the man(ager) who knew too much · the man(ager) who knew too much snehal banerjee uc san diego...

The Man(ager) Who Knew Too Much

SnehalBanerjeeUCSanDiego

JesseDavisUNCChapelHill

NaveenGondhiINSEAD

March2020

Asymmetric Information and Communication within firms Firmsoperateunderincompleteandasymmetricinformation:

- “ontheground”employeesproduceinformationaboutfundamentalse.g.,productdemand,operationalconstraints,newtechnologies

- decisionsaremade“atthetop”byexecutivesandowners

Communicationofinformationiskeytoperformance

Economicsresearchhasfocusedonhowmisalignedincentives,contractualincompletenessandorganizationalfrictionslimiteffectivecommunication…

…buthasignoredaubiquitousregularityexhibitedbyindividuals,withprivilegedinformation:

“thecurseofknowledge”

The Curse of Knowledge Informedexhibitthe“curseofknowledge”whenforecastingothers’beliefs

Hardtoaccountforthefactthatyoudon’tknowwhatIknow

Examples:– Evaluatingforecastersafteracrisis:“shouldhaveseenitcoming”– Ineffectiveteacherswhobelieve“simple”topicsshouldbe“obvious”

Pervasiveandrobustcognitivebiasinforminghigher-orderbeliefs

Expertsarepoorcommunicators:assumeconclusionsare“trivial”

– Teachers,Politicians,Scientists,Economists

What we do Modelofintra-firmcommunicationtostudyhow“curseofknowledge”affects

(i)communication,(ii)informationproduction,and(iii)controlrights

Firmconsistsofprincipalandmanager(agent)

– Managerexertsefforttoacquireinfoaboutprojectproductivity– Managerwantstoover-investinprojectandexhibitscurseofknowledge– Managercommunicateswithprincipal– Principalchoosesinvestment

Communication:costlycommunicationvscheaptalk

ControlRights:Delegationvs.Communication

What we find (1)Thecurseofknowledgehamperscommunication

Lesseffortforcostlycommunication,lessinformativecheaptalk

(2)Thecurseofknowledgeimprovesinformationproduction

Moreeffortexertedforinformationacquisition

(3)Acursedmanagercanincreasefirmvalue

• Especially,whenincentivesarewell-aligned,andcurseisnottoolarge• Themanagercanproducerelevantinformation(e.g.,R&D,consultants)• Theprincipalmaychoosetodelegatetoacursedmanager,butnottoanunbiasedone

Background and Motivation

Perspective taking and the Curse of Knowledge Perspectivetakingor“puttingyourselfinsomeoneelse’sshoes”

Perceivingasituation/understandingaconceptfromanotherperspective

- Perceptual:Visual,Auditory- Conceptual:Thoughts,feelings,attitudes

Criticalforsocialinteractionsandcommunication

Whenforminghigher-orderbeliefs,peoplefinditdifficultto:

• Imagineothersknowwhatwedon’tknow(canleadto“winner’scurse”)• Imagineothersdon’tknowwhatweknowi.e.,“curseofknowledge”CurseofKnowledge≈“hindsightbias”and“knewitallalongeffect”

Examples

ForthoseofuswhorememberlifebeforeGoogleMaps!

Curse of knowledge is ubiquitous and difficult to correct Evaluatingforecasts/diagnoses/liability• Hastie,Schkade,andPayne(1999):Jurorsdecidingnegligence• Anderson,Jennings,Lowe,andReckers(1997):Judgesevaluatingauditors• Arkes,Wortmann,Saville,&Harkness(1981):Physiciansover-estimatelikelihoodof“known”outcome,eventhoughex-anteunlikely

• Kennedy(1995):Auditorsover-estimateabilitytopredictbankruptcyex-post

Expertsarepoorcommunications• Politicians,Scientists,Economists• Teacherswho“movetoofast”“skipsteps”“toomuchmaterial”“notclearenough”• Over-estimatemessageinformativeness/abilityofaudience• Particularlybadatforecastingperformanceofnovices(e.g.,Hinds,1999)

Debiasing:Warnings,Feedback,Accountabilityhavelimitedimpact

(Some) Related Literature Relativelysmallliteratureineconomicsexploring“curseofknowledge”

Camerer,LoewensteinandWeber(1993):exploreeffectsinamarketsetting

• ArgueCoKcanalleviatethelemon’sproblem,andenhancetradeMadarasz(2011):Considerssettingwherereceiverhascurseofknowledge

