URN:NBN: i: s -oa77626
DOI: 10.11143/ ennia.77626
Gene aliza ion, epis emology and conc e e: wha can social
sciences lea n om he common sense o enginee s
SIMONE TULUMELLO
Tulumello, S. (2019) Gene aliza ion, epis emology and conc e e: wha can
social sciences lea n om he common sense o enginee s. Fennia 197(1)
121–131. h ps://doi.o g/10.11143/ ennia.77626
In his essay I deba e c i ically, and somehow play ully, some
assump ions and sho comings o quan i a i e/posi i is social
esea ch, using a dash o common sense ypical o enginee s.
Ci il enginee s, in designing conc e e s uc u es, pa icula ly hose made
up o conc e e, ha e o con inuously conside he e o embedded in he
limi s o a ailable sys ems o calcula ion, ending up adop ing subs an ial
ac o s o sa e y as coun e -measu es. The s udy o esis ance o conc e e
s uc u es is a good me apho o social esea ch; and ye , quan i a i e/
posi i is esea che s, in hei sea ch o “ alsi iable gene aliza ions”, o en
o ge abou he omnip esence o e o , le alone adop he ac o s o
sa e y. In sho , he common sense o enginee s is use ul o cas s some
no -so- equen ly-conside ed doub s o e he capaci y o quan i a i e
me hods and posi i is epis emologies o c ea e gene alizable social
science indings in ace o unce ain y and he complexi y o human
socie ies. By cas ing such doub s, I ad oca e o a mo e elaxed (bu no
less igo ous) app oach o social esea ch and i s complexi y.
Keywo ds: epis emology, quan i a i e esea ch, quali a i e esea ch,
esea ch me hods, pee - e iew, alsi iable gene aliza ions
Simone Tulumello, Uni e sidade de Lisboa, Ins i u o de Ciências Sociais, U ban
T ansi ions Hub, A . P o . A. Be encou 9, 1600-189, Lisbon, Po ugal. E-mail:
[email p o ec ed]
This essay is mo e han he esul o my own ideas and wo k. No only does i s inal o m esul om a li ely
and insigh ul con e sa ion allowed by Fennia’s open e iew p ocess, bu ha e y con e sa ion con inues
wi hin he ex : ha ing Gun am He b, Jouni Häkli and Ossi Ko a aa a gene ously ag eed o publish hei
commen s alongside my essay, I had he oppo uni y o explici ly engage wi h hei insigh s. This allowed me
o s uc u e he essay as ollows: he main ex con ains he main a gumen , which, o he sake o c a ing a
ex I wan ed o be p o oca i e and engaging, is qui e s aigh o wa d and di ec – a imes indeed educ ionis ;
and I used oo no es qui e hea ily o p o ide nuance o he a gumen and engage in con e sa ion wi h he
e iewe s’ commen s. As such, he e a e mo e ways o ead his essay, wi h o wi hou he oo no es, by i sel
o jumping back and o h o he h ee commen s. I am hus e y g a e ul o Gun am, Jouni and Ossi; and o
Edi o -in-Chie Ki si Pauliina Kallio o o e ing he oppo uni y o he open e iew p ocess and he suppo
h oughou he edi o ial p ocess. I am also g a e ul o Ped o Magalhães, who ead an olde e sion o his
essay and wi h whom I ha e discussed issues o epis emology se e al imes. Though hey may no be awa e
o ha , I was inspi ed by discussions wi h a numbe o ellow schola s and iends, including Andy Inch,
And ea Pa oni, Eleono a Tulumello, Ma co Alleg a and Rui Cos a Lopes. This no wi hs anding, his essay’s
many sho coming – and especially he insolence ha may su ace he e and he e – a e my own esponsibili y.
© 2019 by he au ho . This open access a icle is licensed unde
a C ea i e Commons A ibu ion 4.0 In e na ional License.
122 FENNIA 197(1) (2019)Re iews and Essays
P ologue: a quali a i e esea che ’s Re iewe 2
As a esea che ha has been mainly, indeed almos exclusi ely, employing quali a i e me hods and
case s udy esea ch, I ha e los he coun o he imes a pee - e iewe has c i icized my wo k on he
basis o he e ain “ hou shal no gene alize om one case!” When i happens, ha is, almos e e y
ime I submi an a icle, my i s eac ion is I wan o w i e he edi o s an ang y esponse, some hing
on he lines o : “Dea Edi o s, please p o ide pee - e iewe s wi h he ac ual compe ences o assess
case s udy esea ch, ha is, judge whe he , in ligh o he assump ions o case s udy esea ch, my
a icle is capable (o no ) o building ha kind o heo y ha (a i ing e iewe mus know!) case s udy
can indeed p oduce.”1
Bu hen, mos o en, once I ha e ead he eedbacks and edi o ial decision, I would ake a ew days
o le o s eam, be o e si ing down and decide wha o do. E en ually, i I was lucky enough o ge a
Majo Re ision o Re ise and Resubmi ( ha kind o e iews is ne e ollowed by eques s o Mino
Re ision), I would gi e up. I would expand he me hodological sec ion wi h mo e e e ences on “how
o heo ize om case s udy esea ch”, empe he one o he discussion and add a line in he
conclusions ha sounds like: “Al hough he p elimina y indings o his a icle need o be con i med
by u he esea ch wi h wide panels o da a, we can se ou he ollowing conclusions…” Al hough
hey usually help clea ing he pee - e iew and ha e he pape published (as my ellows p eca ious
esea che s ou he e know, his is no a seconda y ma e ), hese changes signal an implici abdica ion
o he (allegedly) supe io ole o gene aliza ion o ha o heo iza ion in knowledge p oduc ion – and,
besides, hey ake space ( he wo d limi …) ha could be mo e p o i ably used o p o ide a be e
desc ip ion o he case o heo e ical discussion.
