WORKING

PAPERS

CONSUMER INFORMATION, PRODUCT QUALITY

AND SELLER REPUTATION

Carl Shapiro*

WORKING PAPER NO. 42

July 1980

nc Bureau fi Ecooomks working papers are preliminary materials circulated to stimulate discussion and critical COIIllllelll AD data cootained in them are in the
pablic domain. Tbis includes ioformation obtained by * Commisoo whidl has become part of public monl. ne aalyses aod coocluions set forth are those
tl the authors aod do IHi aecessarily reflect the vim of other members of the Bureau rl Economics, other Commissioo staff, or the Commissi on itself. Upon
request, single copies of tie paper will be provided. References iu pablicajoas to FfC Bureau of Ecooomia working papers by FfC erooOIIIQs (other than
acknowledgement by a writer that be has atttSS to such unpublished materials) should be cleared with the author to protect the tentative character of these papers.

BUREAU OF ECONO flCS

FEDERAL TRADE COMrtflSSION

WASHINGTON, DC 20580

 Co n s umer I n fo rma t io n , P ro du c t Qua l i ty

a nd Se l l er Re pu ta t io n

C a r l Sha pi r o *
M . I . T . a nd

Feder a l Trade Comm i s s i o n

De c em be r 1 9 7 9

Re vi sed Ju l y 1 9 80

* I wou l d l ik e to tha nk P e t e r Di amond , Jo e Fa rr e l l , Dr e w Fud enbe r g ,

E r i c Ma s k i n , Steve· Sa l o p , R i c h a rd Sc hrna l en s ee , a n d Je a n T i ro l e
for h e l pf u l c o mme n t s a nd sugg e s t i o ns .
 Table of Contents

Intro:luction 1
General Fonrulation of OJality Ch:::>ice by a 11
1-bnop::>list Under Imperfect Information
Once arrl For All Q.lality Ch:::>ice: General Fesults 14
Welfare Implications
24

Personal Learning: An Exten:led &ample 29

A Continuous T .ime Example of Once arrl for All QJ.ality 43
Choice
Product Q.lality Ch:::>ice Over T irre: REmarks and 48
·General Results

A continuous Time M::x:lel with Variable QJ.ality 56

Conclusion 63
lbtes 64

Ap_p:mdi.x
67
 Ec onomi s t s a r e becoming i n c r e a s in g l y aware o f how impe r f e c t

i n f o rma t i on c a n c a u s e a w i d e ra n ge o f ma r k e t impe r f e c t ion s .

A g r e a t d e a l o f e f fo r t h a s gone into c a r e f u l anal y s i s o f

unc e r ta inty a bo u t th e v a r i o u s p r i c e s a t wh i c h a homogenous

1
p roduc t i s be i n g o f f e r ed f o r s a l e . Ye t un c e rta inty about

p r i ce is r e l a t iv e l y minor a n d inexpe n s ive to e l imi n a t e i n

c ompa r i so n w i th unc e r t a in ty about o t h e r p roduct c h a r a c te r i s t i c s

( e . g . , durab i l i t y , s a f e t y o r t a s te ) . Th i s pape r i s c o n c e r n e d

w i th per fo rmance i n ma rke t s whe r e t he p ro d uc t s s o l d cannot

b e f u l l y a n d a c c ur a t e l y eva l ua t e d p r i o r to purch a s e . The

a n a l y s i s wi l l center on how a pro f i t -maximi z in g f i rm choo s e s

product qua l i t y i n an envi r onme nt whe re c on s ume r s c annot

o b s e rve qua l i t y p r i or to purch a s e . I am pa r t i c u l a r l y i n ­

t e r e s ted i n how t he q ua l i ty o f produc t s p rovided depend s on

the mann e r and s peed with whi ch c o n s ume r s gath e r i n f o rma t i o n

abo u t prod uc t s a n d e n t e r a n d l e ave t he ma rke t .

Qua l i ta t i ve unc e rta i n ty i s a w i d e s p r e a d and i mpo r t a n t

f e a t u r e o f ma rket s for mo s t f i rms ' good s a nd s e rv i c e s . Vi r ­

tua l l y a l l s e r v i c e s are impo s s i b l e t o eva l uate un t i l t hey

are u s ed . Th i s inc l ud e s llte d i c a l a n d l e g a l s e rv i c e s , a u t o ­

mob i l e r e p a i r , p l umb i n g a n d e l e c t r i c a l work , e t c . Ano t h e r

impor t a n t c l a s s o f pr o du c t s who se qua l i t y c anno t eas i l y b e

j udged p ri or to pur c ha s e a r e con sumer d urab l e s . I n fa c t , when

a new mod e l a u tomob i l e come s o u t , t h e re s i m p l y i s no way o f

knowing wha t i t s repair r e c o r d wi l l look l ike . A s a f in a l

cla s s o f e xamp l e s , t h e r e a re many p ro duc t s whi c h we buy

 2

q u i t e f r eq ue n t ly wh i ch have unob s e rvab l e a t t r i b u t e s : r e s t au r a n t

mea l s ( ta st e ) and c l o t h i ng ( wi l l i t f a d e o r s h r i nk ) a r e two

e x ampl e s. The r i se in bot h the comp l ex i ty o f p ro duc t s ( cons ider

h i f i equ i pme nt o r e t h i c a l d r ug s ) and t h e f r a c t i o n o f income

spent on s e rv i c e s ha s i n c r e a se d the impo rt a n ce of the s e i n f o r ­

mational prob l e m s .

The per forman c e o f the marke t in such a s e t t ing i s

e s s en t i al to t he eval u a t i o n o f a wide var i e ty o f r e g u l atory

initiatives . The s e i n c l ud e o c cupat ional l i c e n s i n g (o r min imum

qual i ty s t andard s in g e n e r al ) , oc c upat ional h e al t h r e g u l a t i o n s

( th e wor ke r t aking t h e r o l e o f t he c o n s ume r ) , a utomo b i l e

s a fety r u l e s , and a w i d e r ange o f r e g u l at i o n s b y t he Food

and Drug Admin i s t r at i on , F e de r al Tr ad e Commi s s ion , and Con­

sume r Product S a f e ty Commi s s ion. Rec en t a c t i o n s su ch a s

the FTC ' s p ropo s ed u s e d c a r r u l e ar e d e s i gned p r e c i s e ly to

improve the in format ion in the marke t and t he reby enhan c e

per f orman c e . I t i s paramo n t , t h e re f o r e , to de s c r i be marke t

p e r f o rmanc e i n t h e ab s e n c e o f po l i c y i n t e r f e renc e t o be ab l e

t o eva luate t h e wo r t h ine s s o f e i ther mandato ry s tandards

or pub l i c p rov i s io n o f in format ion.

Th e analy s i s i n th i s pape r i s r e s t r i c t e d to t he c a s e

whe re a monopo l i s t c o n t ro l s t he quan t i ty a s we l l a s q ua l i t y

o f t h e g o o d i n que s t ion. S i n ce impe r fe c t i n fo:L.Ii.latie7.t'• gene r a l l y

l e ad s t o s ome marke t powe r , t h i s anal y s i s i s a n e c e s sary

p r e r e quis i te to s tudy i n g many s uppl i e r s in the pre sence o f

 3

qua l i ta t i v e u n c e r t a i n t y . One t re a tme nt o f s uc h ma r k e t s wi t h

many s e l l er s i s Shap i ro [1 9 8 0) .

The p a p e r i s organ i z ed a s fo l l ow s : f i rs t I d i s c u s s a

var ie t y o f mod e l l in g probl ems wh i c h a r i s e o n c e the per f e c t

in forma t ion a s s ump t ion i s removed. Th i s i s meant b o t h t o

f ami l i a r i z e the r e a d e r w i t h t h e prob l ems a n d i n t ro du ce some

propo s ed so l ut io n s . Then I pr e s en t the ana l y s i s in the c a s e

whe r e the s e l l e r s e t s qua l i ty o n c e a n d f o r a l l ( a n e w product ,

fo r e amp l e ) . Fina l l y , I t r e a t t h e ca s e wh e r e s e l l e r s can

vary th e i r qua l i ty over t i me .

 4

P ro du c t Cha r a c t e r i s t i c s a n d Qua l i ty

S in c e t h e i s su e s d e s c r ib e d above r evo lve a round product

c ho i c e by a s e l l e r r a t h e r than s imp l y p r i c in g or output

d e c i s i on s for a g iven p r o du c t , it i s ve r y u s e f u l t o adopt

L an c a s t e r s ' f r amewo r k and v i ew produ c t s as b und l e s o f

c h a r a c te r i st i c s. When produ c t s a r e v i ewed in t h i s wa y , o n e

que s t ion whi c h a r i s e s i s whe ther o r n o t the ma rket prov i d e s

2
t h e s oc i a l ly opt imal m i x o f p roduct c ha r a c t e r i st ic s . S ince

I wi sh to fo c u s o n product qu a l i ty a s o ppo se d t o p r o duc t

va r i e t y , I r e s t r i c t q ua l i ty to a s i ng l e d imen s io n . T hi s

shou l d be thoug h t o f a s some prod u c t a t t r ibute wh i c h i s

d i f f i cu l t o r impo s s i b l e to o b s erve p r i o r t o purcha s e , bu t

wh i c h con sume r s a l l l ik e t o have mor e o f. Examp l e s in c lude

durab i l i ty , s a f e ty ( p robab i l i ty o f no a cc ident}, o r speed o f

service . The r e s t r i c t ion o f qua l i ty to o n e d i me n s ion imp l i e s ,

i n pa r t i c ul a r , t h a t a ny two c on s umer s wi l l a g ree wh i c h o f two

produc t s i s p re f e r r e d a l though they may d i s a g re e a s to how

3
much the added qua l i ty i s wo r t h .

Enodenous Qua l i ty Cho i c e : A Dyn am i c P rob l em

To under s t a nd the p rob l ems wh i c h o c c u r when produ c t

qua l i ty i s unob s e rva b l e p r i o r t o pur cha s e , i t i s impo r t an t

to d i s t ingu i s h the c a se whe re s e l l e r s ho product qua l i ty

f rom the c a s e whe r e t he r e i s an exogenous s upp l y o f p rodu c t s

o f d i f f e rent qua l i t i e s . T h e l a t t e r c a se i s t h e o n e i n t ro du c e d
 5

by Ake r l o f i n h i s s emin a l a r t i c l e U97 0 ] . Un fo rt un a t e ly , mod e l s

wi t h exogenous q ua l i t y supp l y a r e o f l im i t e d u s e f u l ne s s i n

product ma rk e t s . I f t h e p r i c e o f fe red d e pende d on l y on

ave rag e q ua l i ty in the ma r ke t , a n d h i g h e r qua l i ty i t ems a re

mor e co s t l y to produce , then i n a one-shot mod e l t h e r e i s n o

inc ent i ve f o r a g ive n prod u c e r t o provide any t h i ng o t h e r t h a n

min imal qua l ity. Con s equen t l y , the ma rke t w i l l be o ve r r un b y

. . ' 4
mln lma 1 qua 1lty ltems.
.

The s ame r e s u l t o c c u r s in a dynamic

mod e l if c o n s ume r s do not l e a rn about the qua l i ty of i n d i vi d u a l

f i rm s o ve r t ime. The inc e n t i ve to prod u c i ng h i gh qua l i ty i tems

i s that h i gh e r qua l i ty today wi l l c a u s e the demand cu rve i n the

f u t u r e to s h i f t o ut. So p ro d u c t quali t y cho i c e s by s e l l e r s

a re f und amenta l l y dynami c .

