You are on page 1of 11

"#$%&'()* +(),)-% ,$ , .

%/01&2 13 "#$%&',/(1)$

-144%)/ 1) 5$61$(/1

uavld SLark
Columbla unlverslLy


lor !"#$"%"&$#'






7#$/&,-/8 1hls essay conLrlbuLes Lo observaLlon Lheory by commenLlng on
LsposlLo's paper, Lconomlc clrcularlLles and second-order observaLlon: Lhe
reallLy of raLlngs." 1he key quesLlon of LhaL paper ls summarlzed as: Pow does
one calculaLe ln Lhe keyneslan Lhlrd degree (aLLempLlng Lo ascerLaln whaL Lhe
average oplnlon conslders as Lhe average oplnlon) under condlLlons of
dlabollcal clrcularlLy (when uncerLalnLy abouL Lhe fuLure ls generaLed by
aLLempLs Lo predlcL Lhe fuLure)? LsposlLo answers LhaL raLlngs provlde a flxed
polnL of reference noL because Lhey are accuraLe buL because Lhey are hlghly
vlslble. 1he second half of Lhe paper ls lLself a second-order observaLlon. lL
uses anoLher vlewpolnL (LhaL of observaLlon Lheory) Lo relnLerpreL my earller
eLhnographlc and neLwork analyLlc research on flnance.




keywords: CbservaLlon Lheory, aLLenLlon neLworks, flnanclal models, reflexlvlLy,
valuaLlon.




Address all correspondance Lo uavld SLark, ueparLmenL of Soclology, Columbla
unlverslLy, 606 W. 122
nd
SLreeL, new ?ork, n? 10027. dcs36[columbla.edu


2
"#$%&'()* +(),)-% ,$ , .%/01&2 13 "#$%&',/(1)$

-144%)/ 1) 5$61$(/1

uavld SLark
Columbla unlverslLy


l. (") #'+ ",-./0. ' %"1 2)-1 ,3 4'1#5$+&." ?ogl 8erra

1hls quoLe from one of my favorlLe ?ankee phllosophers should be Lhe moLLo of
eLhnographers. WaLchlng, belng Lhere, ln slLu, wlLh eyes and ears open, ln an
aLLlLude of curloslLy, noL knowlng ln advance whaL you are looklng for buL prepared
Lo recognlze lL when you flnd lL - Lhls ls sLlll one of Lhe very besL Lechnlques for daLa
collecLlon. 8uL, of course, eLhnographers have no monopoly on observlng oLhers. ln
facL, one of Lhe key premlses of observaLlon Lheory as arLlculaLed by nlklas Luhmann
and developed by Llena LsposlLo (2011, 2012) ls LhaL socleLy ls consLlLuLed by Lhls
process of muLual observaLlon. ln an earller, exLraordlnarlly rlch paper, LsposlLo
proposed Lhe ldea LhaL Lhe real purpose and funcLlon of Lhe markeL. ls Lo provlde
an arena for Lhe muLual observaLlon of observers" (LsposlLo 2012: 10). ln Lhe paper
under dlscusslon here, she elaboraLes Lhls ldea ln order Lo apply observaLlon Lheory
Lo Lhe fleld of flnance and, ln parLlcular, Lo Lhe sLudy of raLlngs.

8eaders who are noL from norLh Amerlca mlghL be under Lhe lmpresslon LhaL Lhe
?ankee phllosopher" of my openlng senLence ls some new Lngland genLlemen,
perhaps akln Lo Lhe poeL 8oberL lrosL. AlLhough some of 8erra's sLaLemenLs do read
poeLlcally, ?ogl was no counLry genLleman. 8uL he was, lndeed, a ?ankee - a new
?ork ?ankee - one of baseball's mosL ouLsLandlng players, Lhe caLcher on a repeaL
champlonshlp Leam LhaL flelded some of Lhe sporL's greaLesL aLhleLes.
1
ln addlLlon Lo
hls baseball LalenLs, ?ogl 8erra was an accompllshed qulpsLer, famous for plLhy
remarks such as lL's llke de[a vu all over agaln," never answer an anonymous
leLLer," and When you come Lo a fork ln Lhe road, Lake lL." 1rulsms, Lrue, buL
someLlmes more paradoxlcal and always, curlously, revelaLory. AlLhough hls
succlncL accounL of lnflaLlon - A nlckel aln'L worLh a dlme any more" - mlghL seem
Lo be hls mosL Lelllng conLrlbuLlon Lo Lhe Lheory of flnance, ln facL, l'll be uslng
several of hls oLher observaLlons as a devlce for resLaLlng and dlscusslng LsposlLo's
excellenL paper.