• Receiverevaluatesexpertsusingex-postinfo,underestimatesability• Communication:Biasedexpertspeakstoorarelyandtechnically,underestimatesabilityofaudience

Broadlyrelatedtolargeliteratureonstrategiccommunication,disclosure,delegation

The Model

A definition SupposeAdamisbetterinformedthanBethi.e.,info.set𝐼!isfinerthan𝐼"

AdamexhibitsthecurseofknowledgeifhispredictionofBeth’sforecastis

𝐸!"𝐸"[𝑋]& = (𝟏 − 𝝎)𝐸[𝑋|𝐼"] + 𝝎𝐸[𝑋|𝐼!]

i.e.,Adamcannotignoretheinformationheknows(butBethdoesn’t)

Parameter𝜔 ∈ (0,1)characterizesthedegreeofthecurseofknowledge

Note:LawofIteratedExpectations:𝜔 = 0 ⇔𝐸![𝐸"[𝑋]] = 𝐸[𝑋|𝐼"]

Setup 1 – Firm Payoffs Firmvalueis

𝑉(𝑅, 𝑘) = 𝑅𝑘 −12𝑘#

where𝑘representsinvestment(capital)and𝑅reflectsproductivity:

Productivityisgivenby𝑅 = 𝜇 + 𝜃:

– 𝜇ispriorexpectedproductivity(known)and

– 𝜃 ∼ 𝑈 <− $#, $#=isalearnableshocktoproductivity

Giveninformationset𝐼,optimalactionisconditionalexpectationof𝑅i.e.,

𝑘∗ = 𝐸[𝑅|𝐼] = 𝜇 + 𝐸[𝜃|𝐼]

Setup 2 – Preferences and Beliefs (1)Principal(𝑝)wantstochooseinvestmenttomaximizefirmvaluei.e.,

𝑚𝑎𝑥&𝐸B𝑉(𝑅, 𝑘)C𝐼'D ⇒ 𝑘' = 𝜇 + 𝐸[𝜃|𝐼']

(2)Manager(𝑚)hasabias𝒃 ≥ 𝟎towardsover-investmenti.e.,

𝑚𝑎𝑥&𝐸[𝑉(𝑅, 𝑘) + 𝑏𝑘|𝐼(] ⇒ 𝑘( = 𝜇 + 𝐸[𝜃|𝐼(] + 𝑏

andexhibitscurseofknowledgei.e.,if𝐼(isfinerthan𝐼',then

𝐸'B𝐸([𝜃]D = (𝟏 − 𝝎)𝐸B𝜃C𝐼'D + 𝝎𝐸[𝜃|𝐼(]

Setup 3 – Information Managercanpayacost𝑐(𝑝)toproduceasignal𝑥ofprecision𝑝:

𝑥 = M𝜃, withprobability𝑝𝜉, withprobability1 − 𝑝

where𝜉 ∼ 𝑈 <− $#, $#=isindependentof𝜃 “truthornoisesignal”

Note:Weareabstractingawayfromcommonlyknowninformation

Communication:Managersendsmessage𝑑(𝑥),principalinvests𝑘'(𝑑)

- Costlycommunicationwithcommitment- Cheaptalk

Delegation:Managerinvests𝑘((𝑥)

Costly Communication

Themanagercommitsex-antetoanoisymessage𝑦about𝑥withmessage

precision𝜌bypayingacost𝜅(𝜌):

𝑦 = M𝑥, withprobability𝜌𝜂, withprobability1 − 𝜌

where𝜂 ∼ 𝑈 <− $#, $#=isindependentof𝑥

Managerchoosesmessageprecision𝝆andinformationprecision𝒑tomaximize:

𝑚𝑎𝑥),'

𝐸([𝑉(𝑅, 𝑘'(𝑦)) + 𝑏𝑘'(𝑦)] − 𝑐(𝑝) − 𝜅(𝜌)

Cheap talk Managercannotcommittoasignalex-ante

Instead,hesendsacheaptalkmessage𝒅(𝒙)tomaximize:

𝑢((𝑥) ≡ 𝑚𝑎𝑥+𝐸([𝑉(𝑅, 𝑘'(𝑥)) + 𝑏𝑘'(𝑑)|𝑥]

andoptimallychoosesinformationprecision𝒑tomaximize:

𝑚𝑎𝑥'𝐸'[𝑢((𝑥)] − 𝑐(𝑝)

Modelisstylizedfortractability(signalstructure,linear-quadraticvalue,…)

Interpretation:

1.Managerandprincipalstartwithsomecommoninformation(context)

2.Managerproducescostly,incremental,privateinformation𝑥withprecision𝑝

3.Managersendsareport𝑑(𝑥)

• Costlycommunication:𝜌reflectshowwellreport“makesthecase”• “showingthesteps”/makingthecaserequireseffort

• Cheaptalk:credibilityofthereport

CurseofKnowledge:

Managerover-estimateshow“obvious”𝑥is,givencontext/commoninfo

Communication

Costly Communication Optimalmessageprecision𝜌maximizes𝑢e((𝜌) − 𝜅(𝜌),where

𝑢e((𝜌, 𝑝) ≡ 𝐸(B𝑉f𝑅, 𝑘'(-)g + 𝑏𝑘'(𝑝)D@

= 𝑏𝜇 +𝜇#

2 +𝜎#

24𝑝#f1 − (1 − 𝜔)#(1 − 𝜌#)g

• Utilityincreasesinmessageprecision𝜌andinformationprecision𝑝

Moreprecisecommunication/informationincreasesfirmvalue

• MarginalutilityofcommunicationincreasesinsignalprecisionCommunicationandinformationacquisitionarecomplements

Result 1: Under-investment in Communication Result:Cursedmanagerunder-investsincommunicationprecision𝜌

𝜕#𝑢e(𝜕𝜌𝜕𝜔

=𝜕𝑀𝑈(𝜌)𝜕𝜔

< 0

Intuition:

“therightanswerisobviousfromcontext,

soIdon’tneedtoexplainitbetter!”

Consistentwithnarrativeaboutexpertsbeingpoorcommunicators

- Reportsandpresentationsareunclearandfilledwithjargon- Assumethataudiencehas“readthepaper/textbook”

Communication without costs

Curseofknowledgehamperscostlycommunication,butwhatiftalkischeap?

WeshowthereisaCrawford-Sobel,partitionequilibrium:

Apartitionequilibriumwith𝑁partitionsconsistsofcutoffs−𝜎/2 = 𝑠! < 𝑠" <. . . 𝑠# =

𝜎/2,suchthatforall𝑥 ∈ [𝑠$%", 𝑠$],(i)themanagersendsthesamemessage𝑑(𝑛),and

(ii)theprincipaltakesthesameaction𝑘&(𝑛) = 𝜇 + 𝐸[𝜃|𝑥 ∈ [𝑠$%", 𝑠$]]

Result:Thereexistsapositiveinteger

𝑁!"# = ceil#−12+12(1 + 2

𝝈𝒑(𝟏 −𝝎)𝒃 0

suchthatforevery1 ≤ 𝑁 ≤ 𝑁(/0 ,thereexistsapartitionequilibriumwith𝑁

partitions.

Note:When 𝒃𝝈𝒑(𝟏5𝝎)

> 78,thentheonlyequilibriumisuninformative.

Credibility and the effective bias

Credibility/Informativenessdependsoneffectivebias:

𝒃𝝈𝒑(𝟏 − 𝝎)

Theeffectivebias:

• Increasesinover-investmentbias𝒃 strongerincentivestomislead

• Decreasesinprioruncertainty𝝈 communicationismorevaluable

• Decreasesininformationprecision𝒑 cheaptalkhasmorecontent

Result 2: Curse of knowledge ⇒ Less credible cheap talk

Result:Theeffectivebiasincreaseswiththecurseofknowledge𝜔

Intuition:

“therightanswerisobviousfromcontext,soIhaveastrongerincentivetodistortmyreport!”

Curseofknowledgeharmscommunication,evenwhenitrequiresnoeffort!

Increasein𝜔reducesinformativenessthroughchannels:

(i)Decreasesthemaximalnumberofpartitions

𝑁'() = ceil9−12 +

12;1 + 2

𝜎𝑝(1 − 𝜔)𝑏 ?

(ii)Forafixednumberofpartitions𝑁,inducesmorepoolingatthetop

Information Production

Sincecurseofknowledgehamperscommunication,doesitmakeinformation

lessvaluable?