One o he las imes his happened, I sha ed my us a ion on he Facebook page Re iewe 2 Mus
Be S opped, which is equen ed by schola s om he mos di e en backg ounds, ensuing a li ely
deba e among hose ha we e sympa he ic wi h my us a ion and hose who would insis I canno
gene alize om one single case.2 The e is one simple e idence coming ou om his, and o he , mo e
igo ous, deba es (e.g. Fly bje g 2004, 285–286): While quali a i e social scien is s, and especially
hose wo king wi h case s udies,3 a e o en e en oo conscious o he assump ions behind, and he
sho comings o , hei epis emological app oaches, quan i a i e/posi i is 4 esea che s end o
conside “ eal science” ha which s ems om hei epis emological assump ions only – mos o en,
analyses o e s a is ically signi ican samples o expe imen s said o p oduce “ alsi iable
gene aliza ions”.5 Fo ins ance, mos me hodological wo ks abou case s udy esea ch a e ex emely,
i o en excessi ely, cau ious, as i hei au ho s we e expec ing a any momen a Re iewe 2 o shou
“ hou shal no gene alize om one case!”6 On he con a y, in a icles g ounded on a quan i a i e/
posi i is pa adigm, I ha e ba ely ound explici discussions o he epis emological assump ions and
sho comings o hose me hods: hose assump ions go wi hou saying, so o speak, and hei
sho comings a e easily o go en, le alone deba ed (bu see Pepinski 2016, o an excep ion).7
In his essay, I will deba e, somehow play ully, some o hose assump ions and sho comings, using
a dash o common sense ypical o enginee s – be o e emb acing human geog aphy and social
sciences, I ook a mas e ’s in ci il-enginee ing. The common sense o enginee s, I will sugges , cas s
some no -so- equen ly-conside ed doub s o e he capaci y o quan i a i e me hods and posi i is
epis emologies o c ea e gene alizable social science indings. By cas ing such doub s, I hope o
con ibu e o a mo e elaxed (bu no less igo ous) app oach o social esea ch and i s complexi y.
Some insolen ema ks on quan i a i e esea ch
Quan i a i e, posi i is ic-o ien ed social esea ch wo ks, o pu i blun ly, h ough he c ea ion o
simpli ied models o social phenomena. By simpli ica ion I mean he p ocess by which he esea che
would selec a numbe o a iables hey conside su icien and adequa e o c ea e a ealis ic model,
ha is, a model capable o desc ibing a gi en (social) phenomenon wi h accep able accu acy. Mo e
a iables, and mo e nuanced ela ions among he a iables, mo e eliable he model will be, bu a
he cos o mo e wo k and compu a ion. The pe ec model, he socie y i sel , is made up o an in ini e
numbe o a iables, hence he need o in ini e ime o collec da a and in ini e ime o p ocess
FENNIA 197(1) (2019) 123Simone Tulumello
indings – We ne Heisenbe g’s unce ain y p inciple (published in 1927) made p ecisely his poin
wi h ega d o physics.8 The co e o he “scien i ic” wo k quan i a i e/posi i is esea che s do, apa
om he mechanic wo k o collec ing da a and unning he models, is deciding wha a iables, hence
wha da a, a e o be used and how o c ea e links among hose a iables. In so doing, he esea che s
ob iously in luence he esul s on he g ounds o hei judgmen – un o una ely howe e , in he public
domain his is as ob ious as Edga Allan Poe’s Pu loined Le e . In Fly bje g’s (2006, 235) wo ds:
The elemen o a bi a y subjec i ism will be signi ican in he choice o ca ego ies and a iables o
a quan i a i e o s uc u al in es iga ion, such as a s uc u ed ques ionnai e o be used ac oss a
la ge sample o cases. And he p obabili y is high ha (1) his subjec i ism su i es wi hou being
ho oughly co ec ed du ing he s udy and (2) ha i may a ec he esul s, qui e simply because
he quan i a i e/s uc u al esea che does no ge as close o hose unde s udy as does he case-
s udy esea che and he e o e is less likely o be co ec ed by he s udy objec s “ alking back”.
Pu in o he wo ds, all social scien is s employ judgmen based on hei epis emological and heo e ical
assump ions. While quali a i e esea che s employ i mainly “downs eam”, ha is, when c i ically
in e p e ing hei indings; quan i a i e/posi i is scien is s employ mos o i “ups eam”, ha is, in
he design o he models.9 I is exac ly he placemen o judgmen ups eam he p oduc ion o indings
ha which c ea es he illusion quan i a i e/posi i is science is “objec i e”. This is he e y i s eason
o be skep ical o he alleged supe io capaci y o quan i a i e esea ch o p oduce gene alized heo y,
when compa ed wi h quali a i e esea ch. To mo e a s ep u he , le me now use a dash o enginee ing
common sense.