We begln wlLh LsposlLo's argumenL LhaL noLlons of Lhe economy as havlng an lnslde
and ouLslde are mlsLaken. 1here ls no sLance from whlch one can observe Lhe

1
new ?ork MeLs fans wlll demand full dlsclosure so leL's geL lL all on Lhe Lable: As a kld, l
played behlnd Lhe plaLe ln LlLLle League baseball. lellow caLcher ?ogl 8erra and fellow
Cklahoman Mlckey ManLle were my chlldhood heroes. So, of course, l was a ?ankee fan.
3
economy from ouLslde. 1here are dlfferenL perspecLlves, Lo be sure, buL none gazes
lnLo Lhe economy from a poslLlon ouLslde lL. ?ou can observe a loL [usL by waLchlng,
as my ?ankee frlend says. When you do, you see LhaL observaLlons are a parL of Lhe
world, Lhey are parL of Lhe economy. 1hese lnclude Lhe observaLlons by everyday
acLors as well as Lhose by economlsLs - and even by soclologlsLs. lor Lhese reasons,
LsposlLo argues LhaL Lhe framework of embeddedness - wlLh lLs noLlons of economy
'+6 socleLy - ls noL a useful sLarLlng polnL for economlc soclology. SoclologlsLs are
noL ouLslde Lhe economy, observlng lL from some sLandpolnL ln socleLy. lor
LsposlLo, Lhe economy ls noL embedded ln observaLlons. CbservaLlons are parL and
consLlLuLlve of Lhe economy. l agree. And l have argued elsewhere LhaL lL ls Llme for
economlc soclology Lo move beyond Lhe noLlon of markeLs as embedded ln soclal
relaLlons, of value nesLed ln values, of economy resLlng on some maLLress of culLure
(SLark 2009). lnsLead of Lhese neLs, and nesLs, and 8usslan dolls a beLLer meLaphor
for Lhe economy would be LhaL of Lhe moblus sLrlp-Lhe Lopologlcal form wlLhouL
lnslde and ouLslde.


ll. 75'1 8%'#. $- -" #/"46.6 +","63 &".- 15./. '+39"/.:;

ln arL ll of her essay, LsposlLo adopLs Lhe noLlon of Lhe keyneslan beauLy conLesL."
ln such a conLesL, all Lhe [udges are also, ln an lmporLanL sense, Lhe real conLesLanLs
slnce Lhey are compeLlng Lo see who can anLlclpaLe, ln keynes words, whaL average
oplnlon expecLs Lhe average oplnlon Lo be." ?ogl 8erra ls no !ohn Meynard keynes,
buL he does undersLand Lhe dlfference beLween a value lnvesLor and a charLlsL, Lhe
laLLer llke a fashlonlsLa on Lhe resLauranL or Lhe clubblng clrculL who pays aLLenLlon
Lo Lhe muslc or Lhe food only lnsofar as lL necessary for Lhe real game whlch ls
paylng aLLenLlon Lo Lhe crowd. 1he goal ln such a conLesL: Lo anLlclpaLe Lhe cresL and
sLlll be (flguraLlvely) Lhe flrsL ouL of Lhe door when Lhe place geLs Loo crowded [usL
before Lhe muslc sLops" (Lhe sLock plummeLs, Lhe bubble bursLs).

LsposlLo's use of Lhe keyneslan beauLy conLesL ls more sophlsLlcaLed Lhan a slmple
charLlsm. She wanLs us Lo see conLemporary flnance as almosL enLlrely dlsconnecLed
from maLerlallLy. 1he domlnance of derlvaLlves and Lhe prevalence of clrcularlLy
(noLe LhaL keyneslans use specularlLy," uupuy 1989) resulLs ln a slLuaLlon when
you abandon Lhe reference Lo 'ob[ecLlve' crlLerla (Lo Lhe ouLslde world) and
advenLure ln Lhe fleld of oplnlons and soclal sLrucLures." (p. 6). lor LsposlLo, Lhe
meLaphor of Lhe keyneslan beauLy conLesL lndlcaLes LhaL [Lhe] flnanclal world ls
gulded by preclse and noL random crlLerla, whlch however have noLhlng Lo do wlLh
Lhe acLual quallLy of Lhe goods or wlLh Lhe soundness of companles - or wlLh oLher
alleged 'fundamenLals' of Lhe economy" (p. 6). Cne need noL Lo have adopLed some
flnanclal Lheory equlvalenL of Lhe correspondence Lheory of LruLh Lo noLe LhaL lL ls
hyperbollc Lo lnslsL LhaL Lhe flnanclal world has noLhlng Lo do wlLh Lhe soundness of
companles. l wanL Lo focus lnsLead on Lhe noLlon of second-order observaLlons.

