTRACT
O e he pas se e al decades, me a-analysis has eme ged as a widely accep ed ool o
unde s and economics esea ch. Me a-analyses o en challenge he es ablished
con en ional wisdom o hei espec i e ields. We sys ema ically e iew a wide ange o
in luen ial me a-analyses in economics and compa e hem o ‘con en ional wisdom.’
A e co ec ing o obse able biases, he empi ical economic e ec s a e ypically much
close o ze o and some imes swi ch signs. Typically, he ela i e educ ion in e ec sizes
is 45-60%.
—————————
1 Chemni z Uni e si y o Technology; FMM Fellow
2 Chemni z Uni e si y o Technology
3 Cha les Uni e si y P ague
4 Ins i u e o Economic S udies, Facul y o Social Sciences, Cha les Uni e si y P ague; CEPR, London;
Me a-Resea ch Inno a ion Cen e a S an o d
5 Co esponding Au ho . S anley@hend ix.edu. Depa men o Economics, Deakin Uni e si y;
Me a-Resea ch Inno a ion Cen e a S an o d
6 Deus o Business School, Uni e si y o Deus o
7 Depa men o Economics, Deakin Uni e si y; IZA Bonn
8 Vienna Ins i u e o In e na ional Economic S udies (wiiw); FMM Fellow
9 Anglo-Ame ican Uni e si y, P ague
10 Uni e si a de les Illes Balea s, Mallo ca
Con en ional Wisdom, Me a-Analysis, and
Resea ch Re ision in Economics
Sebas ian Geche 1, Bianka Mey2, Ma ej Opa ny3, Tomas Ha anek4, T. D. S anley5,
Ped o R.D. Bom6, H is os Doucouliagos7, Philipp Heimbe ge 8, Zuzana I so a9&
Heiko J. Rachinge 10
1Chemni z Uni e si y o Technology; FMM Fellow.
2Chemni z Uni e si y o Technology
3Cha les Uni e si y P ague
4Ins i u e o Economic S udies, Facul y o Social Sciences, Cha les Uni e si y P ague; CEPR, London;
Me a-Resea ch Inno a ion Cen e a S an o d
5Co esponding Au ho . [email p o ec ed]du. Depa men o Economics, Deakin Uni e si y;
Me a-Resea ch Inno a ion Cen e a S an o d
6Deus o Business School, Uni e si y o Deus o
7Depa men o Economics, Deakin Uni e si y; IZA Bonn
8Vienna Ins i u e o In e na ional Economic S udies (wiiw); FMM Fellow
9Anglo-Ame ican Uni e si y, P ague
10Uni e si a de les Illes Balea s, Mallo ca
Decembe 22, 2023
Abs ac . O e he pas se e al decades, me a-analysis has eme ged as a
widely accep ed ool o unde s and economics esea ch. Me a-analyses o en
challenge he es ablished con en ional wisdom o hei espec i e ields. We
sys ema ically e iew a wide ange o in luen ial me a-analyses in economics
and compa e hem o ‘con en ional wisdom.’ A e co ec ing o obse able
biases, he empi ical economic e ec s a e ypically much close o ze o and
some imes swi ch signs. Typically, he ela i e educ ion in e ec sizes is
45-60%.
Keywo ds. me a-analysis, sys ema ic e iew, con en ional wisdom
JEL classi ica ion. A14,B40,C10
1
1 In oduc ion
Dominan economic heo ies, seminal s udies, and au ho i a i e li e a u e e iews o en
in o m he con en ional wisdom. P ac ical policy ecommenda ion and implemen a ion
equi e knowledge o he speci ic alues o impo an economic pa ame e s; o exam-
ple, he employmen e ec s o minimum wage hikes, e u ns o educa ion, he iscal
mul iplie , he p ice elas ici y o ene gy demand, o he in e empo al elas ici y o sub-
s i u ion. Such con en ional wisdom o en de ines he scope o public policy discussions
and is used o calib a e economic models wi h hese speci ic pa ame e alues. Howe e ,
con en ional wisdom can also lead us as ay.
Me a-analysis is he sys ema ic and s a is ical analysis o all compa able empi ical
es ima es o a speci ic pa ame e . I seeks o summa ize, e alua e, and unde s and wha
we know abou a gi en empi ical economic ques ion, phenomenon, policy pa ame e , o
e ec . Me a-analyses published in he Jou nal o Economic Su eys a e compelled o
ollow guidelines ha speci y minimum s anda ds o he coding, conduc ing, analyzing,
and he epo ing o quan i a i e su eys o economics esea ch (S anley, Doucouliagos,
Giles, e al. 2013; Ha ánek, S anley, e al. 2020). Me a- eg ession analysis (MRA) was
de eloped speci ically o explain and summa ize he ich he e ogenei y ound among
epo ed empi ical economic es ima es (S anley and Ja ell 1989; S anley 2001). By now,
housands o MRAs ha e been conduc ed on economic opics, wi h some hund ed(s) o
new s udies p oduced each yea (Ha ánek, S anley, e al. 2020). MRA, wi h i s abili y
o accommoda e publica ion selec ion bias, was conside ed su icien ly impo an o
unde s anding economics esea ch o de o e a special issue o he Jou nal o Economic
Su eys (Robe s and S anley 2005). Me a-analysis can e eal su p ising u hs abou
economics once publica ion selec ion and mis-speci ica ion biases ha e been iden i ied
and accommoda ed. Thus, a conside able numbe o me a-analyses in economics ha e
ques ioned con en ional wisdom in hei espec i e ields.
2
This pape is a e iew o me a-analyses in he spi i o Ioannidis e al. (2017), Doucou-
liagos and S anley (2013), Doucouliagos, Paldam, e al. (2018), and Geche (2022). The
pu pose o his s udy is o compa e he indings o in luen ial me a-analyses o he ‘con-
en ional wisdom’ abou he same economic ques ion o issue. Wha ha e we lea ned
om me a-analyses o economics? How do hei esul s di e om he con en ional,
ex book unde s anding o economics?