S udying ci il enginee ing; and a en u ed me apho
To begin wi h, le me b ie ly ou line he way ci il enginee s a e ained in I aly. Du ing he i s couple
o yea s (o i e, my deg ee was an in eg a ed bachelo /mas e ), he aspi ing enginee s would s udy
almos exclusi ely heo e ical classes such as Ma hema ical Analysis, Geome y, Classical Physics. In
hese classes, heo e ical p oblems a e always esol ed h ough igo ous ma hema ical me hods.
Du ing hese yea s, “e o ” means he same as “mis ake”. Du ing he hi d yea , he aspi ing enginee s
would s udy Building Science (Scienza delle Cos uzioni), which s a s as a heo e ical class oo. The
s uden s would lea n he ma hema ics behind he equa ions ha could, heo e ically, sol e any eal
s uc u al p oblem. Bu , one day midway h ough he cou se, he p o esso would say ha ,
un o una ely, he e is no way o sol e hose equa ions wi h ma hema ical me hods: hey a e oo
complex. Some simpli ica ions a e hen in oduced ha allow he equa ions o be sol ed, bu a he
cos o in oducing elemen s o unce ain y o he solu ion. Fo he i s ime, he aspi ing enginee s a e
aced wi h he exis ence o he e o . The ollowing yea , in he cou se Building Techniques (Tecnica
delle Cos uzioni), he aspi ing enginee s would lea n how o calcula e a eal s uc u e, being in oduced
o u he p oblems wi h ma hema ics and p ac ice: on he one hand, ha e en he simpli ied equa ions
hey had lea ned in Building Science canno be sol ed once hey a e applied o complex s uc u es
ypical o eal li e; and, on he o he , ha i is no e en possible o know wi h absolu e p ecision he load
eal ma e ials a e capable o abso bing be o e b eaking. E en ually, he enginee s will be ained o use
he so wa e hey will employ in hei p o ession, so wa e ha uses simpli ied models o s uc u es.
This o ganiza ion o he deg ee may well be delusional o some s uden s; bu I happened o enjoy
i , because i o ces he aspi ing enginee s o ace se e al c ises o hei p e ious knowledge, hus
s imula ing hem o keep sea ching o di e en ways o sol e p oblems. Mo eo e , his p ocess ains
enginee s o p elimina ily assess he esis ance o s uc u es a a i s sigh : hey lea n how o look a
he p elimina y design o a s uc u e and apidly assess whe he i may be easible, how i may be
easible and – wha else could be mo e impo an in a capi alis wo ld? – how much i could cos . This
common sense helps enginee s in se e al ways: while calcula ing a s uc u e, hey will s a unning
en a i e models, which will no be oo a om he inal one; o , hey would usually be capable o
wa ning he a chi ec s p e y ea ly ha he pa icula s uc u e hey ha e in mind will p obably no be
easible (o would cos much mo e han hey a e ying o sell o hei clien s) be o e spending ime in
unning complex models. A he same ime, and c ucially o my a gumen , enginee s a e ained o
ecognize whe e e o s appea , and es ima e hei magni ude be o e calcula ing s uc u es.
124 FENNIA 197(1) (2019)Re iews and Essays
Because o he use o simpli ied models and he esul ing e o s, enginee s would e en ually apply
a ac o o sa e y o hei esul s, meaning ha hey will load he eal s uc u e wi h a ac ion o he
maximum loads he model can esis . The ac o o sa e y depends on bo h he geome y o he
s uc u e (an es ima ion o he calcula ion e o ) and he ma e ial used. Wi h ega d o ma e ials, he
ac o o sa e y is smalle when using s eel (in common s uc u es, less han 10%), which is a
homogeneous and highly p edic able ma e ial. The ac o o sa e y becomes much bigge o conc e e:
In common conc e e s uc u es i can go up o 50% – meaning ha he s uc u e will be designed o
esis 1.5 imes he maximum eal loads.
The ac o o sa e y is pa icula ly high o conc e e because o i s complexi y. Conc e e is a
composi e, made o coa se agg ega e (sand, g a el, c ushed s ones) bonded oge he wi h luid
cemen . The a iables ha de e mine he s eng h o conc e e a e se e al: a he mac o le el, he
a e age size o agg ega e g ains, he quali y o he cemen , and he p opo ion o agg ega e, cemen
and wa e used; a he mic o le el, he ways he a ious g ains and he cemen a e o ganized ( he
dis ance be ween g ains, he espec i e posi ion o edges and su aces).
The poin I wan o make he e is ha he s udy o he esis ance o s uc u es made up o conc e e
wo ks as a good me apho o social esea ch.10 Le ’s say each g ain ep esen s an indi idual human
being, while he cemen is he bond o ela ionships and a ec s among hem, cha ac e ized by
complex mu ual ensions. The whole s uc u e, he socie y, is made up o pilla s and beams (and
in e ac ions he eo ), ha is, in e ac ing sub-di isions o he socie y: amilies, g oups, communi ies,
classes, aces, gende s, you name i . Quali a i e case s udy esea ch can be seen, om his pe spec i e,
as he in-dep h s udy o a small piece o conc e e looking a he mic o- a iables. Quan i a i e social
science can be seen as he s udy o he esis ance o he en i e s uc u e, o o a pa o i (a beam, a
pilla , a node…) h ough mac o-le el a iables (shape, es ima ed esis ance o he conc e e…).
