1
Running head: Causali y as Validi y
A icle ype: Sho communica ion
Causali y as Validi y: Some implica ions o he Social Sciences
Ra ael Mo eno and Ra ael J. Ma ínez
Uni e si y o Se ille
Add ess o co espondence:
Depa men o Expe imen al Psychology.
Calle Camilo J. Cela s/n
41018 Se ille (Spain)
Telephone: +34 954557670
Fax: +34 954551784
e−mail: mo [email protected], mce [email protected]
The inal publica ion is a ailable a Sp inge ia h p://dx.doi.o g/10.1007/s11135-006-9060-4
2
Causali y as Validi y: Some implica ions o he Social Sciences
Abs ac
We analyse he concep o causali y in he social sciences, whose de elopmen is
insu icien and lesse han he me hodology de eloped o i s s udy. The na u e o he causal
p ocess as he p oduc ion o e ec s emains unclea and he ela ionships conside ed o be
mani es a ions o ha p ocess canno be aken o p oo o i s exis ence. Gi en hese
di icul ies, we sugges ha , aside om he inhe i ed in e p e a ions, he p ac ice o he
concep o causali y makes e e ence o co ec ly speci ied ela ionships no con ounded by
o he s; cha ac e is ics iden ical o hose which de ine alidi y. In ha way, causali y is
equi alen o he alidi y o a ela ionship. Beyond me ely e-unde s anding causali y, his
p oposal pe mi s he deduc ion ha he empo al p ecedence o he cause is a necessa y
condi ion only o one ype o causali y, making i possible o conside o he ypes, no
admi ed by he adi ional no ion, in which he cause is consequen o simul aneous o he
a iable o be explained. Examples and cha ac e is ics o hese ypes o causali y a e p esen ed
and conside ed o be use ul o he social sciences.
Key Wo ds: Causali y, Validi y, Tempo al asymme y, Simul aneous Rela ions, Non-
di ec ional Causali y, Causal Models, Social Sciences
3
1. In oduc ion
Causali y is a cen al heme in he me hodology o he social and na u al sciences, which
co esponds o he ele an place i holds in ou cul u e (Thompson, 1993). In ha sense,
expe imen al me hodology holds he s udy o causal ela ions as i s objec i e, and he
possibili ies o he e alua ion o causal models o non-expe imen al da a in he social
sciences has inc eased g ea ly in ecen yea s (e.g. Aickin, 2001; Ben le , 1985; Bollen, 1989;
Bollen & Long, 1993; Dalhbäck, 2001; Hayduk, 1987; James e al., 1982; Jö eskog &
Sö bom, 1988; Long, 1983; Mc Donald, 1985; Mu hén, 1987; Seidel & Eichele , 1990).
This me hodological de elopmen has been accompanied by a b oad consensus abou he
na u e and cha ac e is ics o causali y: a) I is unde s ood as a p ocess o p oduc ion o
ansmission o o ces by which he cause b ings abou he e ec ; b) I is conside ed ha ha
p ocess is mani es ed in speci ic ela ionships and ha i appea s when ce ain a iables, such
as common causes and in e ening a iables, which may in luence he componen s o he
ela ionship unde s udy, a e con olled (W igh , 1934, 1960; Simon, 1954; Blalock, 1962;
1964; Duncan, 1966). Rela ionships which do no occu unde hese condi ions a e
conside ed o be spu ious o indi ec , p oduc s o o he a iables, excep in he case o
omission o ano he a iable – e med a supp essing a iable– wi h simila e ec s, bu o he
opposi e sign, on he a iables o a causal ela ionship (Pedhazu , 1982; Bollen, 1989).
Finally, c) I is unde s ood ha , as mani es a ions o a p ocess o p oduc ion, causal
ela ionships a e empo ally asymme ical o di ec ional om he cause o he e ec .