4
AL Lhls polnL Lhe reader musL be remlnded LhaL l am operaLlng wlLh Lhe phllosophlcal
depLh of an Amerlcan baseball player and so mlghL be forglven for belng confused
wlLh all Lhls Lalk abouL Lhe dlfferences beLween Luhmannlan flrsL- and second-order
observaLlons, leadlng Lo keyneslan Lhlrd, fourLh, and flfLh degrees" of clrcularlLy.
LsposlLo labels Lhls secLlon 8eauLy conLesL as second-order observaLlon." 1he
ob[ecL of sLudy of soclologlsLs, we learn, should noL be flrsL-order observaLlons
because [1]hey do noL lead Lhe dynamlcs of operaLlons, whlch focus lnsLead on
second-order observaLlons, Lhe muLual observaLlon of Lhe observaLlon of oLhers, up
Lo very hlgh and seemlngly lnexLrlcable levels of absLracLlon and clrcularlLy" (p. 7).

noL frlvolously: l am confused abouL Lhe concepL of a second-order" observaLlon.
ln Lhe passage [usL quoLed above, a second-order observaLlon ls Lhe muLual
observaLlon of Lhe observaLlon of oLhers." Can l observe your observaLlon? Can you
observe mlne? lf we muLually observe each oLhers' observaLlons, are Lhese Lhen
second-order obseveraLlons"? lf so, how? WhaL would lL mean LhaL l have
observed your observaLlon? Cf course, l cannoL be ln your head, l cannoL observe
your percepLlon. l can, however, observe your sLaLed vlews or lnscrlbed
communlcaLlons abouL your observaLlons.

lmaglne, for example, LhaL you are a securlLles analysL. l can read your reporLs, your
esLlmaLes abouL Lhe earnlngs of a glven securlLy. 8uL Lhls ls no dlfferenL - as an order
of observaLlon, whlch ls Lhe quesLlon aL lssue - from looklng aL a sLock Llcker, or
llsLenlng Lo a CLC on a conference call, or readlng a company reporL. Slnce l cannoL
read your mlnd, my readlng of your lnscrlbed communlcaLlon (your observaLlon") ls
necessarlly a flrsL-order observaLlon. erhaps l am (sLubbornly) mlsundersLandlng,
buL l do noL see how Lhe muLual observaLlon of Lhe observaLlons of oLhers" ls a
second-order observaLlon," when lL seems, lnsLead, a sLralghLforward case of mulLl-
slded flrsL order observaLlons.

WhaL abouL waLchlng you observlng? Could Lhls be a second order observaLlon"?
Cbservlng how you pay aLLenLlon - your posLure, for example, or your degree of
aLLenLlveness - seems Lo me Lrlvlal. ln any case, lL would sLlll be a flrsL-order
observaLlon. erhaps Lhen a second order observaLlon refers Lo your observaLlon
abouL my observaLlon, or mlne abouL yours. 1hls very commenLary could be seen as
a meLa-observaLlon, for example, my observaLlons abouL Lhe observaLlons of
LsposlLo. 8uL meLa or noL, from a flrsL-order versus second-order problemaLlc Lhls ls
no dlfferenL from Lhe securlLles analysL's observaLlons (ln effecL, a reporL abouL
reporLs) whlch we had already esLabllshed ls a flrsL-order observaLlon.

WhaL ls lmporLanL here ls LhaL l do agree wlLh LsposlLo LhaL someLhlng ls golng on,
and especlally so ln Lhe fleld of flnance, LhaL cannoL be capLured wlLh Lhe noLlon of a
flrsL-order observaLlon. lL ls preclsely because Lhls process ls lmporLanL LhaL l Lhlnk
Lhe language of observaLlon" mlghL noL be Lhe mosL approprlaLe analyLlc Lool for
lnvesLlgaLlng Lhe problem. l can observe your observaLlons ln Lhe sense of maklng
3
flrsL order observaLlons of your communlcaLlons abouL your observaLlons. 8uL, as
valuable, (ln many cases, more valuable) Lhan your sLaLed communlcaLlon would be
lf l could have access Lo Lhe lnLerpreLlve schema LhaL gave rlse Lo your
communlcaLed observaLlons. ln oLher words, l would llke access Lo your 9"6.%: ln
some cases you mlghL wlsh Lo reveal your model. 8uL ln Lhe more lnLeresLlng cases,
your model (Lhe schema Lhrough whlch you lnLerpreL Lhe world and make declslons,
lncludlng how you lnLerpreL your observaLlons abouL my communlcaLlons) ls
proprleLary. lf you choose noL Lo reveal lL, lL cannoL be observed. 8uL lL mlghL be
lnferred. We wlll reLurn Lo Lhls quesLlon of Lhe vell of hldden models.

1he quesLlon ls wheLher ?ogl 8erra geLs Lhe lasL word or noL, and lf so, how. erhaps
you can observe a loL [usL by waLchlng. erhaps, as well, Lhere ls much LhaL you
cannoL observe by waLchlng. And perhaps Lhere are some Lhlngs LhaL you can
observe buL noL by waLchlng. 1haL ls, lL could be LhaL, lf l had [usL been observanL
enough, l mlghL have noLlced LhaL ?ogl was Lrylng Lo alerL me all along Lo Lhe
posslblllLy LhaL Lhere could be a dlfference beLween observlng and waLchlng.