The day I emembe ed o my enginee ing aining
Keeping his in mind, le me ecall he day my us y enginee ing common sense was awaken while I
was eading quan i a i e social indings. Du ing he e y same days I was discussing wha is “ eal
science” on Re iewe 2 Mus Be S opped, I happened o ead an a icle on c ime and secu i y in US
ci ies by Ellen, Lens and O’Regan (2012). The a icle has a e y ambi ious and impo an goal, ha is,
es ing he commonplace ha housing ouche s policies heigh en c ime. In he USA, housing ouche s
ha e been gi en o households p e iously li ing in public housing “p ojec s”, whe e c ime a es end
o be high; and, he commonplace goes, he displacemen and dispe sion o hose households will
displace and dispe se c ime, which will hus inc ease in he neighbo hoods o des ina ion. Indeed,
ich e idence exis s o he co ela ion be ween se lemen o households wi h ouche s and c ime
inc ease.11 Howe e , he causal ela ion has always been gi en o g an ed, bu ne e “scien i ically”
e i ied, in public and e en some academic deba es – a p ocess ha accu a ely de ines a
“commonplace”.12 Ellen and he colleagues hence decided o s udy hose causal ela ions, using a
longi udinal analysis (1996–2008) and eg ession models o e panel da a a he census ac le el
om 10 la ge US ci ies. No only does he a icle cas se ious doub s on he commonplace, bu i inds
some e idence o he e e se causal his o y: he au ho s conclude ha ouche holde s may end o
mo e o neighbo hoods whe e c ime is al eady inc easing – he e, en al p ices may be lowe o ha
e y eason and mo e landlo ds may hen accep o en in he ouche ma ke .
While I was eading he a icle, I could eel my enginee ing common sense aising he eyeb ows. Le
me pu his e y clea , I am no c i icizing he alidi y o he igo o he a icle: i is well w i en,
igo ous in he use o me hodology and con incing in i s line o hough ; and, hough I ha e no speci ic
skills o judge he eg essions, he a icle clea ed pee - e iew in a good jou nal and I hus gi e o
g an ed he quan i a i e wo k was well done. No he quali y o he esea ch, a he he unde lying
epis emological assump ions abou he capaci y o hese me hods o gene alize social science indings
s imula ed my enginee ing common sense – in sho , le me be Re iewe 2, o once.
So, wha is he p oblem? As e e y expe ienced esea che knows, collec ing da a abou human
beings and hei in e ac ions is e e y hing bu a simple ask. In o de o collec pe ec da a and be
capable o picking he a iables ha bes i a model, he esea che s should ha e access o e e y
FENNIA 197(1) (2019) 125Simone Tulumello
possible in o ma ion abou each indi idual in he popula ion “sample”. O cou se, collec ing all
possible da a on e e y indi idual would en ail immense amoun o wo k, including de ailed
e hnog aphies. While ollowing he “objec s” o s udy in hei daily li e, and pa aph asing wha We ne
Heisenbe g’s unce ain y p inciple has augh us, he be e e hnog aphic da a is collec ed, he mo e
he ac ion o he “objec s” o s udy is in luenced.13 Back o ou conc e e, pe ec da a means knowing
wi hou ma gin o e o he size and shape o e e y g ain and i s exac posi ion: o do his, we ha e no
o he chance bu ul ima ely b eak i . Indeed, esea che s accus omed o pa icipan obse a ion and
ac ion esea ch a e well awa e o hei ole in changing he e y p ocesses hey a e obse ing.
This is why quan i a i e/posi i is esea che s, whose aim is s udying social phenomena wi hou
in luencing hem, use abo e all agg ega ed da a collec ed by o he pa ies (s a is ical da a) o da a
p o ided by he “objec s” o s udy hemsel es (su eys) – hey cons uc a simpli ica ion o he
a e age composi ion o he conc e e and look mo e widely a he s uc u e, o maybe one speci ic
pilla o beam. This media ion c ea es e o , because s a is ical da a is always a simpli ica ion o he
eali y and su eys, well, o ins ance he e eally is no way o know he ex en o which he esponden
is being since e.
This is well e iden in he a icle by Ellen and he colleagues (2012). The au ho s use adminis a i e
da a and admi hey ha e aced impo an challenges.14 The au ho s employ se e al sma ac ics o
deal wi h such challenges, using linea in e pola ions among a ailable da a, compa ing models ha
make o make no use o p oblema ic da a, compa ing he whole model wi h a smalle model whe e
ci ies wi h p oblems wi h da a ha e been emo ed. The poin is, he e is no exac way o measu e he
e o ha he agg ega e e ec o such p oblems will p oduce. In ac , “e o ” is ba ely quan i ied in
his kind o s udies – wi h he excep ion o he s a is ical e o o eg essions.
Le me s ess ha hese p oblems wi h da a a e no speci ic o his pa icula a icle: e e y se o
s a is ical o agg ega ed da a su e o some kind o e o o , as he p inciple o unce ain y ells us,
he e is no way o collec pe ec da a abou any gi en phenomenon wi hou in luencing i . I know
Re iewe 2 is eady o shou a me I am gene alizing om one single case, bu he ha sh eali y is he e
is no quan i a i e/posi i is s udy ha does no su e o some p oblem wi h da a – i Re iewe 2 has
some doub s, hey may wan o pick a s a is ically signi ican sample o said a icles and look in o
hem, one a e he o he . Mo eo e , and his is ano he common issue wi h his kind o a icles, Ellen
and he colleagues (2012) do no discuss whe he , and wi h wha accu acy, a e he 10 selec ed ci ies
ep esen a i e o he u ban USA.