I n fo rma t i on Flows

It is appa r ent f rom the s e con s i d e r a t i on s th a t the t e c hn o l og y

o f i n f o rmat i on f l ow w i l l b e e s s e nt i a l to a n y s t o ry o f ma rk e t

e q u i l ib r i um. I t i s conve n i e n t to d i s t i n gu i s h three f a c e t s o f

th i s t e c hno l o gy : ( 1) How i n forma t i on f l ow s among con s umer s ,

(2) How muc h a con sume r l e a r n s a bo u t the p ro duc t 's qua l i ty

f rom u s i ng i t , a nd ( 3) how con s ume r s en t e r and l e a ve the

mar k e t. Bo th the n a t u r e o f the p roduct a nd the i n s t i tut ion s

s ur r o und i n g the ma r k e t for th e good i n f lue n c e the way i n wh i c h

i n f o rma t i on f l ows r e ga rd i n g e a c h f i rm' s p roducts .

 6

P e r sona l L e a r n ing

The f i r s t stage i n the p ro c e s s by wh i c h a f i rm's q ua l i ty

today i s t r a n smi t t ed to pote n t i a l future b uy e r s i s through

the ob s e rv at i ons of tho se who have a ct ua l ly u s e d the p roduct .

S o the abi l it y o f con s u me r s to ob s erve product c h a r a c t e r i s t ic s

i s c ruc i al .

P ro du c t a t t r ibute s wh i c h ca n b e ob se rved p r i o r t o pur ch a s e

have b e e n c a l l e d s e a r c h a t t r ibut e s ( by N e l s on [ 1 9 7 0]) . They

p r e s e nt t h e s ame i n f o rma t i ona l probl ems a s do p r i ce s . I n my

o p i n ion , the s e probl ems a r e mi n o r i n c ompa i son wi t h t ho s e

i nvol v ing a t t r i b u t e s wh i ch r e q u i r e u s e t o b e o b s e rve d . The s e

l at t e r c ha r a c te r i s t i c s have be e n d i v i d ed i n to two c l a s se s :

e x pe r ienc e a t t r i butes wh i c h a r e o bs ervab l e a f t e r u s e (e .g. ,

taste ) and c redence a t t r i bute s wh i c h may rema i n unob s e r vab l e

even a f t e r u s e . ( e . g . , the s t ructura l in t e g r i ty o f a n auto­

mob i l e ) .5 It is i n t u i t i v e l y c l e a r t h a t c o n s ume r i n fo rma t ion

prob l ems are mo s t severe f o r c redence p rope r t i e s o f p roduc t s .

I n f ac t , mo s t r e gu l at i on s rega r d i ng p roduct a t t r ibu t e s a r e

d i r e c t e d a t c r e d e n c e a t t r ibut e s . L i ce n s i n g o f doc tor s , mo s t

r e gu l a t i o n s by t h e Food a n d Drug Admi n i s t r a t io n a n d t h e Con­

s ume r Product S a f e ty Commi s s i on a r e e xamp l e s .

F o r mo s t produc t s the d i s t i n c t i o n be tween e xpe r ie n c e

and c redence prope r t i e s is probab l y overdr awn . The r e i s

some l ea r n i n g a bout produc t a t t r ibutes wi t h u s e , but a t

v a r y i ng spe e d s f o r va r io u s produc t s . I t i s very u s e f u l

 7

i n thi s context to t hink o f t h e c o n s ume r a s o b s e rv i n g s ome

outcome wh i c h depends o n bo t h the q ua l i ty o f the prod u c t and

6
some unob s e rv a b l e r andom va r i a b l e . In the c a r s a f e ty e xamp l e ,

i f a c a r h o l d s up we l l in a n a c c i d e n t i t i s not c l e a r whe th e r

t h e m i n ima l dama ge i s due to the type o f a c c i d e n t o r t h e w a y

the c a r wa s made . A s imi l a r l a c k o f o b s ervabi l i t y h o l d s t rue

fo r d r ug s , s e r vi c e s , e t c . I f a l awye r l o s e s a c a s e i t i s h a r d

to know i f he wa s b a d o r th e c a s e wa s we ak.

On e way to i nc o rpo r a t e l e a rn i n g i n to c o n s ume r cho i ce i s

t o a s sume that qua l i ty i s po s i t ive ly r e l a t ed to th e p robab i l i t y

o f repe a t pur c ha s e . Th i s appro a c h ha s been taken i n Schma l e n s ec

[ 1 97 8 ] and Sma l lwood - Co nl i s k 11979]. One prob l em w i t h i t i s t hat

7
i t g i v e s no i n s i ght a s to how c o n s umer s re spond to pri c e.

I have taken a d i f f e r en t approa c h i n wha t fo l l ow s . At

any po i n t in t ime a c o ns um e r ha s s ome e xpe c t a t i on s r e g a rd i ng

p roduc t qua l i ty. Thi s d e t e rmine s the po s i t ion of h i s demand

c urve . The l e a rn ing then c o r r e s pond s t o a d j us t in g e x p e c t e d

qua l i ty towa r d s t rue q ua l i t y . Produc t s for w h i c h qua l i ty i s

ha rd to o b s e rve even a f ter u s e wi l l d i sp l a y s l ow o r l a gged

a d j u s tment of co n s ume r e xpe c t a t i on s. I t sh0u l d b e n o t e d t h a t

t h e ab i l i ty o f con sume r s to d r aw in f e r en c e s a bout the f i rm ' s

qua l i ty from u s i n g the produc t i s not so l e l y techno l o g i c a l l y

determined. To the ext e nt t h a t f i rms have imp e r f e c t qua l i ty

con t r o l a c on s umer who g e t s a bad i tem may have d i f f i c u l t y

 8

knowing whether i t s poo r qua l i t y i s r e p r e se n t a t ive o f t h a t f irm

o r not . For the purpo s e s o f t h i s d i s cu s s ion , I w i l l avo i d t h i s

p ro b l em by a s s uming th a t a l l t h e prod u c t s a f i rm s e l l s on a

g iven d a t e have un i fo rm q ua l i t y .

I n t e r pe r sona l Lea r n i n g and the Ma r k e t f o r I n fo rma t io n

The re a r e q u i t e a f e w potent i a l sourc e s o f i n f o rmat ion

about p r o du c t qua l ity in add i t io n to p e r so n a l expe r i e nc e.

The se ar e (1) Exp e r i e n c e o f f r i e nd s , ( 2) P ub l i c a t i o n s ,

8
0ithe r pub l i c o r p r ivate , s uc h a s C on s ume r Repo r t s , (3)
Adve rt i s ing , and (4) P o t en t i a l s i gna l s of qua l i t y such a s

p r i c e , warran t i e s , o r adve rt i s i n g .

I f the ma r k e t for in forma t i o n a b o u t p r o d u c t a tt r ibut e s

worked we l l , t h e i n f o r ma t i on a l prob l ems i n t h e ma rke t s f o r

fina l good s a n d s e r v i ce s wou l d be s ub s t an t i a l ly r e d u c e d o r

e l imina t e d. T he r e ar e seve r a l r e a s o n s wh y the in f o rma t io n

ma r k e t c annot be expe c t e d to wor k we l l , howe v e r.

The f i r s t two r e a sons r e l a t e to the pub l i c good nature

o f i n forma t io n. The f i r s t i s t h a t i t i s d i f f i c u l t t o prevent

r e s a l e of i n f o rma t io n . A p r i va t e f i rm that s e l l s informat ion

may not be able to surv ive even though the r e is a sub st an t i a l

v a l u e t o the in forma t i on , b e c a u s e there wou l d b e n o re s t r i c ­

t i on s o n i n d i v idua l s who b uy t h e in forma t io n pa s s ing i t a l ong

( free l y ) to non-buye r s . The s e cond r e a son i s due to the

os i t i ve e xt e rna l i ty c re a t e d by in forme d b uye r s wh i c h bene f i t s

u.in f o rmed b uy e r s t hroug h ra i s ing qua l it y ( s imi l a r t o what

 9

happen s i n S a lop- S t i g l i t z [ 1 97 7 )) . T h i s q ua l i ty in f l u e n c e i s n o t

a c c oun t e d f o r by c o n s ume r s wh e n c on s i de r ing whe ther t o b u y

·
in fo rmat ion o r not , and hence too l i t t l e is pur c ha s e d.

A f i n a l r e a son why p r i va t e prov i s i o n o f i n f o rma t io n may

not b e pos s i b l e i s du e to c r ed i b i l i ty prob l em s . Con s ume r s

may fear that the i n forma t ion supp l i e r i s n o t e n t i r e ly c a nd i d ;

p e r h a p s he i s be ing b r ibed by s ome p roduc e r s , o r unde r t h e

i n fl ue n c e o f hi s a dve r t i s er s . A gove rnme n t a l p rovi s io n o f

i n forma t ion m a y be a b l e t o a vo id the s e probl ems . Th i s o f

c o u r s e depends o n how s u c h a n a g e n c y i s s e t u p t o prov i de

i n c e n t ive s f o r t r uth f u l p rovi s io n o f in forma t i o n .

I n o rd e r t o e va l ua t e t h e d e s i r a b i l i ty o f in f o rma t io n

a c t i v i t i e s b y p ub l i c a u t ho r i t i e s , i t i s n ec e s s a r y f i r s t to

un d e r stand the advant age s wh i c h d e r ive f rom imp roved i n f o rma ­

tion . That i s the go a l o f the ana ly s i s to fo l low. The

c ommen t s above are i n tended to h i g hl i ght why the l a i s se z ­

fa i r e l eve l o f i n f o rma t ion c an b e e xp e c t e d to b e s ubopt ima l .

The po s s i b i l i t y that wa r r a n t i e s may prov i d e in f o rma t i on

about qua l i ty wi l l b e r u l ed out in t h i s p ap e r. Spe n c e

[ 197 7 b] ha s shown tha t when qua l i ty r e f e r s to probab i l i t y o f

fa i l ure o f a n i t em , wa r r an t i e s may s e rve a s s i gna l s o f

qua l i ty . F o r many goods this i s po s s i b l e on l y to a v e r y

l im i t e d e xte nt . Take a wa sh ing mac h i n e , f o r e xamp le . A

war ran ty may hold f o r 1 o r 2 y e a r s , but i t i s n o t po s s i b l e

e v e n fo r manu f a c ture r s o f good ma c h ine s to o f f e r comprehen s i ve

 10

wa rran t i e s . T h i s i s b e c a u s e th e manu f a c t urer c a nn o t mon i tor

ma i n t e na n c e or i n t en s i ty of u s e . S u c h mora l h a z a rd and

advers e s e l e c t ion pro b l ems a re i n e v i t a b l e b arri e r s t o t h e

a b s o r pt i on o f pro d u c t q ua l i ty ri s k s b y t h e produc er . In

t h e e xamp l e o f l e g a l s e rv i c e s , i t i s po s s ib l e t o wri t e

inc en t ive c on trac t s , b u t a dve rs e s e l e c t ion l imi t s t h e i r

9
s c o pe .