We iden i y ‘in luen ial’ me a-analyses as hose wi h a leas 100 ci a ions ha we e
published in 2000 o la e , and hose ha we e ecommended by a su ey o membe s o
he Me a-Analysis o Economics Resea ch Ne wo k (MAER-Ne ) (h ps://www.mae -
ne .o g/). Ou o he ull sample o 360 s udies, 72 s udies co e a gene al in e es opic
in economics and include o iginal empi ical es ima es o a ce ain e ec size. We na ow
down u he o hose me a-analyses ha p o ide bo h a simple mean o he o iginal
e ec size and a co ec ed mean, con olling o publica ion bias o o he biases. This
gi es us a inal lis o 24 s udies co e ing he ields o g ow h and de elopmen , inance,
public inance, educa ion, in e na ional, labo , beha io al, gende , en i onmen al, and
egional/u ban economics.
We compa e he cen al indings o he me a-analyses o ‘con en ional wisdom’ as
classi ied by: (1) a widely ecognized seminal pape o au ho i a i e li e a u e e iew;
(2) he assessmen o an a i icial in elligence (AI), he GPT-4 La ge Language Model
(LLM); and (3) he simple unweigh ed a e age o epo ed e ec s included in he me a-
analysis.
Fo 17 o hese 24 s udies, he co ec ed e ec size is subs an ially close o ze o han
commonly hough , o e en swi ches sign. S a is ically signi ican publica ion bias is
p e alen in 17 o he 24 s udies. O e all, we ind ha 16 o 24 s udies show bo h a clea
educ ion in e ec size and a s a is ically signi ican publica ion bias. Compa ing he
bes es ima e om he me a-analysis wi h he con en ional wisdom om he e e ence
s udy, he GPT-4 es ima e, o he simple unweigh ed a e age, he ela i e educ ion
3
in he e ec size is in he ange o 45-60% in all h ee compa ison cases. This is close
o “Paldam’s ule o humb,” acco ding o which publica ion bias ypically in la es he
unco ec ed mean o he e ec size by a ac o o wo (Doucouliagos, Paldam, e al. 2018;
Paldam 2022).
The pape is o ganized as ollows: Sec ion 2 e iews he ela ed li e a u e on he
p e alence o publica ion selec ion bias. Sec ion 3 desc ibes how we selec ed he me a-
s udies in ou inal da ase and he in o ma ion we collec ed. Sec ion 4 p o ides a b ie
quali a i e discussion o he con ibu ion o selec ed me a-s udies. Sec ion 5 hen shows
he quan i a i e esul s om ou su ey. The inal sec ion concludes.
2 Publica ion selec ion bias: a enaissance
Fo many decades, publica ion selec ion bias has been widely ecognized as a se ious
h ea o he alidi y o empi ical science (S e ling 1959; Rosen hal 1979; Lo ell 1983;
Hedges and Olkin 1985; DeLong and Lang 1992; Ca d and K uege 1995; Ioannidis
2005; S anley and Ja ell 2005; S anley 2008; S anley and Doucouliagos 2012; S anley
and Doucouliagos 2014, o ci e bu a ew). Publica ion selec ion bias is he p ocess
o selec ing which esea ch indings o epo based on hei s a is ical signi icance o
hei consis ency wi h con en ional economic heo y. Publica ion selec ion bias is he
consequence o any ype o p e e en ial epo ing o s a is ically signi ican indings, in-
cluding he ile d awe p oblem, publica ion bias, epo ing bias, speci ica ion sea ching,
ques ionable esea ch p ac ices, and p-hacking. As amously exposed by Leame (1983),
epo ed economic empi ical indings a e he consequence o he pa icula speci ica ion
o innume able combina ions o independen a iables, models, and me hods (Sala-i-
Ma in 1997).
E idence o exagge a ed signi icance and e ec size has been widely seen h oughou
he economics esea ch li e a u e. Fo example, a su ey o 64,076 es ima es om 159
a eas o economics esea ch ound ha epo ed esul s a e ypically exagge a ed by a
4
ac o o wo o mo e (Ioannidis e al. 2017). Two highly powe ed eplica ion s udies
o mul iple economics and beha io al expe imen s co obo a e his doubling o e ec s
size (Came e , D ebe , Fo sell, e al. 2016; Came e , D ebe , Holzmeis e , e al. 2018).
Doucouliagos and S anley (2013) show in a me a-me a-analysis o 87 empi ical economics
li e a u es ha mo e compe i ion and deba e be ween i al heo ies (i.e., mo e plu alism)
educes publica ion bias. Me hods o de ec and co ec publica ion selec ion bias we e
in oduced and widely applied o economics in he May 2005 issue o he Jou nal o
Economics Su eys (Robe s and S anley 2005). Since hen i has been s anda d o
in es iga e publica ion selec ion bias when conduc ing me a-analyses in economics.
Recen ly, he e has been a enaissance in documen ing he e ec s o publica ion selec-
ion bias on epo ed economics esea ch and in he de elopmen o new ools o iden i y
and co ec hese biases. F anco e al. (2014) iden i y a se e e unde - ep esen a ion o
null indings, and simula ions show how he me a-analysis o many smalle s udies can
educe hese biases (Hi schaue e al. 2022). B odeu , Lé, e al. (2016) documen he
p esence o p-hacking among 50,000 es s epo ed by op economics jou nals. B odeu ,
Cook, e al. (2020) ind ha op jou nals a e no excep ional in his. Mo eo e , hey
s ess ha al e na i e expe imen al designs di e in he magni ude o publica ion bias.
Ye , op jou nals ha e he powe o educe his h ea o he c edibili y o economics
esea ch. Aska o e al. (2023) unco e e idence om 345 economic me a-analyses ha
manda o y da a-sha ing policies a economics jou nals can be e ec i e in educing ex-
agge a ed e ec s and he se e i y o publica ion bias. Qui e ecen ly, B odeu , Ca ell,
e al. (2023) in es iga e speci ic s ages in he publica ion p ocess and ind ha p-hacking
is p esen p io o submission, somewha mi iga ed by edi o s’ desk decisions, bu again
en o ced by e iewe s who p e e s a is ically signi ican esul s. F ankel and Kasy (2022)
de elop a amewo k o discuss he ade-o be ween non-selec i e publica ion o ind-
ings and policy ele ance unde sca ce jou nal capaci y. An expe imen con i ms ha
a p e e ence o s a is ically signi ican esul s is widely held among economics schola s
5
(Chop a e al. 2023). They p omo e p e- esul e iews as a solu ion, which may be
seen as pa o a wide mo emen o mo e anspa ency and ou ine p e egis a ion
spea headed by Ch is ensen and Miguel (2018).