So, wha does his mean o he main inding o he esea ch, he alue o he a iable chosen o
es he causal his o y e e se o he commonplace? Acco ding o linea eg essions, he a iable is
s a is ically signi ican in he h ee models used: 0.167 (p < 0.01), 0.157 (p < 0.05) and 0.160 (p < 0.05)
(Ellen e al. 2012, Table 4). This means ha , acco ding o he da a a ailable and hei causal model, a
household wi h ouche will be abou 16% mo e likely o mo e o a census ac whe e c ime inc eased
in he p e ious yea . The enginee in me would ask, wha is he ma gin o e o o he model? Is i 0.05
(meaning ha he likeliness in he eal wo ld would s ill be posi i e, be ween 11 and 21%)? O is i 0.20
(meaning ha he likeliness in he eal wo ld could be be ween 37% o sligh ly nega i e)? Looking a
he size o p oblems wi h da a, my enginee ing common sense sugges s, I am a aid, ha he la e is
mo e likely han he o me .
Agains gene aliza ion?
Again, my goal he e is no alsi ying he obus indings a igo ous esea ch has p oduced in ligh o i s
epis emological assump ions. This is o say, in ligh o such assump ions, aking in conside a ion he isk
ha ouche holde s may end up mo ing o neighbo hoods whe e c ime is al eady inc easing makes
a lo o sense, bo h concep ually and p ac ically. These indings help be e unde s and he e ec s o
ouche s (e.g. econside “ eedom o choice” he o ic) and suppo a mo e in o med policy decision
a he local le el (e.g. in ex emely unequal and seg ega ed ci ies, ouche s no accompanied by
u he policies may cause mo e p oblems han hose hey sol e). As o he “gene aliza ion” o a
socio-spa ial end (“whe e do ouche holde s go?”), howe e , we ha e no way o be su e wha he
ma gin o e o is and, ankly, I do no hink he e is a way o be. Does anyone know o any me hod

126 FENNIA 197(1) (2019)Re iews and Essays
o sys ema ically measu e he e o o quan i a i e/posi i is social esea ch ( he e o wi h da a, he
e o wi h he simpli ica ion made by he model, and he cumula ed e o o bo h)?
Remembe , enginee s use ac o s o sa e y p ecisely o accoun o he impossibili y o de e mine
he e o wi h accu acy. And, o a complex ma e ial like conc e e, i is a e y big ac o , mos o en
oo big. Bu i is w i en down in laws, because we all acknowledge ha p o ec ing houses om
c umbling down is mo e impo an han was ing some cons uc ion ma e ial: legal p o ec ions o
his kind show he capaci y, in he poli ical and public sphe e, o unde s and he limi s o enginee ing
s udies and esea ch. Su p isingly, he poli ical and public discou ses seem o ha e much less
capaci y o disce n he limi s o gene aliza ions made in quan i a i e/posi i is social esea ch, which,
by he way, can p oduce mo e signi ican damages han a house c umbling down.15 The magni ude
o damages o he economic c isis s a ed in 2007, oge he wi h he ole ce ain assump ions and
ce ain models used o “p edic ” how ma ke s wo k had in c ea ing he condi ions o he c isis, is a
case in poin .16
So wha ? Taking esea ch
cum g ano salis
I could ha e s opped he e, bu his is exac ly whe e Fly bje g (2001) would ask he “so wha ” ques ion.
Wha is he poin o my c i ique? O , wha does an enginee ing common sense sugges o social
scien is s? On he one hand, i sugges s quali a i e scien is s o be awa e ha he pa icula piece o
conc e e hey a e s udying may no ep esen he whole s uc u e – and, as we all know, his is p e y
well accep ed among e iewe s, pa icula ly Re iewe 2. On he o he hand, i sugges s quan i a i e/
posi i is scien is s o emembe o always use e y big ac o s o sa e y when in e p e ing hei
indings, inasmuch as human socie ies a e a leas as complex as conc e e – o be hones , much mo e
complex, i any hing because conc e e changes e y slowly in ime.17 Un o una ely, his la e
sugges ion is ba ely hea d a ound, le alone lis ened o.
Now, I am no sugges ing Re iewe 2 o shou loud “ hou shal no gene alize om panel da a
indings!” – while, I am a aid, his is a leas as alid a claim as “ hou shal no gene alize om one
case!” Wha I am ad oca ing is ha i is high ime we accep aking he gene aliza ions o social
phenomena based on panels o ac ually-exis ing da a cum g ano salis, ha is, wi h he same cau ion we
ake heo y p oduced h ough quali a i e esea ch. We need o lea n ha he e is no such hing as
“ eal science” and “ha d da a”, as opposed o “high-quali y jou nalism” (as se e al posi i is scien is s
s ill conside e hnog aphy and quali a i e esea ch) – qui e he opposi e, ha he alleged objec i eness
o some me hods is a sha p way o conceal judgmen wi hin he p ocess. On he con a y, social
esea ch would bene i a lo by he in e naliza ion ha ou wo k is always con ingen o some
assump ions, ha is, he accep ance o he e y i educibili y and di e en alue o indings p oduced
h ough di e en me hodological and epis emological lenses. Is no , a e all, Re iewe 2 basically he
incapaci y o accep such an i educibili y and he p e ense o o ce one’s assump ions upon o he s’
esea ch?18 Is no Re iewe 2 he incapaci y o accep ha gene aliza ion is no he hallma k o social
esea ch, and ha he p oduc ion o heo y – as opposed o laws – is as ele an an endea o ?