F i na l l y , i n th i s paper I do not inc l ud e advert i s in g ,

a l though i t s v i rt u e s a s we l l a s f a ul t s mu s t b e ana l y z ed

in the context o f impe rf e c t in forma t io n . Advert i s ing

c a n b e v i ewed as a l tering c o n s umer ' s e x pe c t a t i o n s of q u a l i t y .

S i n ce there i s an obv i o u s i n c e n t ive for produc e rs to over­

ra t e the ir produc t , th e k e y que s t i o n h e re i s why con s umers

pay any a t t e n t ion to such c l a ims . The ab i l i ty o f advert i s in g

to c onvey i n f orma t io n abo ut pro duc t qua l i ty i s some t h ing I

hope to t reat i n the f u ture .

I have t r i e d i n t h i s s e c t i on t o i d e n t i fy f a c tors wh i c h

i n f l uence t he spee d o f l e a rn in g b y c o n s ume rs . T h e a na l y s i s

be low foc u s e s o n t h e re l a t io n s h i p b e tween t h a t s pe e d and

the q ua l i ty o f pro duct c ho s e n b y th e mo nopo l i s t . In

e xampl e s whe re t h e o n l y l ea rn in g i s person a l l e a rn in g ,

i t i s po s s ib l e to b e q u i t e expl i c i t about how the f i rm ' s

demand c urve s h i f t s i n re s po n s e t o l e arn i ng . Such e x ampl e s

p l a c e a n uppe r bound o n t h e in fo rma t i ona l pro b l ems i n t h e

ma rk e t , s i n c e th ere a re in f a c t add i t iona l informa t ion

sourc e s , as d i s c u s s ed above .
 11

Gen e r a l F o rmu l a t i o n o f Qua l i ty Cho i c e b y a Monopo l i s t Un de r

I mpe rf e c t In forma t io n

Con s i de r the prob l em f a c ed b y a f i rm i n s e t t i ng i t s q ua l i ty ,

q. At a n y po i n t i n t ime the f i rm c a n r e a p e x t ra pro f i t s b y

c u t t ing qua l i ty ; t h e re w i l l b e n o l o s s i n revenue s u n t i l

c on s ume r s c a n r e s pond to t he q ua l i ty change.

S i n ce the probl em i s e s s e nt i al l y dyn ami c , let t d e no te

t ime an d c a l l p ( t ) the pr i c e at t ime t , q ( t ) the q u a l i ty , a n d

x (t) sal e s . The re i s a l s o a c o s t f un c t ion c ( x , q ) i n quan t i t y

a n d qua l i t y .

There may a l s o b e c o s t s to c h a n g i n g qua l i ty i . e . , o n c e

a n d f o r a l l c o s t s to i n t ro d uc i n g a new q ua l i ty l i ne . I i gn o r e

th e s e for the mo s t pa r t , e xc e pt to note t h a t they j u s t i fy

the a t t en t i on pa i d t o o n c e - and- f o r - a l l gua l i ty c ho i c e s b e l ow .

I n f a c t we do not u s ua l l y o b s e rve f requent qua l i ty c han g e s

b y s e l l e r s ; pr i c e c ha n g e s o c cu r much mo re r api d l y . Th i s

i s f u r t h e r j us t i f i c a t i o n for the t r e a tme n t i n t h e fo l l ow i n g

s e c t io n .

Pro f i t s a t t ime t a r e g iven by

TI ( t ) = p ( t ) x ( t ) - c (x (t) , q (t) ) .

The c ru c i a l que s t ion i s how x ( t ) d e pe n d s on p ( t ) and pre v i o u s

10
qua 1.1ty c h o 1c e s .
. One approach i s to t r e a t x ( t ) a s t h e

s t a t e va r i ab l e r e pr e senting a l o y a l s e t o f pa tron s . Then

the i n f l ow of n e w custom e r s a n d the o u t fl o w o f d i s sa ti s f i ed

c u s tome r s w i l l i n ge ne r al d e pe n d on q ua l i ty , pr i c e , a nd thP

rtoc k of c u s to me r s itself. Small wood and Conl i sk [ 1 97 9 ] took

 12

th i s a ppro a c h i n. di s c r ete t ime w i th qua l i t y b e in g t he probab i l ity

any g iven c ustomer wou l d r e turn the n e xt p e r i o d . Th i s r e ­

qu i red t h e m to t a k e p r i c e a s f i xe d a n d e xo ge no u s , a n d they

co u l d n o t g e ne r a l l y de te rmin e wha t q ua l i ty a f i rm wou l d

c ho o se . Furth e rmo r e , they di d n o t a l low q ua l i ty changes ove r

11
t ime . Farre l l [ 1 9 7 9] has a l so t a k e n th i s app ro a c h a nd i s

a b l e t o s t udy the opt ima l g ( t ) pa t h fo r a f i rm , aga in t a ki n g

p r i c e a s f i xe d a n d e xogenous .

The approach i n th i s paper i s qu i t e di f f e re n t s in c e I

am i n t e re s t e d in a l lowing p r i c e c ha n g e s ove r t ime , and v i ew

'
the f i rm s reput a t i on a s the st a t e v a r i a b l e r a t h e r than i t s

sa l e s . I n t h is v i ew , previous q ual i ty and s a l e s i n f luence

what var i ous c o n s ume r s t h i nk about t h e q ua l i t y o f a f i r m ' s

product , and i t i s t h i s reputat ion , R ( t ) , wh i c h de t e rmi n e s

t h e l o c a t i on o f t h e f i rm ' s deman d c ur ve . Con se que n t l y ,

the f i rm s • p r i c i n g de c i s i o n s over t ime can be s tudi e d, a n d

consume r s' i ndiv idual deman d c urve s c an be de r i ve d from

12
ut i l i ty f un c t i o n s an d expe c t a t i o n s of gua l i ty .

Th i s appro a c h doe s req u i r e s pe c i f y i ng t he proc e s s b y

whi c h reputa t ion a dj u s t s ove r t ime . I a r g ue s t rong l y fo r

a dapt ive e xpe c t a t i o n s by c o n s ume rs i n r e s po n s e to qua l i ty

chang e s by a se l l e r . The r e i s no ev i de n c e at a l l t o s up po r t

a mo re soph i s t i c a t e d appro a c h b y con s ume r s i n wh i c h they

so lve o u t t h e f i rm ' s opt ima l c o n t ro l p rob l em to f i gure o u t

 13

wha t t o e xp e c t . I n st e a d , e v i de n c e on c o n sumer beha v i o r sug g e s t s

st rong l y t h a t consume r s e xt ra po l a t e f rom r e ce n t e xp e r i en c e t o

13
p r e di c t future p roduct p e r f o rma n c e .

Vi ewin g a f i rm's r eput a t i on a s t h e sta t e var i a b l e i n t h e

dynam i c se t t i ng out l i ne d above , t h e f i rms prob l em i s :

00

( -rt
max .} e (X ( t ) p ( X ( t ) 1 R (t) ) - C ( X ( t ) , q ( t ) )) dt
x(t) 1 q (t)
t=O
.
sub j e c t to R ( t ) = f ( x( t) , q ( t ) , R { t) ) .

R ( O) g i ven .

He r e the i nv e r se deman d c u rve t h e f i rm f a c e s a t t ime t i s

p (x(t) , R ( t ) ) ; it s l o c a tion de pends o n r e puta t io n , n o t q ua l i ty

a t t ime t ( si n c e that wil l not b e o b s e rv e d unt i l l a t e r ) . Now

co nsumer l e a r n i n g wi l l be e mbo di e d in the f ( x , q , R ) f un c t i o n .

T h e discussi o n above about i n forma t ion sour c e s in f l ue n c e s f i rm

be havio r t hrough f . Con s equen t l y , i t wi l l be v e ry impo r t a n t

t o se e how t h e opt ima l c ho ice o f q ( t ) and x ( t ) depend o n t h e

spe c i f i c a t i on o f f .
 14

on ce an d for A l l Q ua l i t y C ho i c e : Gene r a l Resu l t s

I n this se c t ion I sha l l c onsi der t h e prob l em f a c e d by a

monopo l i st when he i s c hoosing o n ce and for a l l wha t qua l i ty

produc t to produc e . I t i s best thoug h t o f a s i n troduct io n o f

a new product , bu t app l i e s a t a n y po i n t whe r e a l o n g - r un

qua l i ty cho i c e i s b e i n g mad e .

Suppose consume r s a l l e xpe c t t h e qua l i ty produc e d t o b e

R • That i s , suppose t h e f i rm h a s a n i n i t i a l reput a t i o n o f

0
R ( R and q a re me a sured i n t he same un i ts ) . F o r now I t ak e
0
R a s beyo n d the control o f the f i r m. I f a dve r t i si ng i s
0
pe rm i t t e d , i t ma y we l l ope r a t e t hrough a l te r i n g R • One c o u l d
0
st udy how muc h the f i rm wou l d a dve r t i se t o i n f l ue n c e R
0
f avorab ly . O f course , suc h a dve r t i s i n g need n o t b e i n fo rma t ive .

R may a l so depend on t he qua l i ty o f pro duc t s a l re a dy i n the

0
ma r k e t ( more o n t h i s be l o w ) . I n the fu l l mo de l whe r e qua l i ty

c hange s ove r t ime a r e permi t ted , R wi l l d ep e n d on previ ous

t
qual i ty a nd sa l e s cho i c es by the f i rm i t se l f .

Suppose , g iven i n i t i a l reputa t io n R the f i rm e l e c t s to

0
produce a pro du c t o f qua l i ty q . It t h e n c hoo s e s a n op ti ma l .

pr i c i n g pa t h ( a n d c o r r e spond i ng sa l e s p a t h ) over t ime to

ma xim i z e the p r e se n t va l ue of i t s p ro f i t s t r eam. Wha t suc h

a p a t h looks l ike w i l l depend on how l ea r n i n g o c c urs b y consume r s.

To r ema i n p e r f e c t l y g e n e ra l , denote the o p t i m i z ed present va l ue

 15

o f pro fits f rom choo sing q g i ve n R b y V ( q , R ) . So long a s

0 0
the r e i s d i scoun t ing we e xpe c t t h i s to b e f in i t e , a n d i n i t i al

r eputa t io n to b e o f some v a l u e , i . e . ,

( Al ) v ( q , R) > 0 f o r a l l q , R >0.
2 0 0

( he r e sub sc r i pt s de note pa r t i a l der i va t i ve s ) .

( A2 ) V ( q , R ) i s f in i t e for a l l (q, R ) and i s c o n t inuou s l y

0 0
dif f er en t i ab l e .

Suc h a V f un c t ion p ermits qu i t e g e n e r a l l ea r n i n g beha v i o r o n

t h e p a r t o f c o n sume r s . I wi l l bound t h e cho i c e o f g f rom

r l ow by a s s um i n g t h a t f o r q O ( t h is i s j u st a norma l iz a ti o n )

consume r s can detect the poor q ua l i ty and wi l l no t buy t h e product.

T h e re fore V ( q , R) = 0 for q O.

In o r d e r to dis cu s s th e f irms optima l qua l ity c ho i c e

when fa c i ng i n i t i a l repu t ation R , i t is ve r y u s e fu l t o

0

i de n t i f y the qua l ity whic h wou l d be c ho s en i n a pe r f e c t

i n fo rma tion wo r l d. De f ine

W (q) = V ( q , g ).