This s udy seeks o con ibu e o his apidly g owing li e a u e by in es iga ing how
he indings om wo dozen me a-analyses o speci ic economic a eas o esea ch compa e
o ecei ed con en ional wisdom.
3 Da a collec ion
To collec he equi ed da a and gene a e ou inal da ase , we ollowed se e al s eps.
Fi s , o iden i y ele an s udies, we sea ched Scopus, Google Schola , and Web o
Sciences (WoS). Second, we employed an expe lis and su eyed MAER-ne membe s
ega ding in luen ial me a-analyses.
The da abase sea ch p oceeded as ollows:
Scopus: We used he sea ch s ing “me a AND analysis OR es ima ” and selec ed
se e al quali ie s: “Economics, Econome ics and Finance,” “A icles in jou nals,” “En-
glish language”, and limi ed he keywo ds o “Me a-analysis” OR “Me a Analysis”. Also,
we employed wo u he eligibili y c i e ia: published in 2000 o la e and s udies ha
ha e 100 ci es o mo e. This yielded 164 s udies.
WoS: We used he que y “SU=Economics AND AK=“me a-analysis” OR AK=“me a”
OR AK=“me a analysis” OR AK=“Me a- analysis” OR AK=“Me a-Analysis” and ap-
plied he c i e ia on publica ion da e and numbe o ci a ions, which esul ed in 57
s udies om ha sou ce.1
Google Schola : We used Ha zing’s Publish o Pe ish, employing he keywo ds “me a-
analysis” and “economics,” and se he sample pe iod be ween 2000 and 2023, selec ing
500 en ies. A e clea ing o “Economic, econome ic and inance published jou nal
1A summa y o he WoS sea ch que y can be ound a
h ps://www.webo science.com/wos/woscc/summa y/5216412a-3195-4339-9976-3860847 306d-
6 0209be/ ele ance/1
6
a icles” and “English language,” 164 en ies emained om ha sea ch. Me a-s udies
wi hou an abs ac and duplica es we e d opped, which yielded a o al o 333 candida e
s udies om he sea ch que ies.
In pa allel, we conduc ed a simple olun a y expe su ey in Feb ua y 2023 o he
membe s o he MAER-ne communi y. We sen he su ey o 150 membe s, asking he
ollowing ques ions:
•Do you hink ha he e ha e been me a-s udies ha ha e o e u ned con en ional
economic wisdom? [YES/NO]
•Which me a-s udies we e mos in luen ial in e ms o o e u ning con en ional
wisdom in economics? I would be especially help ul o us i you could also gi e
some e idence/ easons o you answe .
Wi hin he scheduled ime o wo weeks, we ecei ed 45 answe s (a esponse a e o 30%).
O hem, 29 (65.9%) answe ed YES o he i s ques ion, he emaining 16 answe ed NO.
Rega ding he second ques ion, he expe s sugges ed 27 addi ional candida e s udies.
Thus, in o al ou da ase comp ises 360 me a-s udies co e ing a b oad ange o esea ch
ields ha ha e he po en ial o p o iding esul s possibly challenging con en ional
wisdom in hei espec i e esea ch ield. Based on i le and abs ac sc eening, we
coded hese s udies wi h he ollowing quali ie s:
•= 1 i he s udy is closely ela ed o economics; 0 o he wise.
•= 1 i he s udy opic is o gene al in e es (subjec i ely chosen); 0.5 o unsu e
and 0 o no widely known.
•= 1 i he s udy empi ically syn hesizes p ima y s udies; 0 o he wise.
Addi ionally, we b oadly ca ego ized hem in o “Ag icul u al/Ecological/En i onmen al,”
“Beha io al,” “Heal h,” “Labo ,” “Managemen ,” “Me a_Analy ical,” “Mac o” and
“Policy” o ensu e ha we cap u ed a wide ange o me a-s udies.
Fo he nex s ep, we con inued wi h hose s udies ha quali y in all h ee espec s o
ensu e ha hey en ail an impo an e ec size es ima e ela ed o a con en ional wisdom
7
elec ici y use. We ocus he e on he long- e m elas ici y, which p o ides a mo e comp e-
hensi e measu e o s ee ing e ec s. An ea ly and in luen ial su ey by Dahl and S e ne
(1991), inds an a e age long- e m elas ici y o -0.8 o gasoline demand. In con as ,
he sys ema ic me a-da ase in Labandei a e al. (2017) calcula es a simple a e age o
-0.52. Un o una ely, Labandei a e al. (2017) do no s udy he impac o publica ion
bias o employ any o he co ec ion me hods o a i e a a bes p ac ice es ima e.
We now u n o in e na ional economics, ano he widely co e ed ield in me a-analysis.
Fou ou o he 24 s udies in ou inal selec ion con ibu e o his ield, including he
mos -ci ed me a-analysis in ou selec ion (Disdie and Head 2008). This s udy add esses
he ques ion o how geog aphical dis ance a ec s bila e al ade lows. Thei aim is o
iden i y a “ ypical dis ance e ec ” and ac o s o he e ogenei y. They do so by collec ing
almos 1,500 es ima es om mo e han 100 p ima y s udies. The dis ance e ec is usually
es ima ed as θ, “ he nega i e o he elas ici y o bila e al ade wi h espec o dis ance”
(Disdie and Head 2008, p.39), in a g a i y equa ion. The simple a e age es ima e
om hei s udy, abou 0.9, is on he lowe end o he ange o es ima es acco ding o
con en ional li e a u e su eys. Howe e , i s ill con i ms he ypical “puzzle” in he
li e a u e ha dis ance is much mo e in luen ial on ade lows han would be expec ed
om me e anspo cos s alone. Disdie and Head (2008) also add ess publica ion bias,
using a simple OLS eg ession o θes ima es on hei s anda d e o s. They only ind a
weakly posi i e co ela ion, whe e publica ion bias is ac ually s a is ically insigni ican .
The bias-co ec ed es ima e is he e o e close o he simple a e age, a ound 0.8. Tha
is, he puzzling dis ance e ec s on ade a e sligh ly weakened, hough s ill exis en ,
acco ding o his me a-analysis. Disdie and Head (2008) su mise ha he impac o
publica ion bias migh be weake in his li e a u e since dis ance is o en no he main
a iable o in e es bu a me e con ol a iable in g a i y models. I is also no a di ec
policy conce n as, o example, he e ec s o minimum wages on employmen .