lll. 75. <)1)/. '$+=1 45'1 $1 )-.6 1" ,.:;

ln arL lll, LsposlLo ampllfles Lhe concepL of moral hazard" ln a fasclnaLlng way,
golng beyond Lhe noLlon LhaL lnsurance produces new rlsks. lor her, Lhe mosL
lnLeresLlng aspecL of Lhe clrcularlLy of observaLlons perLalns Lo uncerLalnLy abouL Lhe
fuLure. And Lhe mosL Lroubllng apecL ls LhaL models LhaL predlcL Lhe fuLure can and
wlll, by belng used, brlng abouL a dlfferenL world Lhan Lhe one predlcLed. 1hls ls a
dlabollcal clrcularlLy: 1he more a predlcLlon ls followed, Lhe more lL wlll modlfy Lhe
condlLlons on whlch lL was based, and Lhereby change Lhe world. now we can
undersLand [usL how apL ls Lhls sLaLemenL by our ?ankee phllosopher: 1he fuLure
aln'L whaL lL used Lo be." CbservaLlons abouL Lhe fuLure brlng abouL dlfferenL
fuLures.

1hls ls Lhe shorLesL secLlon of Lhe paper under dlscusslon, parLly because LsposlLo
has wrlLLen elsewhere (LsposlLo 2009, 2012) and aL greaLer lengLh abouL Lhls Loplc.
LsposlLo noLes LhaL any model would need Lo make assumpLlons abouL Lhe acLlons
of oLhers. 1hlngs geL really lnLeresLlng, she argues, when models become more
sophlsLlcaLed and begln Lo Lake lnLo accounL LhaL oLhers are noL slmply acLlng buL
are acLlng on Lhe basls of models (whlch Lhemselves Lake lnLo accounL LhaL oLhers
are uslng models, each of whlch ls probablllLy based). As models become more
sophlsLlcaLed, more powerful, and beLLer able Lo Lake lnLo accounL model rlsk, prlces
become more volaLlle and Lhe sysLem as a whole less predlcLable. 1haL ls, Lhe
/.%$',$%$13 of models conLrlbuLes Lo Lhe )+8/.6$#1',$%$13 of Lhe sysLem: "under Lhese
condlLlons, every rellable forecasL ls desLlned Lo falslfy lLself, because Lhe fuLure
reacLs Lo Lhe expecLaLlons lmposed on lL - where every addlLlonal rellable forecasL
conLrlbuLes Lo an lncreased unpredlcLablllLy of Lhe fuLure" (LsposlLo 2009: 370). l
6
am far from conversanL ln maLLers of probablllLy Lheory so you should noL rely on
my summary. 8uL ln Lhe splrlL of a ?ogl 8erra qulp, lL would read: lL's probably
lmprobable LhaL lmprobabllLy wlll lasL.


lv. >?< 15. 4"/%6 4./. 8./<.#1@ $1 4")%6+=1 ,.:;

Pow Lhen does one calculaLe ln Lhe keyneslan Lhlrd degree (aLLempLlng Lo ascerLaln
whaL Lhe average oplnlon conslders as Lhe average oplnlon) under condlLlons of
dlabollcal clrcularlLy (when uncerLalnLy abouL Lhe fuLure ls generaLed by aLLempLs Lo
predlcL Lhe fuLure)? WlLh everyLhlng ln an uncerLaln moLlon, Lo whaL can l LeLher my
algorlLhm? LsposlLo answers LhaL raLlngs provlde such a flxed polnL of reference. 1o
fulflll Lhls funcLlon, raLlngs do noL need Lo be perfecL. ln facL, Lhelr funcLlon as a
polnL of reference, LsposlLo argues, can be deLached from Lhelr predlcLlve funcLlon.
WhaL maLLers ls LhaL Lhey provlde a common sLandard, a shared and vlslble
reference, an oplnlon LhaL unllke Lhe oLhers ls avallable Lo everyone and LhaL
everyone knows Lo be known also Lo Lhe oLhers." (p. 13). ln a slLuaLlon of Lhe
generallzed lnvlslblllLy of oLhers' observaLlons, lL ls noL Lhe correcLness of raLlngs buL
Lhelr hlgh vlslblllLy LhaL glves Lhem value.


v. ?< 3") 6"+=1 A+"4 45./. 3")=/. &"$+&@ 3")=%% 4$+6 )8 -"9.45./. .%-.:;

LsposlLo's paper prompLed me Lo Lhlnk agaln abouL my own research on flnance. l
dldn'L know exacLly where l was golng ln LhaL work, buL l now see LhaL l am endlng
up qulLe close Lo observaLlon Lheory even Lhough l had noL prevlously undersLood
Lhe expllclL connecLlon. lor example, ln a recenL paper, lrom ulssonance Lo
8esonance: CognlLlve lnLerdependence ln CuanLlLaLlve llnance" (8eunza and SLark
2012), uanlel 8eunza and l ask Lhe quesLlon: Pow do Lraders deal wlLh Lhe falllblllLy
of Lhelr models? ln parLlcular, how do Lhey deal wlLh Lhe facL LhaL, ln ldenLlfylng
paLLerns ln Lhe markeLs, Lhese same lnsLrumenLs can also bllnd Lhe Lrader from
seelng some Lhlngs. As lnsLrumenLs of percepLlon - and lndeed, llke Lhe opLlc nerve
lLself whlch allows us Lo see buL musL also produce a bllnd spoL - models LhaL reveal
also conceal.