Th i s i s the pre sent d i s counted pro f i t s f rom c ho o s in g q und e r

14
p e r f e c t ln
. fo rma tlo n . . I w i l l a s s ume

(A3) W ( O) = 0 , W' ( 0 ) > 0, a nd W ' ( q ) < 0 f o r q l ar g e .

Consequentl y , the qua l i ty c ho se n und e r pe r f e c t i n f o rma tion ,

q * sa ti s f i e s

(1) W ' ( q* ) = 0

( Th e re may b e s e ve r a l roots to (1), but q * i s the be t on e ) .

I t i s now qu i te e a s y to i d e n tify o n e s e n s e in wh i c h impe r f e c t

i n fo rmati o n wi l l l e ad the f i r m to shade o n qua l i ty:

 16

Theorem 1 . U n de r A l , A 2 , and A 3 i f con s ume r s e xp e c t th e f irm

to produ c e a t th e pe r f e c t i n fo rmat io n p r o fit maxi ­

mi zing qua l i ty l ev e l q * , it wil l be op t ima l for

the f irm to p ro duce a lower qua l ity .

Pr oo f : We know W • ( q * ) = V1 ( q * , q * ) + V2 ( q * , q * ) = 0
By Al we know v 2 ( q * , q * ) > 0. Con seque n t ly ,

v ( q* , q * ) < 0. T hu s the curve V ( q , q * ) , th e re l ev a n t

1
one for a firm f a c in g initia l re puta t io n q * , c ut s

W (q) f rom above a t q * . Fur t h e rmo r e , a g a in by Al ,

for q > q * , W ( q ) = V(q, q ) > V (q , q * ) so W ( q ) l i e s above

V(q, q*) . Simil a r l y, for q < q * V ( q, q * ) > W ( q). A

c on s eque n c e o f these f a c t s is that V ( q, q * ) mu s t be as

shown in Figure 1 . T h er e fo r e th e o p t ima l c ho ic e o f

qu a l ity f a c in g init ia l r e p u t a t io n q * , q * * in t h e Fig ur e ,

is l e s s t ha n q * . •
$

FIGURE l
 17

Not e t h e g r e a t g ene r a l it y o f T h e o r em 1 . I n parti cular i t

requ i r e s no c on c a vity a s s ump t io n s .

by
I f we den o t e B ( R ) t h e b e s t c ho ic e o f q ua l it y f a c i n g
0

i n i t i a l r e puta t io n R , t h e above a r gume n t shows tha t f o r

0

R > q* , B ( R ) < R . I t i s nat ural t o a s k whe th e r th e r e is some

o- o o

qua l it y l ev e l q suc h that B ( q ) = q . That is , i s t h e r e s ome

qua l i ty l eve l s uch tha t i f c o n s um er s e xp e c t the firm t o produ c e

a t t ha t ·qua l i ty l eve l i t wil l be optima l f o r t he firm to do so?

Without ma kin g further a s s umption abo ut th e V fun c t i o n

i t i s not po s s ib l e t o conc l ude th a t such a qua l ity l e v e l e xi s t s .

I f V is con cave in it s fir s t a r gument , howeve r , t h e r e wil l b e

suc h a s e l f - f u l f il l ing qua l it y l ev e l :

Theorem 2 . Unde r Al -A3, if V is concave in it s f ir s t a r gument

then there exis t s a q ua l it y l eve l q < q* such tha t

i f con sume r s e xp e c t t h e firm t o produc e q , i t wi l l

i n f a c t be optima l for t h e f irm to do so .

P roo f : Con s ider the fun c t ion v ( q , q ) . We know

1
v1 ( 0 I 0 ) + v2 ( 0 , 0) = wI ( 0) >0 . Al s o v 2 ( 0 , 0) = 0

bec a u s e I a s s u me V(O, R ) = 0 for a l l R . There fore

0 0

A l s o , by t h e a r gumen t above v ( q *, q * ) <0.

1
So , sin c e v 1 ( q, q) is c o n t inuo u s a n d i s po s it ive a t q= O

and n e ga tive a t q=q* th ere mu s t b e s o me qs( O , q * ) such

tha t v ( q , q) = 0 by t h e in t ermediate va lue theorem .

1
 18

Now if V is concave in its first argument and v1(q, q) = 0 ,

then q maximizes V (
q, q), s o facing initial reputatio n q
15
the f. rm p ck s q=q. •
"

It is instructive to cast these results in term of the

B (R ) function mentioned above . The concavity assumption guarantees
0

that B( R ) is a continuous function. Then Theorem 1 tells us that

0

B (R ) lies below the 4 5° line fo r R >q*. The simplest case is

o o-

when there is a single solution to B (q)=q, as in Figure 2 .

It is possible that B (R ) inte rsects the 4 5° line several
0

times, however. Generically there are an odd number of such

inter sections. See Figu re 3 . The shape of the V function
corresponding to Figure 2 is shown in Figure 4.

•
q 'fS
/

R
0

FIGURE 2
 19
q

B(fto)

--
-- ------------------ ---
R
- o
F I GURE 3

FIGURE 4

 20

This self-fulfilling quality level (any of them if several

exist) has some peculiar properties. First of all, the monopolist
does worse at q than he would under perfect information, since

W (q) < W(q *). Therefore, if the monopolist could commit himself
to producing q* and convey this commitment to consumers, he
could do bet ter. This provides one j ustifica tion for warrant ies;
the monopolist could promise to pay customers if they get a
product. of quality less than q*. He would then credibly commit
himself to producing q* and could achieve W (q*). Without the
warrant y, however, after inducing expectations q* he would
want to cut quality, by Theorem 1.
If consumers are sophist icated and solve ou t the firms'
op timal choice problem we would expect to see q produced. It is
a rational expectations equilibrium quality level in t he
following sense: if all consumers expect q the firm will fulfill
16
the . r expectat 1ons.
.
Such calculat ions by consumers is not consistent with observed
consumer behavior, however, which is rather more adaptive and
extrapolative. 1 7 Wit hou t the self-ful filling expectations·
requirement it is not possible to pin down wha t quali ty is
chosen by the monopolist, since R 0 is exogenous. One story
which leads to firms producing q is that there is a sequence
of firms introducing t he produc t, and consumers expect an
entering firm to have quali ty equal to the lowest of all
 21

e x i s t in g f i rm s . Suppo s e t h a t t h e f i rs t f i rm to e n t e r f a c e s

e x pe c t a t i o n s q • I t th en pro d u c e s a p roduc t o f qua l i t y q =B{q ) .

0 1 0
Con s i d e r f i rs t t he c a s e whe re q >q and a s s ume F i gure 2 i s
0
18
the re l evant B ( q ) f un c t i o n . Then B ( q ) < q . The s econd
0 0
f i rm t h e re f ore f a c e s expec ta t i o n s q and p roduces q =B ( q ) . The
1 2 1
pro ce s s converg e s to p ro d u c t i o n of q b y l a t er e n t ra nt s . The

s ame re s u l t ho l d s when q < q . Th i s a rgume n t a l s o go e s through i f

0
con s ume rs expec t a new e n t ra n t ' s qua l i ty to b e tha t o f th e mo s t

rec e nt entrant.

I t i s po s s ib l e to ind i ca t e in th i s framewo rk how t h e s p e ed

o f l ea rn ing by con s umers a f f e c t s the f i rm ' s opt ima l q ua l i ty

c ho i ce . S uppo s e a l l con sumers h o l d t he i r i n i t i a l expe c t a t i o n s

of q f o r K pe riod s , a ft e r wh i c h t h e y wi l l l e a rn t h e t ru e
0
. 20
qua 1 1ty . Denote by u ( q ,q ) th e pro f i t s e a rned i n o n e p e ri o d
0
when a l l con s umers be l i e ve the q ua l i ty i s q b u t in f a c t i t i s
0
q. C learly u < o , u > o . I f a l l l e a rn i n g o c curs a f t e r K pe ri o d s
1 2
a n d t he d i s co unt f a c t o r i s p , O < p < l , then the pre s e n t va l u e o f

p ro f i t s from c ho o s ing q i s

K
l-p K
V(q, q , K) = U ( q , q0 ) + p W (q) .
0
l-p
T h e f i rs t ord e r condi t ion for the cho i c e o f q i s g i ven b y

K
= l -p
r:p-
 22

K
De f in ing s= p , t h e spee d o f l earn ing ( s in c e whe n K ri se s s f a l l s

b e a u s e p < l ) we c a n re wri t e the f irst -order cond i t io n for q a s

( l - s )u ( q , q ) + s ( l- p ) W ' (q ) = 0
1 0
Di f fe ren t i a t ing w i t h re s p e c t t o s y i e l d s

( 1 - S ) \)
11 .:5
- \) + ( 1- p)w I + S ( 1 - p ) w II = 0
1 ds
ds

or

=
ds
( l- s )u (q , q ) + s ( l - p )W" ( q)
11 0

From t he f i r s t -ord er cond i t ion we know W ' ( q) > O a t the q c ho s e n

( be c a u s e u < 0 everywhere ) s o t he n ume ra tor i s p o s it ive . The

1
denom i n a t o r i s nega t ive by t he s e cond order cond i t io n d e f i n i ng
> O and have prov en
the o p t ima l q , so we c a n c o n c lude tha t

Theorem 3 In the ca s e whe re a l l l e a rn i ng o cc urs a f t e r K pe riod s ,

as K i n c rea s e s t he o p t ima l q ua l i ty l ev e l f a l l s , for

a n y g iven in i t i a l r ep u t a t ion .

Co n s e qu e n t l y , t he qual i t y c ho sen appro a c h e s q * monoton i c a l l y from

be l o w a s K fa l l s to 0 .

T h i s compa r a t ive s ta t i c re su l t c a n be s ho wn i n a mo re g en e ra l

s et t i ng . Deno t e the p r e s e n t v a l u e o f p ro f i t s from c hoo s ing q

when the s p e e d o f l e a r n ing i s s b y V ( q , q , s) . Then the f i rs t ­

o
 23

o rd e r condit io n f o r q is V ( q ; q , s ) = 0 a n d t h e c ompar a t ive

q 0
s ta tic s computation give s

v
ds
=- v so > 0 i f a n d o n l y if V > 0 i . e . , i f inc r ea s ed
qq ds qs

s p e ed o f l ea r n in g in c r e a s e s t h e va l ue o f q u a l it y to t h e monopo l is t .

In t h e ca s e whe r e V ( q , q , s )
0
= ( 1-s) V( g , q )
0
+ sW ( g )

this wil l ho l d . I n gene r a l o n e wo u l d n e e d t o lo o k t o s e e how

s en te r ed V to s e e i f V >0.
qs
I t is o n l y s l igh t l y mo r e d if ficu l t to s e e how t h e s e l f-

fu l fil lin g qua l ity l eve l i s in f l ue n c ed b y t h e l e a r n in g s p e ed o f

c o n s ume r s . L et me con s id e r t h e c a s e wh e r e this qua l ity is

unique ; uniqu e n e s s is gua r an t e e d if we a s sume t h e f un c t io n

.

v (q, g) is d e c r e a s ing in g ( so t h e r e ca n o n l y b e a s in g l e
1
root to v ( g , g ) = 0 ) . This a s sumpt ion is e qu iv a l en t t o
1
v q) + v1 2 ( g , g) < 0 for a l l g . The s e l f - fu l fil l in g
11 (q,
qua l ity l eve l , q, i s de f ln ed a s a functio n o f s b y

v (
1
, q, s) = 0. Dif f e r e n t ia t in g with r e sp e c t t o s g ive s

.93.
ds
= v
_ ls
-----
V +V 2
ll l

So a g ain we get > O , und e r t h e a dditio n a l

> 0 e xac t l y wh en v
1s
a s s umpt ion o f v ( g , q) de c l inin g .
1
 24

We l f a r e I mpl i c a t io n s

The a na l y s i s above ind i c a t e s t h a t impe r f e c t i nf o rm at ion w i l l

t e n d to c a u s e a r ed uct io n in the qua l i t y o f produc t s pr o v id e d .