Howe e , such consensus ce ainly does no ansla e in o a clea concep o he
supposed p ocess o p oduc ion, which is he e o e main ained as one o he equen ly
discussed hemes in social science (Do eian, 2001; No e man, 2004). As Kenny (1979: 4)
a i med, "I is di icul o con ey his no ion o causali y o mally, jus as i is o o mally
4
de ine space o ime". The p inciple di icul ies o each one o he indica ed cha ac e is ics a e
he ollowing, espec i ely: a) Tha link is no explained beyond he gene al decla a ion ha
he cause is he gene a ing o ce o he e ec . Wha exac ly he na u e o a causal o ce is and
wha causal o ces in di e en a eas, such as mechanical physics and social p ocesses, ha e in
common, a e ques ions which a e no answe ed, despi e being basic o he pos ula ion o he
exis ence o ha causal o ce; a o ce said o be "almos i alis ic," gi en i s lack o de ini ion
(Kenny, 1979: 4); b) The supposi ion ha he causal o ce has an exis ence which is made
mani es in speci ic ela ionships has no been alida ed, gi en ha he o ce is no de ined
independen ly o he s udied ela ionships. These ela ionships may no be conside ed o be
he mani es a ion o p oo o he exis ence o a o ce which is i sel exclusi ely de ined by
hose ela ionships. I is simply a ci cula a gumen which canno alida e wha i in ends o,
and ; c) Al hough he empo al asymme y o di ec ion o he causal ela ionships is clea ly
iden i ied in e ms o he p ecedence o a a iable, he empo ally asymme ic ela ionship is
no a cha ac e is ic speci ic o causa ion, o example he succession “cockc ow - sun ise”.
Thus, al hough he causal heme in he social sciences p esen s an impo an
de elopmen o o mal p ocedu es, i has se ious di icul y o de ine and alida e he supposed
exis ence o a causal o ce o p oduc ion. As ha e indica ed Glymou e al. (1989: 58),
"ma hema ical ep esen a ion is he easies pa ; he ha d pa conce ns wha in e ences a e o
be d awn om causal claims", which is due p ecisely o he poo de ini ion o hese claims. I
appea s ha he e exis s a concep ual p oblem ega ding he causal p ocess – a p oblem
compounded by he exis ing p ocedu es o he s udy o ela ionships. To ha e p ocedu es o
he pu pose o s udying some hing which is poo ly de ined is a si ua ion which should be
econside ed in o de o eco e he sense o he p ocedu es and o he causal no ion i sel . I is
a p oblem mos likely linked o he insu icien a en ion paid by me hodology, as has been
5
indica ed by o he s: "A basic in oduc ion o he concep o causal heo ies is ha d o ind,
since mos o he ime he discussion is al eady pu in ma hema ical o m" (Sa is &
S onkho s , 1984: 22). F equen ly, me hodological wo k on causali y a ibu es he
insu iciencies in de ini ion o lapses in he philosophy o science. Tha would show ha he
social sciences ha e neglec ed o unde ake he necessa y concep ual analysis, passing i on o
philosophy. None heless, gi en ha ha a i ude has no been sa is ac o y, i would appea
easonable o app oach he causal ques ion om scien i ic me hodology i sel , analysing i s
speci ic bases and p ac ices.
2. Causali y as alidi y
As a s a ing poin , we conside ha he exis ing me hodological p ocedu es and hei
con inuous applica ion exp ess he concep o causali y ac ually used in scien i ic p ac ice – a
concep which, none heless, appea s o be masked by he usual and inhe i ed in e p e a ion o
causali y as a o ce o p oduc ion. In ou opinion, o adequa ely de ine he concep o causali y
does no insis ha one ques ions he ac ion o he supposed o ce, bu ha one ques ions he
ac s which a e unde in es iga ion, and a e conside ed o be easonable, when esea che s
claim o be s udying causali y. Such as wi h any o he concep , i may be ue o causali y ha
in iden i ying he si ua ions in which a communi y conside s he use o he concep o be
adequa e, we a e demons a ing he meaning which ha communi y gi es i in p ac ice
(Wi gens ein, 1953), coinciding o no wi h he way ha ha communi y desc ibes ha
p ac ice i sel . The e o e, o iden i y he cu en concep o causali y in he social sciences
demands he analysis o he ac s used in he me hodological bibliog aphy as cha ac e is ics o
causali y, despi e he in e p e a ions which ha e been supe imposed o e hem; analysing
wha hey do, mo e han wha i is said ha hey do. Tha should be he necessa y and

6
su icien ma e ial o mee he indica ed objec i e.