14
The e ec o dis ance on ade is somewha ela ed o agglome a ion e ec s, a cen al
opic in egional and u ban economics. The s udy by Melo e al. (2009) asks whe he he
li e a u e on agglome a ion inds a genuine e ec on p oduc i i y and which ac o s may
explain di e ences in ou comes. They exploi mo e han 700 elas ici y es ima es om 34
p ima y s udies and ind ha he e ec s a y a lo wi h coun y-speci ics and indus ial
co e age, among o he ac o s. The simple a e age o all elas ici y es ima es is a ound
0.06, simila o he cen al inding o he seminal s udy by Ciccone and Hall (1996).
Howe e , Melo e al. (2009) de ec asymme ic publica ion bias in a o o mo e posi i e
e ec s. The bias-co ec ed es ima e alls o a ound 0.04, no o e h owing bu clea ly
educing he p oduc i i y bene i s o dense agglome a ions om a suspec ed s ong e ec
o a mo e mode a e e ec .
Agglome a ion e ec s may be suppo ed o hampe ed by public in as uc u e and
o he public capi al. The e a e also se e al me a-analyses in he ealm o public inances
and iscal policies. The mos ci ed one in ou selec ion, by Bom and Lig ha (2014),
ocuses on he p oduc i i y o public capi al. I collec s 68 s udies wi h almos 600
es ima es o he ou pu elas ici y o public capi al. The simple mean o he elas ici y
acco ding o he me a-analysis is abou 0.19, only abou hal o he la ge es ima es ound
in he seminal a icle by Aschaue (1989). Mo eo e , Bom and Lig ha (2014) de ec
posi i e publica ion bias in he li e a u e, and a i e a a bes publica ion bias co ec ed
es ima e o 0.11. Ne e heless, his is a sizeable and s a is ically signi ican a e age e ec
o public capi al on ou pu , which leads he au ho s o conclude ha public capi al is in
sho supply in OECD coun ies and could be ex ended o he bene i o socie al wel a e.
Public capi al and ins i u ions may be one o he d i ing o ces o economic g ow h
and de elopmen o coun ies. The s udy by Ab eu e al. (2005) p o ides an excellen
example o an ea ly me a-analysis in economics ha co e s a highly ele an opic,
he “legenda y” measu e o β= 2% a e o condi ional con e gence be ween income
le els o poo and ich coun ies. This alue o 2%, which was es ablished o se e al
15
condi ions and samples by, among o he s, Sala-i-Ma in (1996), has been a majo s ylized
ac in he g ow h li e a u e o many yea s. I posed a puzzling case o he baseline
neoclassical g ow h model o Solow, Swan and Ramsey, which would p edic a much
as e ca ching-up o poo coun ies h ough capi al accumula ion. Ab eu e al. (2005)
collec abou 600 es ima es om he empi ical li e a u e on β-con e gence. They show
ha he dispe sion o es ima es is indeed wide, ques ioning he con idence in a single
measu e o 2% as a “na u al cons an .” Mo eo e , Ab eu e al. (2005) ind a sys ema ic
ela ion o unobse ed he e ogenei y in echnology le els ha , i aken in o accoun ,
aises he alue o β. A he same ime, hey de ec s a is ically signi ican publica ion
bias ha in la es he a e age o epo ed es ima es. They conclude wi h a co ec ed
a e age con e gence a e o 2.9%, which is, howe e , subjec o s ong he e ogenei y.
This inding poin s o he necessi y o ake in o accoun coun y-speci ic o egion-
speci ic ci cums ances and ins i u ions ha canno be cap u ed by a uni e sal g ow h
model.
A mo e undamen al pa ame e ha is ela ed o g ow h and de elopmen , as well
as o he opics in mac oeconomics in gene al, is he elas ici y o subs i u ion be ween
capi al and labo (σ) in p oduc ion unc ions, which has been s udied by Geche e al.
(2022). The size o he elas ici y has impo an implica ions o g ow h and business cy-
cle models, he e ec i eness o mone a y and ax policies, o he unc ional dis ibu ion
o incomes. In many mac oeconomic models, he elas ici y is con enien ly assumed o be
equal o σ= 1 ( he Cobb-Douglas case). Mo e lexible CES app oaches end o assume
a alue o 0.5. Geche e al. (2022) collec mo e han 3,000 es ima es om 121 s udies.
They show ha indeed a simple mean o all es ima es is close o he Cobb-Douglas case
wi h ˆσ= 0.9. Howe e , publica ion bias is p e alen in his li e a u e, whe e nega i e
alues a e implausible and an a ac o o la ge posi i e es ima es exis s. Co ec ing o
his bias and ollowing some bes p ac ices om he li e a u e leads o a consensus es i-
ma e o σ= 0.3, s ongly ejec ing he con en ional Cobb-Douglas assump ion. Unde
16
hese condi ions, labo and capi al a e g oss complemen s. Thus, wage ises in ela ion
o he cos s o capi al may no lead o a s ong eplacemen o labo by capi al. Con-
sequen ly, al e na i e explana ions ha e o be ound o he secula decline in he labo
sha e. I he elas ici y o subs i u ion is a below one, he all in he labo sha e canno
easily be explained by capi al deepening in a neoclassical g ow h model, as in Pike y
and Zucman (2014). Di ec ed echnical change o an inc ease in ma ke concen a ion
a e al e na i e explana ions ha do no hinge on high alues o σ.
I capi al-labo subs i u ion is a undamen al pa ame e in mac oeconomics, so is he
elas ici y o subs i u ion be ween skilled and unskilled labo , a key concep no only in
mac oeconomics bu also in educa ion and inequali y economics. The ecen ly published
me a-analysis by Ha ánek, I so a, e al. (2022) conside s his ela ion, which is o en
assumed o be 1.5 in model pa ame e iza ion. This would imply ha skilled and unskilled
wo ke s a e g oss subs i u es, hough no oo s ongly. Fo easons o iden i ica ion, mos
p ima y s udies ac ually es ima e he nega i e in e se o he elas ici y, which, unde
he con en ional assump ions, would amoun o -2/3. As an impo an new ea u e,
Ha ánek, I so a, e al. (2022) ake in o accoun bo h publica ion bias, which would
ypically lead o in la ed es ima es o he in e se elas ici y (i.e., a downwa d biased
elas ici y), and a enua ion bias, which would d aw he in e se elas ici y owa ds ze o
(i.e., an upwa d biased elas ici y). Thei cen al inding is ha publica ion bias umps
a enua ion bias, and ha an unbiased a e age es ima e o he nega i e in e se should
a he be a ound -1/4, i.e., a s ong subs i u ion elas ici y o close o 4. This implies
ha skill-biased echnical change has a s ong e ec on he ela i e demand o skilled
labo and he skill p emium, s onge han was p e iously held.