T h i s c a n l e ad to e i th e r a g a in o r l o s s i n soc i a l we l fa r e , u s i n g

the c r i te r io n o f con s umer s s u r p l u s p l u s pro f i t s to judge we l fa r e .

The we l f a r e con seque n c e s o f q ua l i t y c ho i ce b y a monopo l i s tmrler

perfect info rmation are stu::lied. in Spence [1975] and Sheshinski [1976]. The
oooclusion o f t heir analysis is t hat , depending on the s hape of the inverse

d emand c urve i n quant i t y and q ua l i ty , p(x , q ) , a monopo l i s t may

e i th e r unde r - o r ov e r- supp l y q u a l i t y . T h i s i s b e c a u s e the

monopo l i s t co n s i der s the e f f e c t o n the ma r g i n a l b u y e r o f a

c ha nge i n qua l i ty , wh i l e s o c i a l we l fare r equ i r e s l ook ing a t t h e

e f f e c t on t h e ave r ag e buye r . As a r e s ul t , g iven the qu a n t i t y

produced , t h e monopo l i st qu a l i ty i s t o o l o w i f a nd o n l y i f

p <0, i.e. , i f and o n l y i f the ma r g i na l con s ume r ' s e v a l u a t i o n

xq
o f qua l i ty i s l e s s than the average con s umer ' s . ( Se e Spence

[1975] p . 419) . Fu rthermo re , the c hoice of q ua l it y ma y in t e r a c t

'

w i th the seve r i ty o f output r e st r i c t io n i f p r i c e e l a s ti c i t y

i s depe n dent o n q ua l i t y.

Con s i de r t h e we l f a r e c o n s equenc e s o f th e s e l f - fu l f i l l i ng

equ i lib rium qua l i ty c ho i c e . I f q u a l i t y a t t h e pe r f e c t i nforma ­

t io n outcome , q * wa s sub-opt ima l , then t he i n forma t iona l p rob l em s

wou l d e xac erbate t h e we l f a r e l o s s e s . I t i s p e r f e c t l y po s s ib l e ,

however , that q * i s s u pra - o pt ima l in whi c h c a se the l a c k o f

i n f o rma t i on cou l d h e l p ma t t e r s . Thi s is a t ypic a l s e c o nd -b e s t

e xampl e o f two imperfec t ion s c o un t e r a c t ing e a ch o the r .

 25

Th e we l f a r e a n al y s i s a lo n g t h e pa t h t o a s te a d y s t a t e , i . e . ,

wh i l e con sume r s a r e l e a r n i n g the true qual i ty , i s con s id e ra b l y

more comp l ex . F or e xamp l e , some c o n s um e r s may purcha s e the

prod u c t ev en though i t in f a ct h a s l i tt l e va l u e to them , b e c a u s e

they ove re s t ima ted t h e qua l i ty . I t i s c l ea r tha t the j o in t

d i s t r i but ion o f v a l u a t i o n s o f qua l i ty a n d expec t a t i o n s o f

q ua l i ty , and how t his dist r ib u tio n i s upd a ted, is c e n t r a l to

a we l fa r e a n a l y s i s .

Cons ider a spe c i a l c a s e : everyone ha s c ommon m i spe r c ep t i o n s

abou t qua l it y , exp e c t in g R in s t e a d o f q . S uppo s e in i t i a l l y

0

tha t pric e e l a stic ity o f d emand i s una f f e ct ed b y expe c ted

21
qual i ty , s o the f i rm's p r i c e is independent o f reput a t io n .

Then if R > q mo re peop l e wil l b u y the product than wou l d under

0

pe r f e c t in forma t ion. This s e ems l ik e it wou l d re pr e s en t

a we l f ar e gain , s in c e monopo l y o u tput i s too low . Even t ho ugh

some o f tho s e who bought o nl y b e c a u s e t h e y expe cted R >q r e gret

0

havin g done so whe .. th ey o b s erve q , t h e re is a s o c ia l gain

t o the i r having purcha s e d t h e product i f the i r va l u a t io n

exceeds t h e c o st of production.

To s k e t c h o ut thi s e xampl e mo r e f u l l y , a s s ume c o n s t an t

r e t urn to s ca l e wit h unit c o s t f un c t ion c ( q ) . Le t c o n s u me r s

b e de s c r ibed by th eir va lua tio n o f q ua l it y 8 s u ch that type

0 h a s ut i l i ty fun c t io n 8q -p f rom buyin g a unit of qua l it y q

 26

22
a t pr i c e p . The d emand c u rve i s g e n e r a te d by a d i s t r i bu t i on of

0 ' s. S uppos e the monopol i st s e t s p r i c e p , qua l it y q , a n d h a s

i n i t i a l r e pu t a t i on R T h en 0 wi l l buy i f a n d on l y i f 0 R >p .
0 0
•

F r om a s oc i a l we l f a r e point of v i ew , 0 s hou ld buy i f a n d on l y

i f 0q>c ( q ) . Depend i n g on t h e r e l a t ive s i z e of £ig) and E

q Ro

t h e r e ma y b e t oo l it t l e or t oo muc h ou t pu t . S e e F i gur e 5 . It

i s p os s i b l e tha t t h e f ir st b e s t out c ome i s a c h i e ved under

mon opoly and imper f e c t i n f orma t i on i f £lsl = . The r e w i l l of

q
0
c ou r se b e d i s t r i but i ona l c on s equenc e s of t he impe r fe c t i on s , but

those are n ot c on s i d e r e d in our s t andard we l fa r e mea s u r e .

A B

- --- - \

£ig_) E
R q
0

F I GURE 5

Ove r - E s t ima t e s of Qua l ity

 27

p
I f R i s a t A t hen t oo man y c on s ume r s pur c h a s e the p r od u ct .
0
T h i s d oe s n ot imp l y that t h e s i t ua t i on wou l d be improved by

e l im i n a t in g the i n f orma t i on a l p r ob l em, h owe ve r . We mu s t c ompa r e

E £
p u r c ha s e s by 8 R w i t h p u r cha s e s by 8 q . T h e l a tt e r s u f f e r s f r om
0
t h e u s u a l monopoly outpu t r e s t r i c t i on .
E
I f R0 i s a t B , t oo few c on s ume r s purc ha s e und e r imper f ec t

i nf orma t i on , but the imper f e c t in f orma t i on impr ove s we l f a r e .

S o m i l d ove r e s t ima t e s of qua l i t y impr ove we l fa r e ( un d e r the

a s sumpt ion of p r i c e e l a s t i c i t y independent of r eputa t i on ) .

The s it ua t i on i s qu i t e d i f ferent i f R < q . Then the

0
pe s s im i s t i c mi s p e r c ept i on s tend t o reduce sa l e s a n d e x a c e r b a t e

wel f a r e probl ems . Se e F i gure 6 b e l ow .

8
E E
q Ra
FIGURE 6

Und e r - E s t ima t e s of Qua l i ty

Now e ve n fewer s a l e s oc cur w i t h expe c t a t i on s R th a n wou l d

0
und e r p e r f e c t i n f orma t i on ; there i s a n un amb iguou s we l fa r e

l os s .

A l l owing reput a t i on t o i n f l ue n ce wha t pr i c e t he f i rm c h a r ge s

pe rm i t s a l mos t anyt h i n g t o oc c u r . For examp l e , i t cou l d b e

t h e c a s e tha t pe s s imi s t i c expec t a t i on s i nc r e a s e d emand e l a s t i c i ty

and th i s c a u s e p r i c e reduc t i ons wh i c h mor e t ha n a f fe c t the l os s e s

men t i oned a bove due t o s u c h exp e c t a t i on s . T ha t i s , if

 28

( ) <
£lS.L P( R0 ) ( )
< .El9..L , even t h ough R < q , then t h e re a r e we l f a r e
q q o
R0
ga i n s t o h a v i n g m i spe r c ept i on s of R ( he r e p ( q) i s the price
0
23
charged when c on s ume r s a l l be l i eve the qua l i ty i s q ) .

Anoth e r we l fa r e e f fe c t , e xp l or e d i n a n examp l e b e l ow,

i s c a u s e d when t h e f irm ' s pr i c ing d e c i s i on s a r e mad e w i t h r e c og ­

n i t i on that t h e y e f fe c t t h e l ea rn in g p r oc e s s a n d h en c e future

demand . S uppos ing that h i gher s a l e s l e ad s t o mor e r a p i d l ea r n ing

by the ma r k e t about true q ua l i ty, the r e is an i n c en t ive t o c u t

b a c k on s a l e s when r eputa t i on e x c e e d s t r ue q ua l i ty . Thi s e x c e r ­

b a t e s t h e u sua l mon opoly we l f a r e l os se s . Conve r s e ly, whe n

reputat i on i s be l ow t rue qua l it y t h e r e i s a bene f i t i n a dd i t i on

t o stat i c ma rg ina l r evenue f r om mak i n g anot h e r s a l e . Name ly,

there is a mor e r a p i d expa n s i on of d emand due t o the i n c r e a s e d

s peed of l e a rn ing a bou t the t r u e qua l i ty . Thi s e f f e c t l e a d s

t h e monopol i st t o expand output r e l a t ive t o i t s s t a t i c p r of i t

ma ximi z n g l eve l , c a u s i n g a we l f a r e g a in a l ong t h e p a th a s c on s um e r s

l earn. I t i s wor t h n ot ing that the se e f f e c t s a r e n ot sma l l ,

b e c a u s e t h e monopol i s t i s n ot a t the s oc ia l opt imum i n t he i r

absence .
 29

P e r s on a l Le a rnin g : An Ex te n d e d E x ampl e

I n t his s e c t ion I a n a l y z e rno ly pricin g and o u t put d e c is io n s

o ve r t ime a s c on s ume r s l ea rn . The l e a rn in g i s r e s t r ic t e d in two

ways which l imit the gene r a l it y o f th e r e s ul t s : {1 ) Al l cons ume r s

24
b eg in with the s ame exp e c t e d qua l ity , and ( 2 ) l e a r ning o c c u r s

o n l y f rom pe r s on a l experien c e : a given con s ume r l ea rn s no thin g

un t il he trie s the produc t , a t whic h time he l e arns it s t ru e q u a l it y .

Con s ume rs dif fer in t h e ir t a s t e s for qua l it y . A cons ume r o f

25
t ype e h a s util ity f un c t ion e q - p , a s above . Hen c e if a l l c o n s u me r s
26
expe c t qua l ity R , then thos e 8 ' s who buy wil l b e {8 = e p/R} .

te by F ( the n umb e r o f c on s ume r s o f type s 8 e. Le t the range ,

o f 8 ' s lie in the c l o s e d in t e rv a l [ 8L ' 8H] wh ere 0 <8L; c a l l F ( 8H ) =N,

S uppo s e a l l con s umers initia l l y b e l ieve the q u a l it y to b e R , whe n

i n f act it is q . Th e initia l demand c urve i s t h u s s ( p ) = N - F ( p/R ) .