Result 3: Over-investment in information acquisition Withcostlycommunication,expectedutilityis:

𝑢e( ≡ 𝑏𝜇 +𝜇#

2 +𝜎#

24𝑝#f1 − (1 − 𝜔)#(1 − 𝜌#)g

Result:Cursedmanagersover-investininformationproductioni.e.,

𝜕#𝑢e(𝜕𝑝𝜕𝜔

=𝜕𝑀𝑈(𝑝)𝜕𝜔

> 0

Intuition:

“SinceI’mgoodatconveyingtherightanswer,

doingresearchismorevaluable”

Over-estimateabilitytocommunicate⇒strongerincentivetoexerteffort

Result 4: A cursed manager can increase firm value Expectedvalueincreasesinbothinformationprecisionandmessageprecision

𝐸[𝑉(𝑅, 𝑘&)] =𝜇*

2 +𝜎*

24𝑝*𝜌*

• Fixedinformationprecision𝑝:cursedmanagersdecreasevalue• Endogenousinfo.precision𝑝:cursedmanagerscanincreasevalue

when𝜔isnottoohigh

Figure 3: Optimal communication and information acquisition under costly communicationThe figure plots the choice of message precision ⇢ (dashed), acquired information precisionp (solid), and expected value of the firm E [V (R, k)] under costly communication, as afunction of the degree of cursedness. The manager optimally chooses message precision ⇢subject to a cost (⇢) = 0

⇢2

1�⇢ and chooses acquired information precision p subject to a

cost c (p) = c0p2

1�p . The other parameters of the model are set to: µ = 1, b = 0.02, � = 1and c0 = 0.01 and 0 = 0.001.

0.2 0.4 0.6 0.8 1.0

0.1

0.2

0.3

0.4

0.5

0.6

0.2 0.4 0.6 0.8 1.00.500

0.501

0.502

0.503

0.504

(a) ⇢ (dashed) and p (solid) vs ! (b) Expected value vs !

c (p) = c0p2

1�p , as before, we assume that the manager’s cost of communicating with precision

⇢ is give by (⇢) = 0⇢2

1�⇢ . The figure plots the optimal choice for both precisions, ⇢ (dashed)

and p (solid), and the expected firm value as a function of the manager’s curse of knowledge,

!.

In this example, there is very little information acquisition or communication when the

manager is rational (i.e., when ! = 0). As ! increases, the marginal utility from information

acquisition increases and the marginal utility from message precision decreases. When ! is

su�ciently large (when ! ⇡ 0.1 in panel (a)), however, this is o↵set by the complementarity

across precisions and so both p and ⇢ quickly increase with !. As panel (b) illustrates, this

leads to an increase in the firm’s expected value since the manager is now acquiring and

communicating more precise information. Notably, this suggests that small changes in the

manager’s bias can lead to large changes in informational e�ciency and firm value.

As the curse of knowledge increases further, the o↵setting impact of the complementarity

begins to dissipate: eventually (when ! ⇡ 0.2) this leads to (i) higher information acquisition,

and (ii) lower message precision. Note that, even in this region, the expected value of the

firm continues to rise, until the rate at which the message precision decreases outweighs the

informativeness of the manager’s signal (near ! ⇡ 0.3). Eventually, the manager’s curse

of knowledge is su�ciently high to drive the optimal choice of message precision to nearly

zero. From this point forward, the expected value of the firm remains relatively insensitive

21

Figure 3: Optimal communication and information acquisition under costly communicationThe figure plots the choice of message precision ⇢ (dashed), acquired information precisionp (solid), and expected value of the firm E [V (R, k)] under costly communication, as afunction of the degree of cursedness. The manager optimally chooses message precision ⇢subject to a cost (⇢) = 0

⇢2

1�⇢ and chooses acquired information precision p subject to a

cost c (p) = c0p2

1�p . The other parameters of the model are set to: µ = 1, b = 0.02, � = 1and c0 = 0.01 and 0 = 0.001.

(a) ⇢ (dashed) and p (solid) vs ! (b) Expected value vs !

c (p) = c0p2

1�p , as before, we assume that the manager’s cost of communicating with precision

⇢ is give by (⇢) = 0⇢2

1�⇢ . The figure plots the optimal choice for both precisions, ⇢ (dashed)

and p (solid), and the expected firm value as a function of the manager’s curse of knowledge,

!.

In this example, there is very little information acquisition or communication when the

manager is rational (i.e., when ! = 0). As ! increases, the marginal utility from information

acquisition increases and the marginal utility from message precision decreases. When ! is

su�ciently large (when ! ⇡ 0.1 in panel (a)), however, this is o↵set by the complementarity

across precisions and so both p and ⇢ quickly increase with !. As panel (b) illustrates, this

leads to an increase in the firm’s expected value since the manager is now acquiring and

communicating more precise information. Notably, this suggests that small changes in the

manager’s bias can lead to large changes in informational e�ciency and firm value.