5 Quan i ying ela i e esea ch e ision by me a-analysis
Many o he a o emen ioned examples, e en hough hey conside e y di e en esea ch
ques ions, seem o sha e a common pa e n: he pa ame e o in e es , a e a ho ough
17
and comp ehensi e collec ion o empi ical e idence, and a e accoun ing o publica ion
selec ion bias as well as in luen ial con ol a iables, is o en smalle in absolu e e ms
han he common wisdom as de i ed om an in luen ial p ima y s udy, a classic li e a u e
e iew, me e con en ions, o when conside ing simply he unweigh ed a e age om he
me a-sample. Such a pa e n has al eady been documen ed in he me a-me a-analyses
o Ioannidis e al. (2017) and Doucouliagos, Paldam, e al. (2018), who show ha e ec
sizes sys ema ically appea in la ed in se e al li e a u es i publica ion bias is p e alen .
We assess his pa e n mo e sys ema ically o ou selec ion o 24 me a-s udies. Table 2
compa es he co ec ed mean om he me a-analysis wi h (i) a na a i e e e ence s udy,
(ii) he answe om an a i icial in elligence (AI), and (iii) he unweigh ed simple mean
om he me a-analysis.
(i) Fo each o he me a-s udies, we sea ched o a con en ional wisdom poin es ima e
om a na a i e e e ence s udy. This seminal pape can be a ecen con en ional
li e a u e su ey o a highly ci ed and well-published p ima y s udy ha se he one
o ollow-up p ima y s udies in he espec i e li e a u e. Impo an ly, he na a i e
s udy needs o p o ide a p e e ed es ima e o he pa ame e o in e es . The e e ence
s udy is ci ed in column (2), and i s quali a i e as well as quan i a i e assessmen a e
gi en in columns (3) and (4) o Table Table 2.
(ii) Al e na i ely, we also asked an AI, speci ically he la ge language model (LLM)
GPT-43, o a bes possible poin es ima e o he pa ame e s o in e es in ou 24 me a-
s udies. The gene ic ques ion o he AI o each o he 24 ields eads as ollows:
Please p o ide an es ima e o he e ec o [ esea ch ques ion o he me a-
analysis] based on all ele an li e a u e up o yea [publica ion da e o he
me a-analysis]. Tha is, he es ima e should e lec he s a e o knowledge
p io o he publica ion o he me a-analysis [ i le o he me a-analysis] on
3GPT-4 has he ad an age ha i is well-es ablished and has access o an up- o-da e da abase. While i
is no open-access, i p o ed mo e powe ul in p o iding a quan i a i e assessmen han open-access
al e na i es like Cha GPT o he Bing LLM.
18
Table 2: Con en ional wisdom and esul s om he 24 selec ed me a-analyses
Seminal S udy Con en ional Wisdom (CW) GPT4 Me a-Finding
Me a-S udy (1) Re e ence (2) Quali a i e (3)
Quan
(4) AI
CW
(5) Sim-
ple Mean
(6) Co -
ec . Mean
(7) Pub
Bias
Ab eu e al.
(2005)
Sala-i-Ma in
(1996)
+: poo coun ies ca ch
up
2.00 2.00 4.30 0.30 yes
Ashen el e e
al. (1999)
Psacha opoulos
(1994)
+: school yea s inc ease
ea nings
0.09 0.09 0.07 0.07 yes
Bandie a e al.
(2021)
Gneezy e al.
(2003)
−: women espond less o
pe o mance pay
-0.28 [n/a] 0.08 0.07 [n/a]
Bom and
Lig ha
(2014)
Aschaue
(1989)
+: public capi al en-
hances p oduc i i y
0.39 0.18 0.19 0.11 yes
Disdie and
Head (2008)
Ande son and
Newell (2003)
+: bila e al ade in-
c eases wi h p oximi y
1.30 0.95 0.91 0.80 no
Doucouliagos
and S anley
(2009)
B own (1999)−: highe minimum wage
educes employmen
-0.08 -0.10 -0.19 0.04 yes
Doucouliagos,
S anley, and
Giles (2012)
OECD (2012)+: la ge bene i s om
imp o ing heal h/sa e y
3.90 6.00 9.50 1.66 yes
Feld and Heck-
emeye (2011)
Bénassy-Qué é
e al. (2005)
−: highe ax a es e-
duce FDI
4.79 2.50 3.35 1.74 yes
Fid muc and
Ko honen
(2006)
A is and
Zhang (1997)
+: synch onous busi-
ness cycles o CEECs and
Eu o A ea
0.60 0.60 0.15 0.16 no
Geche (2015) Ramey (2019)+: ax cu s s ongly in-
c ease GDP
2.50 0.65 0.54 0.61 no
Geche e al.
(2022)
Knoblach and
S öckl (2020)
+: close o uni y (Cobb-
Douglas)
0.75 0.95 0.90 0.30 yes
Ha ánek and
I so a (2011)
Ja o cik
(2004)
+: spillo e s om o eign
a ilia es o local i ms
0.38 0.75 0.88 0.18 yes
Ha ánek
(2015)
Hall (1988)+: highe shi s con-
sump ion o u u e
0.50 0.35 0.50 0.07 yes
Ha ánek,
I so a, e al.
(2022)
Can o e e al.
(2017)
−(in e se): |ε|<1
(skilled and unskilled
labo g oss subs i u es)
-0.67 -0.57 -0.56 -0.27 yes
Imai e al.
(2021)
Augenblick e
al. (2015)
1-β>0: people a e
p esen -biased
0.07 0.20 0.04 0.01 yes
Kaise e al.
(2022)
B uhn e al.
(2016)
+: bene i s o g ea e i-
nancial knowledge
0.23 0.20 0.19 0.13 yes
Koe se e al.