S imi l a r l y , the f u l l y in forme d demand c urve is z ( p ) = N - F ( p/q ) .

I n the diagrams th e s e are d r awn a s l in e a r demand curve s ; th a t c o r ­

r e s ponds to t ak in g F to be t h e uniform dist ribution , bu t is tot a l l y

unne ce s s ary.

Und e r the type of l e a rnin g a s s umed , the pos ition of the c ur r e n t

deman d c urve depends o n l y on wh i ch c on s um e r s h ave previous ly t ried

t h e prod u c t ( and there fore l e arned) . This in turn depeods o n l y o n

the l owe s t pric e p r e vio u s l y ch a rge d .

E a c h pe riod the firm choo s e s a price . It s o b j e c tive i s the p r e ­

s en t va l ue o f pro f i t s . In ge n e r a l the p r i c e c h a r ge d w i l l vary w i th

t i me as mo r e c on s ume r s l e arn the tr ue qua l ity o f the produc t . Th e

pr ice c h a r ged wil l n o t ge n e ra l l y b e the one which ma x i m i ze s s t a tic

 30

prof i t s , b e c a u s e the firm mus t a c c ount f or the e f f e c t o f t hi s p e r i od's

pr i ce on the deman d c urve in future pe r i od s .

S i n ce th e s i t ua t i on i s f undamen t a l ly d i f fe re n t d e pe n d i n g upon

wh e t te r i n i t i a l e xpe c t a t i on s a r e ove r - or unde r - e s t im at e s o f t rue

qua l i ty , I s ep arate the an a ly s i s i n t o two c a s e s : R > q and R < q.

e R
H

e q
H

N
F I GU RE 7

Initia l Re p ut a t i on Exc eed s Tr u e Qua l ity

( CASE I )
 31

Case I :
R > q ( Re f e r / t o F i gu r e 7 )

C a l l the p r i c e c h a n ged i n p e r i od t , P t ·

Lemma 1 : I f fo r s om e T , P T P T- l'

then P t = P T f or a l l t > T .

P r oof :
I f PT P T - 1' th en th e r e i s n o s h i ft i n the d emand c ur ve

due t o the s a l e s made dur i n g p e r i od T. Th i s i s b e c a u s e

n o new p e o pl e t r i e d t he p r od uc t d u r i n g pe r i od T . Con­

s e q uen t ly , i f it was op t i ma l f or t h e f i rm to c h a r ge P T

·fa c i n g t h i s d e mand c ur ve dur i n g p e r i od T , i t i s op t im a l

for it t o d o t h e s ame th i n g a ga in i n p e r i od T+l. •

S i n ce we kn ow th e p r i c in g s equence i s mon ot on i c a l l y d e c r e a s i n g

over t..i.rre e x c ept p os s i b l y for a f i n a l j ump i n p r i c e , it i s n a t u r a l

t o a s k h ow l ow P t ge t s , and wh e r e i t f i n a l l y en d s up.

I t i s a t th i s po i n t n e c e s s ary t o d e s c r i b e wh at th e d emand curve

l ooks l i ke a s a fun c t ion of the l owe s t p r i c e previo us l y c h a r ge d ,

. Den ote t h i s demand c u r ve b y x (p, ). We m u s t d i s t i n gui s h two

s ub c a s e s :

T h e d emand c u r ve x ( p , p) i n th i s c a s e i s s hown i n F i gure 8 . It

i s c on s t r uc t e d a s fo l l ows: T h er e a r e s (p) c on s umers who h ave p re ­

v i ous l y t r i ed the produc t . Sin c e t hey upda t ed f r om R t o q , the i r

d mand i s repre s en te d by the po rt ion of the z(p) c ur ve f r o m x=O t o

x=s ( p) . T h os e wh o h a ve y e t to try the pro duc t h a ve de mand r e pr e s e n t e d

bv s (p) f r om x = s ( p ) to x = N . I t i s add i n g th e s e two d e mand c ur ve s

t oge th e r wh i ch gi ve s x ( p , p).
 , 32

For p > p no one will buy, since all those who previously did
now know better. 8 q < p < p some uninformed consumers will
For
H
buy, namely those who value the product enough to buy but not
enough to have done so at p: eHq < 8R < p. There are just
s(p)-s(p) of these, since s(p) have already learned. For
z-1 {s(p) )<p<eHq some informed consumers will buy and some do not.
Exactly those e s.t. eR >p (informed) and 8q>p will do so. There
are z(p) of these. Also some uninformed buy: those e s.t. 8R<p
and 8R>p. There are s(p)-s(p) of these. Finally, for very low
p (p<z-1 (s(p})) all those who are informed buy,s(p), and some un­
informed do as well: z(p)-s(p) of these. Summing up, we get the
demand curve x(p, p) shown in Figure 8 . Algebraically, for
p > 8Hq we have

p 2 p
A

s(p)-s( p) eHq <p<p

s ( p)-s < f> l + z < p l z - 1 < s < p l l <p<eHq

s (p} p <z -(s
1 (p)).
 33

F I GURE 8

Case I-B . E< 8H9·

Th e de s c r i p t i on o f demand i s s omewh a t s impl e r i n th e c a s e wh e r e

"'

p < e Hq b e c a u s e tho s e wi l l in g to p a y the m o s t a r e now the i n forme d

who va l ue q ua l i t y the mo s t , r a th e r than the un i n forme d who a r e

o ve r l y opt imi s t i c . S e e F i gure 9 b e l ow f o r t h e s hape o f x (p , p) in

this cas e . Aga in th i s curve i s de r i ve d b y a d d e d two o t h e r c urve s

toge th e r h o r i z on t a l l y : the s ( p ) p
c urve f r om s ( ) to N a n d the z ( p )
A
A

c u r ve f rom 0 t o s ( p ) . A l geb ra i c a l l y , when p < 8Hq we ge t

z (p) p 2:. p

x ( p,
p)
=
p
És ( )

z ( p ) + s ( p )
-

-
z
É1 ( s (p)) < p < p

l
s (p) -i s
p < z
É (p))

 34

,..
p

FIGURE 9

Notice that even though both s(p) and z(p) exhibited declining
marginal revenue, x(p, p) need not.
Now we can proceed to analyze the optimal pricing sequence
{pt} chosen by the monopolist. Denote by p* the profit maximizing
price facing the fully informed demand curve z(p). Then
Lemma 2 For some T, Pt p* for all t > T.
 35

Proof: By Lemma 1 , there are two cases to consider. One where

price falls monotonically, and another where price remains
the same after some date. In the latter case, price could
not remain at p > p* because thefirm could gain profits

both in the short- and long-run by chaYging p* instead of

p (charging p forever yields per period profits of
(p-c(q))z(p)<(p* -c(q))z(p* ) and x(p, p) always lies on or
above z(p) so it is feasible to earn (p*-c(q))z(p* ) every
period.) In the case where {pt} declines monotonically,
it must approach some p. Then if p>p* the per period profits
pproach ( -c(q))z(p)< {p*-c(q))z(p*) so approaching
cannot be optimal. •

Lemma 3 : For some t, Pt < p*.

Proof: Suppose not. Then {pt} p* by Lemmas 1 and 2 , and profits
approach (p*-c(q))z(p*). The demand curve approaches x(p, p* )
which looks like Figure 10 below. I have drawn in the as­
sociated marginal revenue curve as well.
 36

FIGURE 10

Since the curve x( p, p*) has a kink at p*, the marginal revenue
is not defined there. But the marginal revenue curve associated
with X (pIp*) is shown as MR on Figure 10. It can be seen that
( p ,p*)
X
the firm can make one period profits in excess of (p*-c(q))z(p*)
by setting p<p* , since MRx( p, p*)>c ( q) for p<p*. In fact, it
could earn as much as the shaded triangle's area in a one-shot ex­
ploitation of consumer's initial misperceptions. Whether it will
want to do this all at once depends on the exact shape of demand
and the discount rate, but the basic point is established: at some
point the price will fall below p* to reap some gains from the
mL. informed. II
 37

/
Theorem 4 When consumers learn only from personal experience
and begin with common overexpectations about product
quality, the monopolist's prices over time fall mono­
tonically, eventually falling below the fully-informed
monopoly price, and then jump back to that price for­
ever after.
Proof: In view of the Lemmas, all that needs to be shown is that
price does eventually return to p* after it has fallen
below it. If not, suppose {pt } p<p*. Then profits
approach (p-c (q))z (p} < ( -c (q))z (p*) and it would be better
for the firm to charge p* forever rather than · II
Case II R<q
The situation is entirely different when consumers are skeptical
about a product's attributes. Now the firm will tend to sell more
than the static monopoly profit maximizing level, because more sales
today shifts out the demand curve tomorrow. While this might repre­
sent a welfare gain, it must be balanced against the fact that con­
sumers are less likely to buy, even if the product is valuable to
them, because they underestimate its value.
The initial demand curve s (p) and the fully-informed demand
curve z (p) are shown below in Figure 11. I have also drawn in
x (p, p ), the demand curve the firm faces when p is the lowest price
previously charged.
 38

'·

"
p

FIGURE 11

Quality Exceeds Initial Reputation

Let p*(R) be the static profit maximizing price facing common

expectations R (i.e. facing s(p)), and p*(q) be the profit maximi­
zing price facing perfect information demand, z(p). Let x*(R) and
x* (q) be the corresponding levels of sales and n*(R), n*(q) the
profit levels. Then there is a simple case which can be fully de­
scribed:
Lemma 4 If x*(R) x*(q) then the monopolist first charges p*(R)
and forever after charges p*(q).
Proof: Observe that the monopolist cannot possibly make more
than n*(q) during any period, and cannot possibly make
more than n*(R) the first period. So if making n*(R)
followed by n *(q) is feasible, it is optimal. If
 x* (R) x* (q( then after making n* (R) the first period
he faces a demand curve such that the quantity-price
combination (p* {q), x* (q)) is feasible. Hence the pro­
posed regime must be optimal. II
While it is possible that x* (R) x* (q) it does not seem to be
the usual case. This is because such a relationship implies that

p* (R) is considerably below p* (q) i. e. the pessimistic expectations

cause much more elastic demand than accurate expectations. If this
inequality does not hold we can prove
Theorem 5 If x* (R) < x* (q) then in the long run the monopolist
will not sell as many units when he faces initial
expectations R<q as he would under perfect information.
His long-run'per period profits will be below their
perfect information level, and there will be an added
welfare loss as a result.
Proof: The analysis is significantly aided by reference to Figure 12
below. There I have drawn the initial and fully informed
profits as a function of sales. That is
n (x; R) = (s-1 (x)-c (q)] x and

n (x; q) = (z-1 (x)-c (q)]x

Since R<q, or equivalently s (p) <z (p) so s-1 (x) < z-1 (x), n (x,R) <
n(x, q). I have drawn the case to which this Theorem refers, namely
x* (R)<x* (q).