In conclusion, and beyond he poli ics o pee - e iew, a e we eady o emb ace he ac ha
con adic ions a e inhe en exp essions o he complexi y o he human (and non-human) wo ld?19
And, well, le me conclude wi h a pe sonal sugges ion. Always discuss you esea ch ideas and
me hods wi h an enginee be o e unning complex models o making leng hy on- ield esea ch: ha
can help sa e a lo o ime!
Epilogue: he shel e we ha e
(Social) science has been adi ionally imagined as a go hic ca hed al, a pe ec cons uc ion ha will
be comple ed and pe ec ed in he momen he keys one will be pu in i s place. O cou se, scien is s
ha e always been awa e ha he p ocess was complex, slow and pain ul; and ha ex e nal e en s
may had o ced econs uc ing a pa o e en econside ing some hing in he ounda ions. Bu he
keys one has always been, and s ill seems o be o many, he ul ima e goal, he ul ima e answe –
“42!”, Douglas Adams would say.20
FENNIA 197(1) (2019) 127Simone Tulumello
Then, his has been pu in o deba e, om he unce ain y p inciple all he way o decons uc ion,
pos -s uc u alism, pos -mode nism and he like. In ime, he building has been sa aged and seems
now o be mo e a messy s uc u e made up o uins, shining glasses, shacks. The e is he neoclassic
glass and s eel skysc ape om which economis s enjoy he eal wo ld ollow di e en pa hs, while
sipping champagne, blaming s a e egula ion, and accusing people and poli icians o no abiding by
he ules o he sys em. The e is he pos -s uc u alis ield o uins o he elen less c i ique, whe e
he “so wha ” ques ion echoes pe ennially. The cons uc ion seems o be now made up o pa s ha
do no in e ac o , wo s , c ea e mu ual s uc u al p oblems.21
Bu , a e all, i is impo an o acknowledge he scien i ic cons uc ion o i s agili y, no solidi y;
o i s con inuous need o main enance, e u bishing and es uc u ing. Fo , p eca ious as i is, i is he
only shel e we ha e.
No es
1 Th oughou he ex , I dis inguish pu pose ully be ween gene aliza ion and heo y as wo a he
di e en goals o social esea ch and knowledge p oduc ion (see especially he concluding sec ion “So
wha ?”) – Jouni Häkli sugges ed I should pay special a en ion o his dis inc ion.
2 Re iewe 2 Mus Be S opped is a quasi-i onic o um o sha ing hange and discussing bad pee -
e iew. The discussion is publicly a ailable he e: www. acebook.com/g oups/ e iewe 2/
pe malink/10153956215715469/.
3 He e and a e wa ds, I use case s udy as he main poin o e e ence when discussing quali a i e
me hodological app oaches o wo easons: i s , because i has ecen ly assumed a cen al ole in
quali a i ely-o ien ed u ban and geog aphic esea ch; and, second, because i is he me hodology I
ha e mos expe ience wi h.
4 Le us no o ge – as Ossi Ko a aa a and Gun am He b co ec ly poin ed ou – ha no all quan i a i e
esea ch seeks, om a posi i is pa adigm, o build global socie al “laws”, hence he use o
“quan i a i e/posi i is ” h oughou he ex .
5 I am awa e ha he dicho omic opposi ion I adop h oughou he essay is a simpli ica ion o he
landscape o social esea ch – and, a imes, uns he isk o building wo “ca ica u es”, as poin ed ou
by Jouni Häkli. Fo one, he dicho omy quali a i e e sus quan i a i e/posi i is me hodologies/
epis emologies is o en uzzy, as he e is a g owing “g ay a ea” (a de ini ion sugges ed by Häkli) made
up o expe imen a ion wi h di e en app oaches, he use o mixed me hods and c oss- e iliza ion
among long-sepa a ed me hodological and epis emological “ on s”. In e ospec , Hanson (2008)
a gues ha he dicho omy is mo e “appa en han eal” h oughou he his o y o social esea ch.
Mo e ecen ly, a g oup o schola s based a he Sciences Po Médialab ( ounded by B uno La ou ) has
de eloped an a gumen abou he capaci y o digi al me hods and big da a analysis o o e come he
quan i a i e/quali a i e di ide, c ea ing a mo e “con inuous” sociology (Ven u ini e al. 2017). And ye ,
he eali y o ac ually-exis ing social esea ch is cha ac e ized by ie ce deba e and con aposi ion
among di e en schools o hough , he consequences o which we all ha e, soone o la e , came o
expe ience, o ins ance when ecei ing a epo by Re iewe 2.
Gun am He b also spo ed an imbalance in he quan i a i e/posi i is e sus quali a i e dicho omy.
Indeed, me hods and epis emologies a e no s aigh o wa dly and di ec ly associa ed (see B yman
1984, o a discussion). And ye , wi h ega d o he a gumen I de elop on he ela ionship be ween
me hodologies/epis emologies and he p oduc ion o social esea ch, I see ha , independen ly om
hei epis emological o ien a ion, quali a i ely-o ien ed schola s mo e o less ag ee on he limi s o
hei own epis emological assump ions – and hence “quali a i e esea ch”, h oughou he ex , means
“quali a i e esea ch in o med by a di e se se o epis emological assump ions”. A any a e, I ag ee
wi h He b whe e he sugges ed I owe he eade sha ing openly my own app oach, because his
in o ms my pe spec i e o e he issues a s ake. I see my pe sonal epis emological endea o as he
sea ch o c i ical heo y, which I unde s and o ha e a wo old meaning: on he one hand, “a u hless
c i icism o e e y hing exis ing, u hless in wo senses: he c i icism mus no be a aid o i s own
conclusions, no o con lic wi h he powe s ha be” (Ma x 1978 [1844], 13; emphasis in he ansla ion
128 FENNIA 197(1) (2019)Re iews and Essays
quo ed); and, on he o he hand, a heo y ha seeks o os e , in o m and suppo ans o ma i e
ac ion (see Ma cuse 2010).