(2008)
Be nd and
Wood (1979)
+/−: C-E complemen s
o subs i u es
0.43 0.50 0.47 0.46 [n/a]
Labandei a e
al. (2017)
Dahl and
S e ne (1991)
−,|ε|<1: gasoline no -
mal inelas ic good, sub-
s an ial long- un ε
-0.80 -0.70 -0.53 -0.53 [n/a]
Longhi e al.
(2005)
Ca d (2001)−: highe labo supply
educes wages
-0.15 -0.15 -0.12 -0.04 no
Melo e al.
(2009)
Ciccone and
Hall (1996)
+: agglome a ion en-
hances p oduc i i y
0.06 0.04 0.06 0.04 yes
Nijkamp and
Poo (2005)
Blanch lowe
and Oswald
(2003)
−: wage cu e
downwa d-sloping
-0.10 -0.10 -0.12 -0.08 yes
Reynaud and
Lanzano a
(2017)
Egan e al.
(2009)
+: ecosys em se ices in-
c ease alua ion o lakes
153 [n/a] 315 153 yes
Rose and S an-
ley (2005)
Rose (2000)+: cu ency unions in-
c ease ade
1.20 1.15 0.86 0.39 yes
Voo en e al.
(2019)
Heckman e al.
(1999)
+: ALMP imp o e labo
ma ke ou comes (long
un)
0.03 0.10 0.02 0.004 yes
No es: The able compa es he indings o he 24 selec ed me a-analyses wi h hose om a e e ence
s udy in he espec i e ield and he con en ional wisdom es ima e om GPT-4.
he same opic. The es ima e should ake in o accoun all a ailable scien i ic
s udies, no jus one p ominen s udy. A he same ime, he es ima e should
igo ously summa ize he con en ional wisdom in he li e a u e in yea [pub-
lica ion da e o he me a-analysis]. Answe like an economis and expe in
his ield. P o ide he bes possible poin es ima e o he e ec oge he wi h
he co esponding 95% con idence in e als.
The answe ega ding he poin es ima e gi en by he AI is documen ed in column
(5). No e ha he AI some imes only p o ides a ange o es ima es, o which we ake
he simple a e age. In wo cases, he AI did no espond wi h a quan i a i e assessmen .
Ne e heless, in mos cases, he AI ga e an in o ma i e and delibe a i e answe , includ-
ing a poin es ima e and a con idence in e al. The ull answe s a e p o ided in he
supplemen a y ma e ial. On a e age, he AI’s poin es ima e is qui e close o he esul s
om ou own selec ion o seminal con en ional s udies (which was done be o ehand on
a di e en lap op).
(iii) Ou hi d compa ison (in column 6) is he simple unweigh ed mean o es ima es
included in he me a-analysis, which is usually gi en in he desc ip i e s a is ics o he
me a-s udy. Such an unweigh ed a e age o a b oad se o p ima y s udies does no
accoun o any co ec ions o publica ion bias o bes p ac ices. One migh expec
his measu e o di e subs an ially om he es ima e o he na a i e e e ence s udy.
While his is pa ly he case o indi idual esea ch ques ions, on a e age, he igu es do
no di e oo much. This migh poin o he pe o ma i e powe o seminal s udies in
se ing an es ablished e e ence alue o he pa ame e o in e es .
The h ee e e ence alues can be compa ed o he co ec ed mean om he me a-
s udy as documen ed in column (7). Usually, his co ec ed mean e e s o an es ima e
om he me a-s udy a e co ec ing o publica ion bias and/o de ining a bes p ac ice
es ima e. Me a-analyses ha e applied a ious app oaches o such co ec ions in he
pas , and only ecen ly, has he ield con e ged o es ablished guidelines and s anda d
20
es p ocedu es (S anley, Doucouliagos, Giles, e al. 2013; Ha ánek, S anley, e al. 2020;
I so a e al. 2023). Thus, he e is no single cohe en way o ex ac ing he co ec ed
mean om he espec i e me a-analysis. P ima ily, we e e ed o a p e e ed es ima e
om he me a-s udy and we documen ou choice in he supplemen a y ma e ial i mo e
han one such candida e es ima e is a ailable in he me a-s udy.
I u ns ou ha qui e o en he co ec ed mean om he me a-analysis is subs an ially
close o ze o (o o he null hypo hesis) han all o he h ee compa ison measu es. Ve y
o en, his lowe alue is d i en by some so o publica ion bias: 17 o he 24 s udies
de ec a s a is ically signi ican publica ion bias (column 8).
In o de o compa e and quan i y his pa e n ac oss s udies, we se up Rela i e
Resea ch Re ision (R3) indices o ou h ee compa ison me ics. The R3is calcula ed
as ollows:
R3j
i=MCMi−CW j
i
CW j
i
(1)
whe e MCMiis he me a co ec ed mean om ield i, and CW j
iis he con en ional
wisdom acco ding o compa ison me ic j( om he na a i e s udy, he AI, o he
simple mean o he me a-s udy). The R3index has he ollowing use ul p ope ies: i
gi es he pe cen age change o he absolu e alue o he con en ional wisdom e ec size
due o he me a-analysis. The pe cen age change is posi i e in cases when MCMiand
CW j
iha e he same sign and MCMiexceeds he CW j
iin absolu e alue (an upwa d
e ision). I is nega i e and be ween 0% and -100%, when MCMiis close o ze o
han CW j
i(a downwa d e ision). I exceeds -100% in cases o a sign e e sal o he
con en ional wisdom.4
4No e ha he R3index can be ans o med in o he esea ch in la ion (RI) index o Ioannidis e al.
(2017), which is de ined as RI =CW
MCM −1and hus co esponds o RI =−R3
1+R3. The R3index is
mo e use ul in ou case as i signals downwa d e isions owa ds ze o and e e sals wi h he same
nega i e sign and mono onously inc easing magni ude, while upwa d e isions ecei e a posi i e sign.
Fo he RI index, upwa d e isions and e e sals would ha e he same sign, which would ende he
a e age o he index ambiguous.