The profits obtainable as a function of sales when x is the

mavimum number of sales previously made (alternatively: p=s-1 {x)
is : he minimum price previously charged) is shown in F igure 13 as
TI (>{I X) •
 ')(

Fi gure 13

 "' '

Several points can now be made clear. First of all, the monopo­
list will never sell more than x* (q) or less than x* (R). This is
because either of these actions loses money in both the short- and
long-run. Selling less than x* (R) sacrifices profits this period
and fails to inform very many consumers that the product is better
than they had thought. Selling more than x* (q) sacrifices short-run
profits, whichever n(x, ) the firm is facing, while informing cus-
tomers who the firm will not want to sell to anyway. That is the
key: there is a cost to informing more customers that the product
is better than they had believed. This is done through introductory
offers in this example. I have implicitly assumed that it is impos­

sible to cut price only to new customers; the firm sets one price
each period.
Once x* (q) customers are informed, there is no point in informing
more, because the firm does just as well facing n(x, x* (q)) as facing
n(x; q).
What will the actual sales path look like? Again, the trade­
offs to be made between short-run sacrifices and long-run gains
depend on the discount rate. But it is easy to see that, so long
as the discount rate lies between 0 and 1 , ( 1) at some time the
firm will sell more than x* (R), and ( 2 ) the firm will never sell
as many as x* (q). The firm will not repeatedly sell x* {R) because,
to the first-order, there are no losses from selling a bit more
(TI (x*(R)"I R) = 0 ) but there are real long-run gains (TI (x* (R);q)>O).
X X
 S imilarly, there/ is no point in pushing sales all the way up
to x* (q) because the long-run gains are ln (x*(q); q ) = 0 and
r x
(i.e., there are short-run
"

the short run gains are TI

X
(x* (q} x) < 0
)
losses from expanding output. Consequently, the monopolist will
utl imately sell X E(x* (R), x* (q)) forever after. This provides less
profit than he would obta in under perfect informat ion s ince
n(x*(q); q)>n ( x, x). And it entails a welfare loss s ince the monopol ist
already has restricted output below its soc ially optimal level.
I s hould po int out that this welfare loss is exacerbated by the
inab il ity of the monopolist to prov ide selective discounts to new
customers. If he could do that, he would f ind it prof itable to
iuform more consumers, through int roductory offers, that the true
quality is q .
The results in this example o f personal learning carry over
if learn ing is not immediate in response to use of.the product,
so long as it takes only fin itely long. Slower learning will in­
fluence the opt imal sales (price) path, but the qualitative con-
elusions of Th eorems 4 and 5 carry over.
 43

A Continuous Time Example of Once with For All Quality Choice

/

In this section I move away from studying how individual

'
consumers learn to focus on one aspect of the monopolist S optimal
sales path in the presence of aggregate consumer learning . That
aspect, referred to above in the welfare section, is that when
sales levels influence learning, if reputation exceeds quality
the firm will cut back on sales relative to static profit
maximization. As in the above section, this model takes
quality choice as given and studies the sales path over time.
Denote quality by q, reputation by R, and sales by x. The
inverse demand curve the firm faces depends on reputation,
and is written p(x, R). The firm faces a control problem of
the type discussed in the in troduction, with the restriction
that quality is a once-and-for-all choice. This problem
fits precisely into the framework analyzed in Theorems 1 and
2:

CXl

V(q, R( o )) = rna x l( e-rt [p( x(t), R( t))x(t) - c( x(t ) , q J J dt

X( t) 0
.
s.t R = sx(q-R)
R( o) given.
The analysis here will focus on what the path x(t) looks like.
To avoid other effects, I assume constant returns to scale :
c(q , x) = xc {q).
.
The specification of R is cruc ial to the optima l con trol
problem . The idea behind the equation R = sx(g - R) is that
 44

the speed with which reputation adj usts towards true qual ity
depends positivel y on/ the l evel of sal es . This occurs for
two reasons : (1) A given customer updates his expectations
more completel y, the more experience he has with the product ,

and ( 2 ) the more customers who try the product, the more
learning there wi l l be regarding true qua lity (and the more
subsequent interpersonal communication about the firm's quality).
As we wil l see , this causes the monopolist to cut back on sales

when R>q in order to retard the deterioration of reputat ion .

The reverse effect occurs when R<q. The parameter s measures

the speed with which consumers learn from using the product .

The present value Hami ltonian for the firms contro l

probl em is

H(x , R, A ) = p(x , R)x - xc(q) + A S X(q -R).

Assuming P (x, R) > 0 we know that reputation must have

R
posit ive shadow value , i. e , , A>O. The assumption in Theorem 1

that v2 > o is exact l y that A > O .

Denote margina l revenue by MR (x , R)

(i. e MR(x, R) = +
xp (x , R)) .
p(x , R)
X
Then the necessary conditions for the optimal path inc lude

MR(x, R) - c(q) = ' S(R-q )

xp (x, R) - A S X = r A-
R
We can see from the first equation that MR > c exact l y when R>q.
Assuming d ecl ining MR (MR <0) this mean s that sa l es are cut
X
back re l ative to static maxim i z ation exa c t l y when R>q.
 45

Assuming that MRR> O , i . e , that increased reputation

increases marginal revenue, we know the curve in x-R space
along which MR(x, R) = c is upward sloping, as in Figure 1 4 .
By the first equation above, since A > O we know that sqn(MR-c)=sqn(R-q),
so the optimal regime never enters the shaded regions.

Figure 14

To investigat e the dynamic system induced ir x-R space by

the necessary conditions above, eliminate A and solve for x and
R as functions of x and R. This is done in the A pendix. The
resulting equations of motion are

-sx
MR- c ( q ) r
X = MR (q -R) [ xp
£t + R-q sx J
X x
.

R = sx ( q - R)
 46

When p = 0 , a central case, the x = O curve corresponds to MR=c .

xr
/
This case is drawn in Figure 1 5 below. Since MR-c

Figure 15
Q
A
and R-q of the same sign in the relevant regions, we then get
sqn (x) = sqn (q-R) outside the shaded regions. Unle ss p is
.

xr
a large negative number, the expression in brackets in the x
.

equation above will be positive and we will hav e sqn x = sqn ( q-R )
To complete the solution to th e optimal control problem,
we must see which path is best, starting at a given R { O ) .
Fortunately, it is easy to rule out all paths except the one
leading to the steady state . Begin with the case R( O ) >q .
We know any path which enters the shaded regions from Figure 14
ca nnot be optimal. Consider those path s which lead to the
x= O boundary. These involve closing down with a high re putation .
 47

T h i s c annot b e opt ima l und e r t h e a s sumpt io n t h a t po s i t ive p ro f i t s

/

c a n b e e ar n e d i n t h e s t e a dy - s ta t e , b e c a u s e a f i rm w i t h R> q can

at l e a s t d up l i ca t e the a c t i o n s of a f i rm w i t h reput a t i o n q .

T h i s p rov e s that t h e opt i ma l r e g ime when R { O ) > q invo l ve s fo l lowing

th e s e p a ra tr i x i nto s t e a d y state f ro m wh e r e it i n te r s e c t s

R=R ( 0) .

T h e a n a l y s i s i s muc h t h e same wh en R ( O ) < q . Now we mu s t

r u l e o u t p a t h s wh i c h l ea d o f f to x :oo as q> R . The s e l ea d to

l ar ge l ? s se s a s x grows a n d h e n c e c annot b e o p t i ma l . Aga in

t h e opt i ma l r eg ime goe s t o t he po i n t ( x * , q ) in the F i gu re s .

Th e r e l a t i on s h i p be twe e n s a l e s l eve l s in t h e p r e s e n c e
s
o f l ea rn i n g x * ( R ) , a n d un d e r s t a t i c pro f it max imi z a t io n x ( R )
s
i s dep i c t e d b e l ow . The fact that x * ( R ) > x ( R) i f and o n l y i f

R< q do e s not d e pend on t h e a s s umpt i on that M R > O . Th e we l fa r e

R
c o ns e qu en c e s o f t h i s b e h a v i o r we r e d i s c u s s e d in a p r e v i o u s

sect ion .

x*(rt)

wi th learning

Fi «;ure 16
 48

P roduc t Qua l i ty Cho i c e Ove r T ime : Remark s a n d Gen er a l Re s u l t s

Wh i l e i n some c s e s produ ce r s c hoo s e product a t t r ib u t e s

o n c e and f o r a l l , a s st ud i ed a bove , in o t h e r c a s e s i t i s

po s s ib l e f o r t hem to c hange th e i r qua l i ty over t ime . In t h i s

sect ion , I s tudy t h e b e h a v i o r o f a mo nopo l i st who c a n a l te r

h i s qua l i ty e a c h pe r io d . Thi s i s one po l ar c a s e - no c o s t s

t o c hang ing qua l i t y - wh i l e th e e a r l i e r a na l y s i s i s t h e o th e r ­

l ar g e co s t s to c hang in g qua l i ty so t h a t a o n c e a n d for a l l

c ho i c e mu s t b e mad e .

A c ent r a l i s sue in s tudy in g q ua l it y c ha n g e s o ver t ime

s the fo l l ow i ng : un der what c i rcums t a nc e s ( co s t f un c t i o n s ,

dema n d fun c t ion s , and con s ume r l e ar n in g ) wi l l a monopo l i s t

i i nd i t opt ima l to s e t t l e down to some st ea dy - s t a t e qua l i ty

l e ve l ? T h e a l t e rn a t ive i s to o s c i l l a t e , r ep e a t e d l y runn i ng

up a r e put a t i on and then m i l k i ng i t . I f t h e o sc i l l at i o n s

we re opt ima l , w e mi g h t e xp e c t a l t e rn a t iv e mechani sms i n s te a d

o f r eput a t io n t o ar i s e to c e r t i f y o r c o n t ro l t h e monopo l i s s

product qua l i ty .

F i r st I d i s c u s s t h e n e c e s sa r y c o n d i t i o n s f o r a qua l i t y

l e v e l t o c o n s t i tut e a s t e a dy s t a t e . Aga in w e s e e tha t imp e r ­

f e c t in forma t i o n c a u s e s the monopo l i s t t o redu ce qual i ty .

T h e way i n wh i c h c o n s umer s ' expe ct a t io n forma t io n a f f ec t s

t h e s t e a dy - s t a te qua l i t y l eve l i s an a l y z e d . F i n a l l y I mak e

s ome ob s e r va t i o n s on when t h e s u f f i c i e n t cond i t i o n s wi l l b e

sa t i s f ie d so t h a t s tayin g in o r go i n g t o st e a dy s t a t e i s

opt ima l .
 49

T h e f i r st ma in r e s u l t r e g a r d i n g s t e ad y - s t a t e q u a l i ty l e ve l s

i s t h e a n a l og ue o f Tpeorem 1 in a wor l d whe r e q ua l i ty c a n b e

changed over t ime :

Theorem 6 . So l ong a s repu t a t ion h a s po s i t i ve va l ue , a n y s te a d y -

s t a t e q ua l i t y l ev e l mu s t l ie b e low t h e pe r fe c t

i n f o rma t i on q ua l i t y l e ve l .