As the curse of knowledge increases further, the o↵setting impact of the complementarity

begins to dissipate: eventually (when ! ⇡ 0.2) this leads to (i) higher information acquisition,

and (ii) lower message precision. Note that, even in this region, the expected value of the

firm continues to rise, until the rate at which the message precision decreases outweighs the

informativeness of the manager’s signal (near ! ⇡ 0.3). Eventually, the manager’s curse

of knowledge is su�ciently high to drive the optimal choice of message precision to nearly

zero. From this point forward, the expected value of the firm remains relatively insensitive

21

Result 3’: Cursed managers acquire more information

Withcheaptalk,expectedutilityis:

𝑢e( ≡12𝐸(

r𝐸([𝑅 + 𝑏|𝑥]# − s𝐸([𝑅 + 𝑏|𝑥] − 𝑘'f𝑑(𝑥)gt#u

Result:Holding𝑏fixed,utilityincreasesinperceivedinformativenessof𝑑

• Utilityincreasesintrueinformativenessi.e.,𝑁and𝑝

• Utilityincreaseswithcurseofknowledge𝜔

• Forafixed𝑁,acursedmanageracquiresmoreinformationi.e.,

𝜕#𝑢(𝜕𝑝𝜕𝜔

=𝜕𝑀𝑈(𝑝)𝜕𝜔

> 0

Fixing𝑁,asmallincreasein𝜔 ⇒increaseinformationprecision

But,sufficientlylargeincreasein𝜔 ⇒decreasein𝑁(/0andinfo.precision

Sawtoothpatterninoptimallychoseninformationprecision

Result 4’: Cursed managers can increase firm value Expectedvalueincreaseswithinformativenessandcommunication:

𝐸[𝑉(𝑅, 𝑘'(𝑑)] = 7#(𝜇# + var(𝜃) − 𝐸[var(𝜃|𝑑)])

• Decreaseswithcurseofknowledgewithexogenousinformationprecision• Canincreasewithcursewheninfoprecisionisendogenous

Whencurse𝜔isnottoolarge,infoproductiondominatescommunication

Delegation versus Communication

Result 5: Delegate if and only if bias is sufficiently small

Result:Withexogenousinformation/messageprecision,thedelegation

decisiondoesnotdependonthecurseofknowledge

• Costlycommunication:Delegateiff

𝑏# <𝑝#𝜎#(1 − 𝜌#)

12

• Cheaptalk:Delegateiff

𝑏# <𝑝#𝜎#

12

Note:Commitmentimprovescomm⇒Delegatemoreoftenwithcheaptalk

Withendogenousprecisionchoice,curseofknowledgeaffectsinfoprecision,andsoaffectsdelegationdecision

Theprincipalmayretaincontrolwitharationalmanager(𝜔 = 0),butdelegatetoacursedone

CostlyCommunication CheapTalk

Otherparameters:𝜇 = 1, 𝜎 = 1, 𝑐! = 0.02, 𝜅! = 0.001

Extension: Verifiable Disclosure “Intermediatecase”betweencheaptalkandcommitment

Result:“Disclosureontop”versus“Disclosureatextremes”

(i) When 9$'(75:)

≥ 𝐵∗,managerdisclosesiff𝑥 ≥ 𝑥

(ii) When 9$'(75:)

< 𝐵∗,managerdisclosesiff𝑥 ≥ 𝑥OR𝑥 ≤ 𝑥

Result:Forfixedprecision𝒑,curseofknowledgereducesdisclosureandvalue

Result:Fixingequilibrium,higher𝝎canleadtohigherinformationacquisition,sovalueisnon-monotonicwith𝝎

Result*:Delegationdecisiondependson𝝎,evenwithfixed𝒑,fordisclosureatextremesequilibrium

Conclusions

Isthecurseofknowledgereallyacurseorablessing?

Asymmetricinformation⇒Curseofknowledge

Cursedmanagersreducefirmvaluewhen:

– Informationprecisionisexogenous(e.g.,reporting,riskmanagement?)

– Incentivesaremisaligned

– Informalinternalcommunication(lackofcommitment)

Cursedmanagerscanincreasefirmvaluewhen:

– Informationprecisionisendogenous(e.g.,marketresearch,R&D)

– Curseisnottoolargeandincentivesarewell-aligned

– Formalcommunicationsystems(abilitytocommit)

Enhancing/fosteringcommunicationencourageseffortinexpertise

Cursed Academics? Curseofknowledgeprovidesoneexplanationforwhymanyexcellentresearchersarealsobadatteaching– Morecursedacademicsoverinvestinexpertise,underinvestinteaching

Analysishighlightscomplementarityinresearchandteaching

Justlikeyourdeansays!