Pow does Lhe Lrader avold such cognlLlve lock ln? 1he answer ls LhaL Lraders
leverage Lhe facL LhaL oLher Lraders are observlng from a dlfferenL vanLage polnL.
1he Lraders aL Lhe merger arblLrage desk we sLudled could noL observe whaL ls on
Lhelr rlvals' screens. 1haL ls, as a Lrader l cannoL observe your observaLlons dlrecLly,
and l don'L have access Lo your model. WhaL l would llke Lo do ls make reasonable
lnferences abouL your model. 8eunza and l show LhaL, ln Lhe case of merger
arblLrage, Lraders place on Lhelr screen an lmage of Lhe spread ploL" whlch Lhey
sklllfully use as a represenLaLlon of Lhe aggregaLe vlews of Lhelr rlvals.

7
When Lhe spread ploL moves ln a dlrecLlon dlfferenL from one's own esLlmaLes,
Lraders can ask, WhaL am l mlsslng?" and make correcLlons ln Lhelr models. ln
lLself, waLchlng Lhe spread ploL ls a flrsL order observaLlon. 8uL when Lhe spread
moves ln a dlfferenL dlrecLlon Lhan Lhe esLlmaLes derlved from my proprleLary
model, Lhe resulLlng LrlangulaLlon ls a second order observaLlon LhaL allows me Lo
make lnferences abouL 5"4 3") '/. $+1./8/.1$+& 15. 4"/%6 whlch can cause me Lo
reflecL on (Lo Lhlnk agaln abouL) my own model. Such reflexlve modellng" can help
an lndlvldual Lrader Lo avold dlsasLer. 8uL lL should come wlLh a warnlng label: when
Lhe sysLem lacks requlslLe dlverslLy, Lhe cognlLlve lnLerdependence can creaLe
poslLlve feedback LhaL ylelds an arblLrage dlsasLer - such as Lhe $2.8 bllllon ln losses
Lo merger arblLrageurs (lncludlng Lhe Leam we sLudled) ln Lhe CL-Poneywell deal.
When Lhe sysLem lacks dlverslLy of vlewpolnLs, Lhe same pracLlces LhaL do prove
effecLlve ln mlLlgaLlng lndlvldual cognlLlve lock ln can lead Lo a collecLlve lock ln of
enormous proporLlons.

8eunza and l base our argumenL on exLended eLhnographlc observaLlons of merger
arblLrageurs ln Lhe derlvaLlves operaLlon of a ma[or lnLernaLlonal lnvesLmenL bank
on Wall SLreeL. 1hls paper was drawn from observaLlons of one merger arblLrage
desk ln one Lradlng room (ln facL, furLher llmlLlng our accounL Lo whaL Lransplred on
a slngle mornlng). ln a subsequenL paper, MaLLeo raLo and l use a very dlfferenL
meLhod - a sLaLlsLlcal analysls of 10,933,662 palrs of securlLles analysLs' esLlmaLes
on uS publlcly llsLed flrms' earnlng per shares - Lo sLudy Lhe effecLs of soclal
sLrucLures of observaLlon on valuaLlon.

ALLenLlon neLworks: A 1wo-Mode neLwork vlew on valuaLlon" (raLo and SLark
2013) bullds on Lhe observaLlonal Lheory prlnclple LhaL valuaLlon depends on Lhe
conLlngenL vlewpolnL of Lhe observer and on Lhe vlews expressed by Lhe observed.
1he observer's vlewpolnLs and observed vlews are for us embedded ln Lhe evolvlng
Lwo-mode (agenLs-asseLs) neLwork sLrucLures of aLLenLlon LhaL characLerlze flnanclal
markeLs. Cur argumenL sLarLs wlLh a slmple quesLlon: WhaL does lL mean Lo focus
on a flnanclal asseL?

Cne way Lo Lhlnk abouL Lhls ls as a slngular relaLlonshlp of an acLor Lo Lhe asseL.
AnoLher, qulLe popular way among soclologlsLs, ls Lo Lhlnk abouL an acLor examlnlng
an asseL ln relaLlonshlp Lo an absLracL caLegory. We Lake a dlfferenL vlew: lnsLead of
poslLlng LhaL lL ls Lhe sLrucLure of #%'--$<$#'1$"+ LhaL guldes valuaLlon" (Zuckerman
2004: 411), we argue LhaL lL ls Lhe sLrucLure of '11.+1$"+ LhaL guldes valuaLlon. ln
place of argulng LhaL valuaLlon ls embedded ln soclally consLrucLed caLegorles, we
argue LhaL lL ls shaped by neLworks of aLLenLlon.