6 E en Yin (1994), au ho o Case S udy Resea ch, now a i s six h e ised edi ion and possibly he mos
used e e ence in his ield, seems o belie e case s udy is a mino esea ch me hod ha , being
incapable o building “social science gene aliza ions”, should mos ly be used as a p elimina y o
explo a o y ool. Ano he example o his app oach is an o he wise ado able a icle on pe suasion in
case s udy esea ch by Suggelkow (2007). Fly bje g (2004, 2006) is, o he bes o my knowledge, he
schola ha has been mos s aigh o wa d in ad oca ing o , and ully exploi ing, he po en iali ies o
case s udy esea ch o heo iza ion.
7 This may no be pa icula ly he case o human geog aphy, Ossi Ko a aa a sugges s. I ag ee, and
wish o specula e ha his may be in pa due o he ela i ely young his o y o human geog aphy
when compa ed o disciplines such sociology, an h opology o poli ical science – all disciplines wi h
longe and ie ce epis emological/me hodological deba es, om which human geog aphy could
lea n om. The main a ge o my e lec ion is wha Ko a aa a sugges ed e ming he “quan i a i e
posi i is pa adigm o discou se” (a de ini ion I ully emb ace). Wha p omp ed me o w i e his essay
is ha i seems o me ha he quan i a i e/posi i is pa adigm is s ill p e y s ong wi hin he whole
body o social sciences – maybe no as an ideology bu indeed as a “p ac ice” o esea ch and esea ch
e alua ion (see Fly bje g 2004, 285–286) – and in pa icula in disciplines like sociology, social
psychology o poli ical science (e.g. Desch 2019).
8 Acco ding o unce ain y, he e is no way o know wi h absolu e accu acy bo h he posi ion and
speed o a pa icle a a gi en momen , because he mo e ca e ul he obse a ion, he bigge he
impac o e he pa icle’s ajec o y. One o he implica ions is ha o ecas ing he u u e ajec o y o
a single pa icle – and by ex ension o any sys em – is, pu e and simply, impossible. I is qui e su p ising
o me how his p inciple, a basic ene o na u al sciences, is basically igno ed in many s ands o
social sciences, whe e i should be a uism. The ul ima e e sion o he e usal o acknowledge
unce ain y in social sciences is epi omized in g and claims, by some ad oca es o big da a, abou he
“end o heo y”: “Scien is s no longe ha e o make educa ed guesses, cons uc hypo heses and
models, and es hem wi h da a-based expe imen s and examples. Ins ead, hey can mine he
comple e se o da a o pa e ns ha e eal e ec s, p oducing scien i ic conclusions wi hou u he
expe imen a ion” (P ensky 2009; see Ki chin 2014, o a c i ical o e iew).
9 Jouni Häkli co ec ly poin ed ou ha subjec i e judgemen plays “ups eam” in quali a i ely-d i en
esea ch as well, as i shapes “concep ual, philosophical and on ological s a ing poin s” (in his wo ds).
I pe ec ly ag ee, and his is why I use o he e m “mainly” in his sen ence. S ill, I ha e wo esponses:
one, ha he concep s o posi ionali y and e lexi i y ha e been de eloped – in quali a i ely-o ien ed
esea ch – exac ly o acknowledge and emb ace he ole o one’s own judgemen (and e en p ejudice);
and, wo, ha my main poin he e is emphasizing he gene alized lack o such an acknowledgemen
in quan i a i e/posi i is esea ch.
10 As Jouni Häkli and Gun am He b commen ed, his is a “delibe a ely mechanis ic” (in Häkli’s wo ds)
and, indeed, educ ionis me apho . Indeed, my goal is no so much using he s udy o conc e e o
e lec on social esea ch la u sensu, bu a he using i o ocus on he di e ences among me hodological
and epis emological app oaches.
11 A ca ea is necessa y om a c i ical c iminological s andpoin (e.g. Su he land & C essey 1978;
Reine 2016). One should always emind ha c ime s a is ics desc ibe “ epo ed c imes”, ha is, hose
c imes ha a e known o he police and he judicia y, which de ini ely do no co espond o he o ali y
o c ime as a social phenomenon. Repo ed c ime is hea ily in luenced by he likeliness ha a
pa icula c ime is epo ed, by epo ing me hodologies and by police p io i ies – ce ain c imes, o
ins ance d ug c imes, a e almos ne e epo ed and a e hus egis e ed only when ac i ely en o ced
by he police, which may p io i ize his o ha ypology o c ime, his o ha loca ion o hei ac i i y.