Encouragingbetterteachingpracticesmayimproveresearchproductivity

– Whatdoesn’twork:Warnings,feedback,experience– Whatdoeswork:Consideringalternativeoutcomes(counter-explanation)

Thank you!

Extension: Verifiable Disclosure

Suppose𝑚observes𝑥withprobability𝑞,andnothingotherwise

– Aninformedmanagercanchoosewhethertodisclosenothing(𝑑 = ∅)or

disclosethesignalperfectly(𝑑 = 𝑥)

– Anuninformedmanagercannotverifiablydiscloseheisuninformed

Denotetheequilibriumbeliefafterno-disclosureby𝜇∅ = 𝐸[𝜃|𝑑 = ∅].

Optimalinvestmentbyprincipal:

𝑘'(𝑑) = M𝜇 + 𝑝𝑥, if𝑑 = 𝑥𝜇 + 𝜇∅, if𝑑 = ∅

But,manager’scurseofknowledgeimplies:

𝐸([𝑘'(𝑑)|𝑥, 𝑑 = ∅] = 𝜇 + (1 − 𝜔)𝜇∅ + 𝜔𝑥

Result:Let𝐵∗ = %&'('(&'()+(

.

1. If ,-.(&'/)

> 𝐵∗,thenthereexists𝑥 ∈ (−𝜎/2, 𝜎/2),suchthatthemanagerdisclosesiff𝑥 ≥ 𝑥

(“disclosureontop”)

2. If ,-.(&'/)

≤ 𝐵∗,thenthereexists�̄� < 𝑥 ∈ (−𝜎/2, 𝜎/2),suchthatthemanagerdisclosesiff𝑥 ≥

𝑥or𝑥 ≤ �̄�(“disclosureatextremes”)

Intuition:

benefitsfromhigherinvestment(nodisclosure)

vs.

highervaluefrommoreinformedinvestment(disclosure)

Whenbiasissufficientlylarge,wehavestandard,disclosureontop

Whenbiasissmall,wehavedisclosureatextremes

• Smallerbias⇒informedinvestmentdominateswhenxissufficientlylow

• Smallerbias⇒Disclosureismoreinformative(fulldisclosurewhenb=0)

Impact of curse of knowledge on expected value

Effectivebias 9$'(75:)

increaseswiththecurseofknowledge

• Whenbiassufficientlylow,disclosureregiondecreaseswith𝜔• Whenbiassufficientlyhigh,disclosureregionisinsensitiveto𝜔

Intuition:Cursedmanagerbelievesrightansweris“obvious”

⇒incentivetowithholddisclosureishigher

Result:Forfixedsignalprecision𝒑,expectedvalueis:

• unaffectedby𝜔for“disclosureontop”equilibrium• decreasingin𝜔for“disclosureatextremes”equilibrium

Endogenous Information Acquisition

Result:Holding𝒃fixed,utilityincreasesinperceivedinformativenessof𝒅:

• Utilityincreaseswithtrueinformativeness(i.e,𝑝)

• Utilityincreaseswiththecurseofknowledge𝜔

Forfixedequilibrium,utilityincreasesmorewith𝑝forcursedmanageri.e.,

𝜕#𝑢(𝜕𝑝𝜕𝜔

> 0

Moreover,utilityishigherforthedisclosureatextremesequilibrium

Aswithcheaptalk,anincreaseincursednesscanincreasefirmvalueby

increasinginformationacquisition!

Cursed managers may increase firm value Forfixedprecision,curse(weakly)decreasesvaluevialessdisclosure

But,italsoincreasesmarginalvalueofinformationprecision,whichcanincreaseinformationproduction

Result:Delegationdecisiondependsoninformativenessofequilibriumdisclosure:

(i) Forthe“disclosureontop”(lessinformative)equilibrium,delegateiffover-investmentbiasissufficientlysmall

(ii) Forthe“disclosureatextremes”(moreinformative)equilibrium,delegateiffcurseofknowledgeissufficientlylarge

Disclosureontopequilibrium:

Forfixed𝑝,delegationdecisionindependentofcurseofknowledge𝜔(likecheaptalk)

Disclosureatextremesequilibrium:

Forfixed𝑝,disclosureintervaldependson𝜔,andsodoesdelegationdecision