We deflne an '11.+1$"+ +.14"/A as an evolvlng neLwork creaLed by mulLlple agenLs
allocaLlng Lhelr aLLenLlon and expresslng Lhelr [udgmenLs across mulLlple slLuaLlons.
valuaLlon, we argue, ls shaped by an acLor's locaLlon (or vlewpolnL) wlLhln such an
aLLenLlon neLwork. 1haL ls, as a flrsL sLep, we propose Lo sLudy Lhe relaLlonshlp
8
beLween paylng aLLenLlon and allocaLlng aLLenLlon. locuslng aLLenLlon and
allocaLlng aLLenLlon are noL so very dlfferenL. 1he ob[ecLs across whlch one allocaLes
aLLenLlon are Lhe ground agalnsL whlch Lhe flgure can be seen. lf we as researchers
can know Lhe oLher ob[ecLs LhaL an acLor has ln her fleld of vlew, Lhen we know Lhe
vlewpolnL from whlch she makes an assessmenL.

ln assesslng a focal slLuaLlon, acLors can make assoclaLlons, analogles, and
comparlsons wlLh Lhe oLher slLuaLlons LhaL are presenL ln Lhelr 8"/1<"%$" "< '11.+1$"+.
Speclflcally, a feaLure vlewed as sallenL for evaluaLlng one lssue mlghL be recognlzed
as relevanL for anoLher. 1haL ls, Lhe lssues across whlch an acLor allocaLes her
aLLenLlon wlll shape Lhe properLles LhaL are selecLed as sallenL and worLhy of
conslderaLlon when assesslng Lhe focal slLuaLlon.

We refer Lo Lhls as Lhe 0$.48"$+1- .<<.#1. Cur flrsL proposlLlon ls LhaL valuaLlon ls
perspecLlval: Cne's assessmenL of an lssue ls shaped by one's vlewpolnL, glven by
one's conLlngenL porLfollo of aLLenLlon. We hypoLheslze, speclflcally, LhaL Lwo acLors
who assess a glven slLuaLlon vls-a-vls a slmllarly (dlfferenLly) composed porLfollo of
oLher slLuaLlons are more llkely Lo auLonomously converge (dlverge) ln Lhelr
lnLerpreLaLlons of Lhe glven slLuaLlon.

vlewpolnLs are Lhe flrsL buL noL Lhe only sLep ln developlng an observaLlonal
neLwork approach Lo valuaLlon. 8ulldlng on Lhe second relaLlonal properLy of
aLLenLlon ln a Lwo-mode observaLlonal neLwork (l.e., llnks among Lhe compeLlLors
who pay aLLenLlon Lo Lhe same markeL lssues), we expecL LhaL markeL acLors are
more llkely Lo come across Lhe assessmenLs of Lhe compeLlLors who focus Lhelr
aLLenLlon on Lhe same lssues. When Lwo compeLlLors allocaLe Lhelr aLLenLlon across
more slmllar porLfollos of problems, Lhelr vlews become promlnenLly vlslble Lo each
oLher. AssoclaLlons made by one acLor become noLlceable Lo Lhe oLher and vlce-
versa. Conversely, muLual exposure would be llmlLed when Lwo compeLlLors are noL
ln Lhelr respecLlve flelds of vlslon because Lhey are allocaLlng Lhelr aLLenLlon Lo
dlfferenL markeL aspecLs.

1hus, our second proposlLlon, referrlng Lo Lhe 0$.4- .<<.#1, ls LhaL valuaLlon ls doubly
perspecLlval: acLors' valuaLlons are noL only shaped by Lhelr conLlngenL vlewpolnLs,
glven by Lhelr fleeLlng porLfollos of aLLenLlon, buL also by Lhe vlews of oLhers, whlch
Lhemselves are shaped by Lhelr changlng vlewpolnLs. We, Lherefore, furLher
hypoLheslze LhaL, Lhe more (less) Lwo acLors have encounLered Lhe same Lhlrd
acLors' vlews on Lhe oLher slLuaLlons Lo whlch Lhey have noL been aLLenLlve [olnLly,
Lhe more Lhelr lnLerpreLaLlons of a glven slLuaLlon wlll converge (dlverge).

We LesL Lhese proposlLlons ln Lhe conLexL of securlLles analysLs, whom we mlghL
Lhlnk abouL as professlonal observers. ln parLlcular, we sLudy Lhe end of year
earnlngs esLlmaLes LhaL securlLles analysLs make abouL Lhe flrms ln Lhelr porLfollo of
9
coverage. Cur flndlngs supporL Lhe ldea LhaL an acLor's poslLlon ln an observaLlonal
neLwork - vla vlewpolnL and selecLlve exposure Lo oLhers' vlews - shapes valuaLlon.