Mo e han ha , c ime i sel is a socio-poli ical cons uc ion, as many ac i i ies ha cause ha m a e no
legally de ined as c imes – hink o he ac ha “hono killing”, ha is, killing a wi e, is s ill legal in many
coun ies and was legal in many mo e jus a ew decades ago. As such, using c ime s a is ics o
conclude ha “c ime is high in he p ojec s” may be p oblema ic in he i s place. Mo eo e , one could
hypo hesize ha (Ellen and he colleagues (2012) did no conside his possibili y), ha ing policing
FENNIA 197(1) (2019) 129Simone Tulumello
been his o ically pa icula ly agg essi e wi h (poo , mos ly Black o La inx) people li ing in public
housing, he dispe sion o hose households may be ollowed by a “dispe sion o policing” and, ce e is
pa ibus, con ibu e o he dispe sion o ( epo ed) c ime. Tha said, I will none heless conside epo ed
c ime as c ime – like he a icle unde analysis does – o he sake o he a gumen and because
inco po a ing hose e lec ions would jus add s eng h o he poin I will make abou he unce ain y
su ounding indings based on c ime da a.
12 In pa icula , Ellen and he colleagues (2012) we e p omp ed by he conclusions o a jou nalis ic
epo on he case o Memphis published on The A lan ic (Rosin 2008).
13 Fo a b ie p esen a ion o he p inciple, see no e 8 abo e.
14 A lis o some o he simpli ica ions and p oblems hey poin ou (Ellen e al. 2012, 557–558 and
Appendix 1) ollows: c ime da a a e collec ed a he census ac scale o all ci ies bu one, whe e hey
a e a ailable a he “neighbo hood” le el (usually made up o wo o h ee census ac s); ouche da a
a e missing in some ci ies du ing some pe iods o ime (in some ci ies hey a e no a ailable o he
la ge majo i y o yea s in he pe iod o s udy); anomalies on ouche da a a e ound and a ibu ed o
geocoding p oblems (census ack ID is missing, each yea , in 8 o 20% o cases; in abou 2% o ac /
yea s alues de ia e sha ply om p eceden and ollowing yea s); demog aphic da a a e a ailable
only o 1990, 2000 and an a e age o 2005–2009.
15 The mos e iden example o his is he simpli ica ion and spec acula iza ion o esea ch indings by
mass media – epi omized by sen ences like “scien is s say…”, “acco ding o [inse highly- anked
uni e si y]…” While esea che s canno au oma ically be blamed o he simpli ica ions and dis o ions
o science epo ing, his end has ed back in o he way quan i a i e/posi i is social esea ch is
ca ied ou , as esea che s a e inc easingly pushed o p oduce no el, con i ma o y, “g oundb eaking”
indings in o de o publish in op- anking jou nals. One o he e ec s is he g owing conce n wi h
“p-hacking”, he use o a ious s a egies ( om da a mining o unduly in luencing da a collec ion
echniques) o o ce ully ex ac s a is ically signi ican indings om as collec ions o da a – see he
o e iew by Head and colleagues (2015) and he (in) amous case o he e ac ion o se e al a icles
by ood beha io scien is Wansink (Resnick & Belluz 2018).
16 In his wo k on aus e i y poli ics, Bly h (2013, 32 .) ocuses on he ole played by models – based on
neoclassical unde s andings o economics and used in he inancial indus y o measu e he isk o
loss om in es men s – in jus i ying he decisions ha b ough o he 2007 inancial c ash. Acco ding
o he me ics p o ided by such echnologies, sys emic c ises o inancial ma ke s – hose e y c ises
ha ha e ecu en ly happened du ing he las cen u y o so – should basically ne e occu .
17 Gun am He b sugges ed ha he me apho is no ully applicable o wo easons: i s , because
conc e e has no agency o , a he e y leas , i has much less agency han humans ha e; and, second,
because he es ing o a s uc u e is essen ially abou ue o alse (“will he conc e e hold?”), while
social esea ch deals wi h less clea ou comes. Beyond ei e a ing ha his me apho is abo e all
use ul o ocus on di e ences among pa adigms (see no e 10), le me add a couple e lec ions. Fi s ,
le us no o ge ha he e a e bo h posi i is and pos -posi i is pe spec i es ha would concep ualize
agency qui e di e en ly. Fo ins ance, adical s uc u alism would sugges ha he indi idual human
being unde a capi alis sys em is no eally mo e ee o ac han an indi idual g ain wi hin a conc e e
conglome a e; and Ac o Ne wo k Theo y would maybe con es ha he poin is measu ing and
compa ing he agency o conc e e and human beings – and a he ad oca e conside ing agency as he
ne wo k o ela ionships among hem. A any a e, second, i seems o me ha applied quan i a i e/
posi i is esea ch p io i izes seeking “solu ions” o social p oblems – ha is, i o en mimics he ue/
alse app oach o s uc u e es ing. He e, I wish o add ha he co e endea o o c i ical social esea ch
(see no e 5) is p ecisely ha o ques ioning he de ini ion o social p oblems as opposed o seeking
di ec , (allegedly) neu al and echnical, solu ions o hem – see Gus ield’s (1989) discussion o he
ela ionship be ween “poli ical issues” and “social p oblems”, and my ansposi ion o his a gumen o
he ield o u ban secu i y (Tulumello 2017).
18 I belie e i ually anyone has, a leas once in hei li e, ecei ed (and gi en!) a e iew ha sugges ed
ejec ion on he basis o commen s ha basically denied he e y epis emological o on ological
assump ions o he manusc ip unde analysis. The common commen on he impossibili y o
gene alize om one case o en ells p ecisely o he incapaci y, on he side o he e iewe , o