In he me hodological bibliog aphy o he social sciences, p incipally cons i u ed o
s uc u al equa ion models (SEM), causali y is e e ed o in e ms o ela ionships (e.g. Bollen
& Long, 1993; Dalhbäck, 2001; Hayduk, 1987; Jö eskog & Sö bom, 1988; Seidel &
Eichele , 1990). A la ge pa o he ela ionships o many a iables xj wi h o he a iables yi
may be indica ed by he addi i e linea model
yi = γi1x1 + γi2x2+...+ γiqxq + ζi (1)
whe e each ij is he expec ed change in yi o a one-uni change in x1 when he es o he
a iables a e main ained cons an , and whe e ζi ep esen s he andom componen o yi o he
e ec s o a iables such as e o s o measu emen o cha ac e is ics which a e unknown o o
li le ele ance and a e he e o e no included in he model as independen a iables.
The men ioned exp ession akes in o accoun ha he a iables o each s udied
ela ionship may be ela ed o o he a iables which hey should no con ound. Thus, when
s udying each ela ionship (yi, xj) in equa ions like (1), i is equi ed ha he a iables
con ained in ζi, beyond me ely ha ing minimal e ec s and cons an a iance, a e no ela ed
among hemsel es no among he a iables xj, meaning ha E(ζi)=0, V(ζi)=σζi2, COV(ζiζi')=0
o all ii', and COV(ζixj)=0, o all i and j. This demands he inclusion o he maximum
numbe o ele an a iables, and demands ha hei con ol is ca ied ou by ope a ions
p e ious o da a collec ion and/o by s a is ical means, in expe imen al o non-expe imen al
esea ch. This is wha S eye e al. (2000) ha e indica ed in e ms o causal unbiasedness and
uncon oundedness. Addi ionally, when ep esen a i e inclusion and con ol o a iables a e
no su icien , assump ions o pa ial es ic ions mus be made, e alua ing he possible
ele ance o each one o hem by successi e eplica ions.
Some conclusions may be ex ac ed om wha has been exp essed. While he
7
adi ional no ion o causali y is unde s ood as a p ocess o p oduc ion o ac s, which is
mani es ed in con olled ela ionships be ween a iables, in p ac ice egula ed by he scien i ic
me hodology o SEM, i occu s ha : 1) The meaning o causali y is exclusi ely e e ed o as
adequa ely speci ied o ep esen ed ela ionships, no con ounded by o he a iables. 2) These
wo cha ac e is ics also de ine he alidi y o a ela ionship, (Ke linge , 1973; Shadish e al,
2002). In consequence, o iden i y a ela ionship as causal is simply, in p ac ice, o iden i y i
as alid. 3) Causali y hen is o med by he ela ionships subjec o s udy and by hose o he s
which mus no con ound he i s , which means ha hese and o he ela ionships a e equally
as necessa y and ha nei he is su icien o es ablish causali y. Aside om his conjunc ion,
i is ue ha "co ela ion does no p o e causa ion" and ha "a lack o co ela ion does no
disp o e causa ion" (Bollen, 1989: 52). Finally, 4) he no ion se s he causal heme in
iden i iable e ms and escues i om in e p e a ions wi h p oblems o de ini ion o
e i ica ion; he adequa ely speci ied and no con ounded ela ionships should be ecognised
as necessa y and su icien ma e ial o cons i u e he meaning o causali y, ins ead o being
unde es ima ed as simple mani es a ions o supposed unde lying p ocesses ne e de ined
independen ly wi h espec o hose ela ionships.
To assume he p e iously enume a ed ideas pe mi s a di e en and pe haps mo e
signi ican way o hinking abou he wo cha ac e is ics o he adi ional no ion: Causal
o ce, mani es ed in ela ionships. Ins ead o saying ha a ela ionship e lec s causal o ce
s udied h ough speci ied and no con ounded ela ionships, i may be said ha his
ela ionship is e med causal. This las poin desc ibes he ci cums ances in which he concep
o causali y is applied in scien i ic p ac ice, di e en om he supposi ion o a o ce based on
da a conside ed o be mani es a ions o ha e y o ce. The causal heme he e o e becomes a
p oblem o he way i is plan ed a he han o no ha ing been sol ed. The concep ion o
8
causali y as he ansmission o a hidden o ce would, by de ini ion, be a supe luous p oblem
which could be abandoned.