Cur analysls shows, ln Lhe flrsL lnsLance, LhaL an analysL's esLlmaLe of Lhe end-of-
year earnlngs per share of a glven securlLy ls shaped by Lhe oLher securlLles ln her
fleld of vlew. ln Lerms of odolny and Plll-opper's (2004:91) lnslghL LhaL valuaLlon
Lakes place from Lhe parLlcular orlenLaLlon of an lndlvldual Lo an ob[ecL of
exchange," we found LhaL, when evaluaLlng a glven securlLy, an analysL ls noL faclng
LhaL securlLy alone. ln place of a slngular relaLlonshlp - a glven analysL Lo a glven
securlLy - we found a more mulLl-slded seL of relaLlons. 1he securlLy ls noL alone. As
our flndlngs lndlcaLe, lL ls evaluaLed ln Lerms of Lhe oLher securlLles LhaL are ln Lhe
analysL's fleld of vlew.

Cur analysls furLher demonsLraLes LhaL analysLs' esLlmaLes are lnfluenced by Lhe
vlews of oLher analysLs wlLh whom Lhey shared sLock coverage and LhaL Lhese
effecLs are ampllfled when lndlvlduals shared aLLenLlon paLLerns wlLh Lhe same Lhlrd
parLles. Pow does a glven analysL search when she knows LhaL she has llmlLed
cognlLlve ablllLles? Cur answer began wlLh a slmple proposlLlon: 1he analysL ls noL
alone. Agaln, Lhe relaLlonshlp beLween analysL and securlLy ls noL a slngular one -
Lhere are mulLlple analysLs evaluaLlng LhaL securlLy, each of whom ls slmulLaneously
evaluaLlng oLher securlLles. Clven llmlLed lndlvldual cognlLlve ablllLles, analysLs
leverage Lhls mulLl-slded relaLlonshlp. !usL as Lhe vlew of Lhe focal sLock ls noL only
shaped by Lhe lnformaLlon on LhaL securlLy buL also by Lhe oLher securlLles LhaL form
Lhe background, so we argue LhaL Lhe vlew of Lhe focal sLock lL ls noL shaped only by
Lhe vlews of oLhers abouL LhaL securlLy buL also by Lhelr vlews of oLher securlLles
LhaL are noL shared.

lf my vlews are shaped by my perlpheral vlslon and yours are shaped by your
perlpheral vlslon, Lhen Lo Lhe exLenL LhaL we muLually lnfluence each oLher, we can
say LhaL my vlews are shaped, ln parL, by your perlpheral vlslon.

Cur soclologlcal accounL of valuaLlon explolLs Lwo-mode neLworks as a meLhod of
analysls. Cb[ecLs are locaLed wlLhln a neLwork sLrucLure of aLLenLlon glven by Lhe
acLors who observe and evaluaLe Lhem. Meanwhlle, acLors are also locaLed wlLhln a
sLrucLure of aLLenLlon glven by Lhe Lles LhaL connecL Lhem Lhrough Lhe ob[ecLs Lhey
observe and evaluaLe. noLe Lhe pecullar feaLure of Lhls neLwork. 1here are no dlrecL
Lles among Lhe agenLs. 1hey are noL proxlmaLe because of some personal
connecLlon. 1helr locaLlon ln Lhe soclal space of aLLenLlon - Lhelr proxlmlLy Lo or
dlsLance from each oLher - ls a funcLlon of Lles formed Lhrough ob[ecLs. ln mapplng
Lhese neLworks, we charL soclo-cognlLlve neLworks.

Whereas problems llke Lhe keyneslan beauLy conLesL are sLlmulaLlng economlsLs Lo
Lhlnk abouL lnLersub[ecLlvlLy (lullbrook 2001), we Lhlnk abouL our adopLlon of Lwo-
mode neLwork analysls as a meLhod for sLudylng $+1./",2.#1$0$13.
10
vl. Concluslon: ?1 '$+=1 "0./ B1$% $1=- "0./:;