21
Table 3: Rela i e Resea ch Re ision (R3) indices
Me a-S udy R3 Seminal R3 AI R3 Me a
Ab eu e al. (2005) -85% -85% -93%
Ashen el e e al. (1999) -24% -24% -7%
Bandie a e al. (2021) -124% [n/a] -18%
Bom and Lig ha (2014) -73% -39% -44%
Disdie and Head (2008) -38% -16% -12%
Doucouliagos and S anley (2009) -155% -141% -122%
Doucouliagos, S anley, and Giles (2012) -57% -72% -83%
Feld and Heckemeye (2011) -64% -31% -48%
Fid muc and Ko honen (2006) -73% -73% 6%
Geche (2015) -75% -6% 13%
Geche e al. (2022) -60% -68% -67%
Ha ánek and I so a (2011) -53% -76% -80%
Ha ánek (2015) -85% -79% -85%
Ha ánek, I so a, e al. (2022) -60% -53% -51%
Imai e al. (2021) -83% -94% -72%
Kaise e al. (2022) -43% -36% -32%
Koe se e al. (2008) 7% -8% -2%
Labandei a e al. (2017) -34% -25% 0%
Longhi e al. (2005) -72% -72% -64%
Melo e al. (2009) -35% 11% -33%
Nijkamp and Poo (2005) -23% -23% -35%
Reynaud and Lanzano a (2017) 0% [n/a] -51%
Rose and S anley (2005) -68% -67% -55%
Voo en e al. (2019) -87% -96% -80%
Median -62% -60% -50%
Mean -61% -53% -46%
No es: The able p esen s he calcula ions o he h ee ela i e esea ch
e ision (R3) indices o he 24 inal me a-analyses acco ding o he in-
o ma ion in Table 2.
22
The esul s o he R3indices a e gi en in Table 3. Fo 11 o he 24 s udies, he
downwa d e ision is -50% o mo e ex eme, consis en ly among all h ee R3indices. In
17 cases, a leas one o he R3measu es indica es such a s ong downwa d e ision. In
wo ins ances, he co ec ed e ec size e en swi ches sign. While he h ee R3indices
di e o each single case, conside ing hei means and medians shows ha hey a e
as onishingly simila , alling wi hin a close ange om abou -45 o -60%. No e ha his
pa e n does no di e much be ween s udies ha en e ed h ough he expe su ey
and hose om he da abase sea ch. Tha is, he a e age co ec ed mean om a me a-
s udy in ou sample educes he con en ional-wisdom e ec size by abou hal . This
is a con i ma ion o Paldam’s ule o humb ha he a ious incen i es o publica ion
selec ion in la e he a e age es ima e ypically by a ac o o 2 (Ioannidis e al. 2017;
Paldam 2022). I also esona es wi h Came e , D ebe , Holzmeis e , e al. (2018) who
show ha highly-powe ed eplica ion s udies o expe imen s in social sciences epo , on
a e age, only hal o he e ec size o he o iginal s udy.
6 Conclusion
In a su ey o me a-analyses in he spi i o Ioannidis e al. (2017), Doucouliagos and
S anley (2013), Doucouliagos, Paldam, e al. (2018), and Geche (2022), we ha e ound
ha many me a-analyses o e u ned con en ional wisdom in hei speci ic ields by ex-
ploi ing comp ehensi e da ase s o empi ical es ima es and by de ec ing publica ion bias.
On a e age, es ima es sh ink by abou hal in absolu e e ms when compa ing he un-
weigh ed a e age and he mean beyond publica ion bias, con i ming “Paldam’s ule”
(Ioannidis e al. 2017; Paldam 2022). This inding also esona es wi h Came e , D ebe ,
Holzmeis e , e al. (2018), who show ha highly-powe ed eplica ion s udies o expe i-
men s in social sciences epo , on a e age, only hal he e ec size o he o iginal s udy.
Ou analysis lends suppo o he po en ial o me a-analysis o b ing o wa d imp o e-
men s ega ding a mo e obus calib a ion o model-pa ame e s, as well as he economic
23
Hi schaue , No be , S en G üne , and Oli e Mußho (2022). Fundamen als o S a is-
ical In e ence. Cham: Sp inge In e na ional Publishing.
Imai, Taisuke, Tom A. Ru e , and Colin F. Came e (2021). “Me a-Analysis o P esen -
Bias Es ima ion using Con ex Time Budge s”. In: Economic Jou nal 131.636, pp. 1788–
1814.
Ioannidis, John P. A. (2005). “Why mos published esea ch indings a e alse”. In: PLoS
medicine 2.8, e124.
Ioannidis, John P. A., T. D. S anley, and H is os Doucouliagos (2017). “The Powe o
Bias in Economics Resea ch”. In: Economic Jou nal 127.605, pp. 236–265.
I so a, Zuzana, H is os Doucouliagos, Tomáš Ha ánek, and T. D. S anley (2023). “Me a–
analysis o social science esea ch: A p ac i ione ’s guide”. In: Jou nal o Economic
Su eys.
Ja o cik, Bea a Sma zynska (2004). “Does Fo eign Di ec In es men Inc ease he P o-
duc i i y o Domes ic Fi ms? In Sea ch o Spillo e s Th ough Backwa d Linkages”.
In: Ame ican Economic Re iew 94.3, pp. 605–627.
Kaise , Tim, Annama ia Lusa di, Lukas Menkho , and Ca ly U ban (2022). “Financial
Educa ion A ec s Financial Knowledge and Downs eam Beha io s”. In: Jou nal o
Financial Economics 145.2, pp. 255–272.
Knoblach, Michael and Fabian S öckl (2020). “Wha De e mines he Elas ici y o Sub-
s i u ion be ween Capi al and Labo ? A Li e a u e Re iew”. In: Jou nal o Economic
Su eys 34.4, pp. 847–875.
Koe se, Ma k J., Hen i L.F. de G oo , and Raymond J.G.M. Flo ax (2008). “Capi al-
ene gy subs i u ion and shi s in ac o demand: A me a-analysis”. In: Ene gy Eco-
nomics 30.5, pp. 2236–2251.
Labandei a, Xa ie , José M. Labeaga, and Xi al López-O e o (2017). “A me a-analysis
on he p ice elas ici y o ene gy demand”. In: Ene gy Policy 102, pp. 549–568.
30
Laibson, Da id I. (1997). “Golden Eggs and Hype bolic Discoun ing”. In: Qua e ly
Jou nal o Economics 112.2, pp. 443–477.
Leame , Edwa d E. (1983). “Le ’s Take he Con Ou o Econome ics”. In: Ame ican
Economic Re iew 73.1, pp. 31–43.
Longhi, Simone a, Pe e Nijkamp, and Jacques Poo (2005). “A Me a-Analy ic Assess-
men o he E ec o Immig a ion on Wages”. In: Jou nal o Economic Su eys 19.3,
pp. 451–477.