3. No necessi y o empo al asymme y in causal ela ionships
The empo al asymme y o di ec ionali y o ela ionships, unde s ood as he
p ecedence o one a iable, is an ine i able cha ac e is ic o he assump ion o causali y as a
p oduc i e o ce. The supposed ansmission o ha o ce be ween a iables may only occu i
he cause p ecedes he e ec . In consequence, simul aneous ela ionships a e usually uled ou ,
e en conside ing ha in ela ionships o such appea ance - such as some om non- ecu si e
models - he in e al be ween cause and e ec is smalle han he obse a ion in e al. Wi hin
his amewo k, empo al asymme y o p ecedence is no ques ioned, as i is assumed o be a
necessa y condi ion o he ela ionship and any o he possibili y would be absu d. Keeping
he p ecedence o he cause cons an , he co a ia ion be ween cause and e ec is s udied, bu
no i ha co a ia ion depends on he cause p eceding he e ec . The same occu s in he case
o ecip ocal causali y, in which he cause p eceden a ime is he e ec a iable a a
pos e io ime +k, which assumes bidi ec ionali y o successi e empo al asymme y.
The basis di e s om he no ion o causali y as alidi y unde s ood in e ms o
unbiasedness and uncon oundedness. This no ion does no include e e ence o empo al
asymme y o causal ela ionships, which may indica e ha his cha ac e is ic is no a
necessa y condi ion o all e ec s. This means ha he cause o necessa y condi ion o a
alid ela ionship mus no necessa ily be p eceden in ime, and may ha e se e al
consequences. One may be ha i empo al asymme y o di ec ion is no a necessa y
condi ion o causal ela ionships, hen he o de o occu ence o de e mined a iables is one
hing, and qui e ano he is i ha p ecedence is necessa y o ha ela ionship. Fo ha
9
eason, he empo al asymme y o a de e mined causal ela ionship mus no be unde s ood as
he simple p ecedence o one a iable, bu as he condi ionali y one ela ionship may ha e
wi h espec o he empo al p ecedence o some o i s a iables. Also, i he empo al
p ecedence o one a iable is no a condi ion o all causal ela ionships, ha means ha he e
may be alid causal ela ionships wi hou his p ecedence o he a iable conside ed o be he
cause, occu ing simul aneous o he e ec o ollowing i . In o he wo ds, in acco dance wi h
he no ion o causali y as alidi y used in me hodological p ac ice, one may speak o h ee
ypes o causal o alid ela ionships in unc ion o he empo al loca ion o he cause o
necessa y condi ion: a) Di ec ional causali y wi h p eceding cause; b) Di ec ional causali y
wi h consequen o pos e io cause; and, c) Non-di ec ional causali y, wi h cause simul aneous
o he e ec . Gi en he no el y o he la e wo ypes, i is con enien o explo e hei
cha ac e is ics and p esen possible examples.
Abundan examples may be ound o di ec ional causal ela ionships wi h consequen
cause in he psychological li e a u e o ope an condi ioning, wi h wide gene aliza ion ac oss
species. In any o i s a ia ions, he phenomenon is desc ibed as he inc emen o hose
esponses o he o ganism which a e ollowed by de e mined e en s o ein o cemen . The
pos e io ein o ce is conside ed he cause o he inc ease o he beha iou , and he key
c i e ia o dis inguish be ween beha iou di ec ed by i s consequences and ha di ec ed by
p e ious s imuli which occu as pa lo ian condi ioning. Ano he example o his di ec ional
causali y wi h consequen cause is ound in he explana ions o in en ional ac ion o
an icipa ion-guided ends o goals, such as ha ound in Ac ion Psychology (e.g. Ha é e al,
1985).
In e e ence o non-di ec ional causal ela ionships, some p ima y examples may be
ex ac ed om he wo k o Kuhn (1977) on causali y in physics. Ha ing objec i es qui e
16
No e.
1. The p oposed ypes may ep esen he ou a is o elian ypes o causali y. The ela ionship o he
p eceden cause appea s o co espond o causali y o he e icien ype, hose o consequen cause o
inal causali y and hose non-di ec ional causal ela ionships o he o mal and ma e ial ypes, in which
a a iable is explained by being alidly ela ed o a ep esen a ion o model o i s s uc u e o o i s
componen elemen s, espec i ely.