Pelnz von loersLer's We don'L see whaL we don'L see" mlghL well have been a ?ogl
8erra aphorlsm - for boLh men appreclaLed LhaL a good LauLology can be
lnformaLlve. uoes observaLlon Lheory have a bllnd spoL? Cf course, lL musL. 1o Lhe
exLenL LhaL lL provldes a lens Lo see, lL musL also conceal some elemenL or momenL
or lnsLance from observaLlon. Lvery Lheory has a bllnd spoL. ln place of Lhe slngular
l am a Luhmannlan" (or Lhe equally slngular, l am an eLhnographer," or l am a
neLwork analysL"), Lhe correcLlve ls blnocular LheoreLlcal vlslon. 1o Lhe plea, Ch, buL
surely you, [usL llke l, need an ldenLlLy," one can reply LhaL a real ldenLlLy ls one LhaL
ls wlLh Lhe dlscrepancy, aL Lhe dlfference, wlLhln Lhe dlssonance. uanLe expressed lL
sllghLly dlfferenLly ln Lhls passage from Lhe urgaLorlo of 1he ulvlne Comedy: llx
noL Lhy mlnd on one place only."
uanLe's ln[uncLlon ls good advlce Lo address Lhe problem of geLLlng Lrapped ln your
own successes. A soclologlcal double vlslon can help Lo avold such cognlLlve lock-ln.
Cf course, double vlslon ls a klnd of malady, Lhlngs are ouL of focus. 8uL focus" can
be overraLed, especlally lf lL's Lhe slngle-mlnded varleLy.
We so ofLen hear advlce, wheLher lL ls Lo organlzaLlons or, for example, Lo our
sLudenLs: CeL focused!" 8uL, conLlnulng wlLh Lhls vlsual meLaphor, Lhere ls also
someLhlng Lo be sald abouL Lhe lmporLance of perlpheral vlslon. lL's crlLlcal for
aLhleLes. lL's a useful and necessary sklll for movlng very qulckly LogeLher wlLh many
oLher people, golng ln dlfferenL dlrecLlons, as l've been aware when navlgaLlng from
one subway llne Lo anoLher durlng rush hour ln Lhe 1lmes Square subway sLaLlon.
And lL's vlLal for organlzaLlons. ln hlghly uncerLaln seLLlngs, you should noL be
locked-ln looklng ahead (ln Lhe doubly mlsLaken vlew LhaL Lhe fuLure can be <"/.-..+
and LhaL lL musL necssarlly be '5.'6) buL musL also be aLLenLlve Lo Lhe movemenL
LhaL ls happenlng around you. erlpheral vlslon achleves awareness of LhaL
movemenL.
ln sclence, as for organlzaLlons, Lhe blnocular has beneflLs. We are bllnd Lo our bllnd
spoLs, von loersLer Lells us. We cannoL flnd a polnL from purely ouLslde, as lf ln
some klnd of aerlal soclology, ob[ecLlvlLy and/or reflexlvlLy were a funcLlon of
dlsLance. We are always lnslde. 1here ls no moblus sLrlp soclal Lheory. ln place of no
lnslde/no ouLslde we can operaLe ln anoLher Lopologlcal form - Lhe sLrucLural fold -
lnslde more Lhan one communlLy (vedres and SLark 2010, ue vaan, vedres, and
SLark 2012). ln Lhls way, we can sLrlve for reflexlvlLy as a properLy noL of an
lndlvldual buL of a collecLlvlLy. WhaL's beLLer Lhan an observaLlon? A conversaLlon.



11
9%3%&%)-%$

8eunza, uanlel, and SLark, uavld. 2012.lrom dlssonance Lo resonance: cognlLlve
lnLerdependence ln quanLlLaLlve flnance." C#"+"93 '+6 !"#$.13, 41: 383-417.

de vaan, MaLhl[s, 8alazs vedres, and uavld SLark. 2012. Came Changer: 1he
Lopology of creaLlvlLy ln vldeo game developmenL." Columbla unlverslLy
CenLer on CrganlzaLlonal lnnovaLlon, Worklng aper Serles.

uupuy, !ean-lerre. 1989. Common knowledge, common sense." 75."/3 '+6
D.#$-$"+ 27: 37-62.

LsposlLo, Llena. 2009. 1he CerLalnLy of 8lsk ln Lhe MarkeLs of uncerLalnLy." ln
Wolfgang Pafner/Pelnz Zlmmermann, eds., E$+F.+F G/"+F$+=- H81$"+ I/$#$+&
J"6.%-K CL8"-$1$"+ '+6 M88/'$-'%. 8erlln/Peldelberg: Sprlnger, pp. 339-372.

LsposlLo, Llena. 2011. 75. N)1)/. "< N)1)/.-: 75. 7$9. "< J"+.3 $+ N$+'+#$+& '+6
!"#$.13. ChelLenham: Llgar.

LsposlLo, Llena. 2013. 1he sLrucLures of uncerLalnLy. erformaLlvlLy and
unpredlcLablllLy ln economlc operaLlons." C#"+"93 O !"#$.13 42: 102-129.

lullbrook, Ldward. ed. 2001. ?+1./-),2.#1$0$13 $+ .#"+"9$#-K M&.+1- '+6 -1/)#1)/.-.
new ?ork and London: 8ouLledge.

raLo, MaLLeo and uavld SLark. 2013. ALLenLlon neLworks: A Lwo-mode neLwork
vlew on valuaLlon." Columbla unlverslLy CenLer on CrganlzaLlonal
lnnovaLlon, Worklng aper Serles.

SLark, uavld. 2009. 75. !.+-. "< D$--"+'+#.K M##")+1- "< P"/15 $+ C#"+"9$# Q$<..
rlnceLon unlverslLy ress.

vedres, 8alazs and uavld SLark. 2010. SLrucLural lolds: CeneraLlve ulsrupLlon ln
Cverlapplng Croups." M9./$#'+ R")/+'% "< !"#$"%"&3 113(4): 1130-90

Zuckerman, Lzra W. 2004. SLrucLural lncoherence and sLock markeL acLlvlLy."
M9./$#'+ !"#$"%"&$#'% S.0$.4, 69(3): 403-432.

You might also like