Lo ell, Michael C. (1983). “Da a Mining”. In: Re iew o Economics and S a is ics 65.1,
pp. 1–12.
Melo, Pa icia C., Daniel J. G aham, and Robe B. Noland (2009). “A me a-analysis
o es ima es o u ban agglome a ion economies”. In: Regional Science and U ban Eco-
nomics 39.3, pp. 332–342.
Nijkamp, Pe e and Jacques Poo (2005). “The Las Wo d on he Wage Cu e?” In:
Jou nal o Economic Su eys 19.3, pp. 421–450.
OECD (2012). Mo ali y Risk Valua ion in En i onmen , Heal h and T anspo Policies.
OECD.
Paldam, Ma in (2021). “Me hods Used in Economic Resea ch: An Empi ical S udy o
T ends and Le els”. In: Economics 15.1, pp. 28–42.
– (2022). Resea ch me hods in economics. Tech. ep.
Pike y, Thomas and Gab iel Zucman (2014). “Capi al is Back: Weal h-Income Ra ios in
Rich Coun ies 1700-2010”. In: Qua e ly Jou nal o Economics 129.3, pp. 1255–1310.
Psacha opoulos, Geo ge (1994). “Re u ns o in es men in educa ion: A global upda e”.
In: Wo ld De elopmen 22.9, pp. 1325–1343.
Ramey, Vale ie A. (2019). “Ten Yea s A e he Financial C isis: Wha Ha e We Lea ned
om he Renaissance in Fiscal Resea ch?” In: Jou nal o Economic Pe spec i es 33.2,
pp. 89–114.
31
Reynaud, A naud and Denis Lanzano a (2017). “A Global Me a-Analysis o he Value
o Ecosys em Se ices P o ided by Lakes”. In: Ecological economics : he jou nal o
he In e na ional Socie y o Ecological Economics 137, pp. 184–194.
Robe s, Colin J. and T. D. S anley, eds. (2005). Me a- eg ession analysis: Issues o
publica ion bias in economics. Malden, Mass.: Wiley-Blackwell.
Rose, And ew K. (2000). “One money, one ma ke : he e ec o common cu encies on
ade”. In: Economic Policy 15.30, pp. 8–45.
Rose, And ew K. and T. D. S anley (2005). “A Me a-Analysis o he E ec o Common
Cu encies on In e na ional T ade*”. In: Jou nal o Economic Su eys 19.3, pp. 347–
365.
Rosen hal, Robe (1979). “The ile d awe p oblem and ole ance o null esul s”. In:
Psychological Bulle in 86.3, pp. 638–641.
Sala-i-Ma in, Xa ie X. (1996). “The Classical App oach o Con e gence Analysis”. In:
The Economic Jou nal 106.437, p. 1019.
– (1997). “I Jus Ran Two Million Reg essions”. In: Ame ican Economic Re iew 87.2,
pp. 178–183.
S anley, T. D. (2001). “Whea om Cha : Me a-Analysis as Quan i a i e Li e a u e
Re iew”. In: Jou nal o Economic Pe spec i es 15.3, pp. 131–150.
– (2008). “Me a-Reg ession Me hods o De ec ing and Es ima ing Empi ical E ec s in
he P esence o Publica ion Selec ion”. In: Ox o d Bulle in o Economics and S a is ics
70.1, pp. 103–127.
S anley, T. D. and H is os Doucouliagos (2012). Me a Reg ession Analysis in Economics
and Business. New Yo k: Rou ledge.
– (2014). “Me a- eg ession app oxima ions o educe publica ion selec ion bias”. In: Re-
sea ch Syn hesis Me hods 5.1, pp. 60–78.
32
S anley, T. D., H is os Doucouliagos, Ma ga e Giles, e al. (2013). “Me a-Analysis o
Economics Resea ch Repo ing Guidelines”. In: Jou nal o Economic Su eys 27.2,
pp. 390–394.
S anley, T. D., H is os Doucouliagos, and John P. A. Ioannidis (2017). “Finding he
Powe o Reduce Publica ion Bias”. In: S a is ics in Medicine 36.10, pp. 1580–1598.
S anley, T. D. and S ephen B. Ja ell (1989). “Me a- eg ession Analysis: A Quan i a i e
Me hod o Li e a u e Su eys”. In: Jou nal o Economic Su eys 3.2, pp. 161–170.
– (2005). “Me a- eg ession Analysis: A Quan i a i e Me hod o Li e a u e Su eys”. In:
Jou nal o Economic Su eys 19.3, pp. 299–308.
S e ling, Theodo e D. (1959). “Publica ion Decisions and Thei Possible E ec s on In e -
ences D awn om Tes s o Signi icance - O Vice Ve sa”. In: Jou nal o he Ame ican
S a is ical Associa ion 54.285, p. 30.
Voo en, Mel in, Ca la Haele mans, Wim G oo , and Hen ië e an den Maassen B ink
(2019). “THE EFFECTIVENESS OF ACTIVE LABOR MARKET POLICIES: A
META-ANALYSIS”. In: Jou nal o Economic Su eys 33.1, pp. 125–149.
33
Imp in
Publishe
Mac oeconomic Policy Ins i u e (IMK) o Hans-Böckle -Founda ion, Geo g-Glock-S . 18,
40474 Düsseldo , Con ac : mm@boeckle .de, h ps://www. mm-mac o.ne
FMM Wo king Pape is an i egula online publica ion se ies a ailable a :
h ps://www.boeckle .de/de/ mm-wo king-pape -22457.h m
The iews exp essed in his pape do no necessa ily e lec hose o he IMK o he Hans-Böckle -Founda ion.
ISSN 2512-8655
This publica ion is licensed unde he C ea i e commons license:
A ibu ion 4.0 In e na ional (CC BY).
P o ided ha he au ho 's name is acknowledged, his license pe mi s he edi ing, ep oduc ion and dis ibu ion o he ma e ial in
any o ma o medium o any pu pose, including comme cial use.
The comple e license ex can be ound he e: h ps://c ea i ecommons.o g/licenses/by/4.0/legalcode
The e ms o he C ea i e Commons License apply o o iginal ma e ial only. The e-use o ma e ial om o he sou ces (ma ked wi h
sou ce) such as g aphs, ables, pho os and ex s may equi e u he pe mission om he copy igh holde .