Proo f : One po s s ibl e d e v i a t ion from s t e ady s t a t e i s a o n c e a n d

for a l l change in q ua l i ty . S in c e , by Theorem 1 , for any

qu a l i ty l eve l a t l e a st a s g r e a t a s q * ( the p e r f e c t

in forma t ion q ua l ity l ev e l ) i t i s pre f e rab l e to c u t

q ua l i ty ( fo reve r ) r a t h e r t h a n to ma inta i n q ua l i t y ,

such qu a l i ty l eve l s c annot be s te a d y s t a t e s . •

To anal y z e mo r e c a r e f u l l y wha t t he s t e a d y s t a t e q ua l i ty

l eve l wi l l be , con s i d e r t h e f o l lowing s e t up : E ac h y e a r t

t h e f i rm c a n c hoo s e a q ua l i ty l eve l q . C o n s ume r expe c ta t i o n s

t
a t t h e be g i nni ng o f t h e y e a r a r e s umma r i z ed in the reputa t i o n ,

R . T h e f i rm can vary p r i c e s ( an sa l e s ) t h ro ughout t h e y e ar

t
i f i t d e s i re s . T h e r e s u l t i n g pro f i t s d ur i n g t h e y e a r a r e
. 28
wr ' t t en V ( q t R ) B y t he e n d o : t h e y e a r , repu ta t i on w i l l
t
adj ust to R in a mann e r y e t to b e s p e c i f i ed . T h e f i rm ' s
t+l
o b j e c t i ve f un c t ion i s
00
t
w = L: p V (g ' R )
t t
t= O
where R i s g i ven .
0

Theorem 7 . Let W ( g ) = V { g , q) be concave in q . If R = yR + { 1-y) q

t t_1 t­
s
whe r e O < y < l , t h en an} s t e ady - s t a t e qu a l i ty l eve l q

mu s t be l e s s t han t h e f u l l i n f o rma t i o n p ro f i t ­
 so

max im i z in g q ua l i t y l eve l q * , a n d g r e a t e r than the l owe s t

s e l f - f u l f i l i ng qua l i ty l eve l q ( th i s q * and q re f e r s

to t h e g iven V f un c t i o n ) .

s
Proo f : S uppo s e the f i rm i s i n s t e a dy - s t a t e a t qua l it y q . Con­

s id e r a sma l l o n e t ime i nc r ea s e i n q dur in g p e r i od 0 ,

s
f o l l owed by a r e tur n to q forever . The e f f e c t on the

str eam o f pro f i t s i s

dW - v s s t s s dR t
(q q ) + p V 2 (q q )
1..
_

dq 1 I I dq
0 t= l 0

= yR + ( 1-y ) q so
0 0

dR
l = 1-y .
dg
0
And R = y R + ( 1 - y ) q so

2 l l

dR
l = y ( l-y) .
y
dg
o
dR
t t- 1
In g en e r a l = y
( 1-y) , t> l .

dq
0

So
00

s s s s t t l-y
= V (q I q ) + V 2 (q , q ) [ L p y ( )
l t= l y --

s
If q is to be a steady- s tate, it must be the c a se
W
d s
that d = 0 at q=q .
qo
Therefore we get
 51

or

p ( l -y )
s s s s
v (q , q ) + 1 - py V2 ( q I q ) = o.
l /

F o r q q * we know v 1 ( q , q ) + V 2 ( q , q ) < 0 s i nc e W ( q ) i s c on cave .

p ( 1-y) p ( l-y)
S in c e v > o a nd 0 < 1 - p y < 1 we t h e n know v ( q , q ) + 1 - p y v (q,q} <O
2 1 2
for q q * . Th e r e f o r e no qua l i t y l eve l a t o r a bove q * c a n q ua l i f y

a s a s t e ady s t a t e . ( A s we k now f rom Th eorem 6 )

N e x t con s i de r qua l i ty l eve l s b e low t h e lowe s t s e l f - f i l l in g

q ua l i ty l eve l q . T h e y k e y t h i n g about q i s t h a t f o r q < q we h a ve

p ( l-y)
v ( q , q) > O . Con s e qu ent l y w e a l so have v ( q , q ) + 1 - p y v ( q , q ) > O ,
1 1 2
so t he s e qua l i t i e s c a nno t be s t e a dy - s t a te s .

•
 52

S i nc e t h e i nc en t ive t o c u t q ua l i ty i s g r e a t e r whe n d e t e c t io n

i s d e l ay ed , there i s l owe r q u a l i t y i n s tea d y st a t e und e r s uc h

c i rcum s t a nc e s . T h e s e l f - f u l f i l l in g qua l i ty , q , i s t he o n e w h i c h

wou l d b e c ho s en i f t he re wer e no future a f t e r t hi s y e a r . I f there

w e r e no l e a r n ing abou t q u a l i ty unt i l t h e y e a r wa s ove r , =0 ( Ak e r lo f )

In g e n e r a l , howev e r , even i gno r ing f u t u r e y ea r s i t doe s not pa y

to produce th e lowe s t po s s ib l e q ua l i t y : t h i n k i n g o f a u tomo b i l e

model yea r s a s a n e xamp l e , wo rd may g e t o u t wi t h i n t h e y ea r t h a t

the c a r p e r f o rm s poo r l y . S in c e future y ea r s do ma t te r , q ua l i t y

c h o s e n wi l l e x c e e d q ) h i g he r future r ep u t a t i o n i s o f p o s i t iv e va l u e .

Lt q t h e re i s no l o s s , up to t h e f i r s t -order , o f a n i n c r e a s e i n

q ua l i ty , and t h e r e a r e r e a l g a i n s i n t h e future .

Not i c e that a s the i n t e r e s t r a t e a p p ro a c h e s 0 , i e . , a s p l ,

the s o l u t i o n approa c h e s q * . Al so , a s p O so t h a t future

ye ar s ma t te r v e r y l i t t l e a t a l l , th e s t e a d y - s t a t e qua l i ty leve l
 53

goe s t o q. I n f a ct , i t i s po s s i b l e t o s e e how t h e speed w i t h

whi c h con s ume r s upda t e the i r e xpe cta t i o n s i n f l ue nc e s t h e s t e a dy ­

/

s t a t e qua l i ty l eve l :

Theorem 8 . When a s teady s t a t e q ua l i t y leve l e x i s t s i n the c o n t e x t

of Theo r em 7 , i t i s h i g h e r ( 1 ) t h e h i gh e r i s the

we i g h t p l a ced o n r e ce n t q ua l i ty by con sumer s i n

f o rming r e put a t i o n and (2) t h e h ig h e r the d i s co un t

29
fa ctor ( i . e . , t h e l owe r t h e i n te r e s t r a t e ) .

Proof : The s t e ady- s t a t e qua l i ty l e ve l , d e f in ed by ( * ) above ,

p ( l -y )
depe n d s o n the f a c t o r s = l - p y O< s< l . Ob s erve tha t

< 0 and > 0. So h i gh e r d i s count f a c to r and mor e

we i g ht o n qua l i t y in reputa t io n fo rma t io n c a u s e s t o r i s e .

s
Al l t h a t n e e d s t o b e s hown , t h e re fore , i s t ha t > 0. i
( * ) wi t h r e sp e c t to s y i e l d i n g
D i f f e rent i a te
s
dq
( V + V 2 + sV 1 + sv 2 )
ll l 2 2 ds + v = 0 or
2
s
d
_3_
ds

s
If q i s a s t eady- s t a t e , the d enominator i s n e ga t ive by
s
the s e cond order c ond i t i o n s . Therefore > 0 s ince;
v > 0 and the Theorem i s proven .
2

Examp l e : Suppo s e the f i rm produc e s o n e pr o j e c t e a c h p e r iod so

tha t sa l e s are not a con t r o l va r i a bl e . T h i s wou l d a p p l y

t o a l awyer , f o r e x amp l e , who doe s one c a s e eac h p e r iod

and can choo s e how hard to wo r k on the ca s e . Le t t h e

pr i c e he can e a r n fo r one p ro j e c t be a f un c t i on o f h i s

r eputa t i on n ( R) . Le t t h e co s t o f p rod u c i n g a pro j ec t

o f qua l i t y q b e c ( q ) . A s sume p ' > O , p " < O , c ' > O , c " > O .
 54

Unde r p e r f e c t i n fo r ma t io n t h e s e l l e r wou l d se q to
/

max imi z e p (q) - c (q)

q

S o q * sat i s f i e s P 1 ( q * ) = C 1 ( q* ) . Und e r impe r f e c t

i n f o rma t ion w e h a v e V ( q , R ) = p ( R ) - c ( q ) so t h e f ir st ­
p ( l-y)
o r d e r cond i t io n i s - c 1 ( q ) + 1-py p 1 ( q) = 0

whe n R = yR + ( 1-y) q • W r i t in g t h i s a s
t t-l t_ 1
£9.
C I ( q) = sp 1 ( q) , O < s < 1 we h a ve c " ( q ) d s = p 1 ( q ) + sp " ( q ) d s
p I (q)
or : > 0
C II ( q ) - s p II ( q )
•

ds

I n t h e c a s e y = O so R = q we get C 1 ( q ) = PP 1 ( q) .
t t_ 1
s e e F i gu r e 1 7 b e l o w .

q,s 'l
FIGURE 17
 55

I n g en e r a l th er e i s n o guarantee that a n o pt i ma l s te a d y ­

s t a t e qua l i ty l e v e l exi st s . I f V ( q , R) i s c on c a ve , then we c a n

b e s u r e t h a t i t doe s . P art o f t h e prob l em , however , i s tha t

the spe c i f i ca t io n o f how re pu t a t i on chan g e s wh i c h wa s u sed i n

The o r ems 7 a n d 8 d o e s n o t i nc o r po r a t e a s a l e s t e rm i n repu t a t io n

a d j u s tmen t . I f the f i rm c a n bu i l d up a good reput a t ion b y s e l l i n g

o n e g o o d i t em a n d y e t s e l l ma ny poor i t em s b e f o r e r eputa t i on

d im in i sh e s , i t w i l l nev er b e o p t i ma l to s t a y i n s t e a d y s t a t e .

I n gen e ra l , o sc i l l a t ion s wi l l b e d e s ira b l e wh enever a

f i rm c a rt e arn mo re i n the p ro c e s s o f r unn i n g down i t s r epu ta t ion

than i t c o st to bu i l d i t up . Th i s wi l l depend on the pr e c i s e

mechan i sm b y wh i ch r e pu t a t i o n i s f o rmed . I con s i d e r o ne

p l a u s i b l e spe c i f i ca t ion of reput a t ion a d j u s tmen t i n t h e mod e l

b e l ow . An open prob l em i s how t h e a c t i vi t i e s o f i nd i v i d u a l

c on s ume r s - bo th in format i on g a th e r i n g a n d expe c t a t i on a d j u s tmen t ­

i n f l u en c e t h e w a y in wh i ch r eputa t i o n move s . A mod e l wh i ch

a ddr e s s e d that prob l em wo u l d b e a b l e t o t r a c e through the imp a c t

o f impro ved cornrn, . i c a t ion o r i n f orma t io n ga th e r i ng o n the

f irm ' s s t eady- s t a te qua l i t y l e ve l .

 56

A C on t i nu o u s T im e M o d e 1 w i th Variable Q.lality _

I n t h i s s e c t i o n I an a ly z e ano t h e r mo d e l o f t h e f o rm d i s -

cu s s e d i n t h e i n t r o du c t i o n . I t i s v e ry s imi l ar t o t he e ar l i e r

c on t i nu ous t im e m o d e l , w i t h t he a dd i t i on o f q ua l i ty as a c ont r o l

variable . F o rm a l ly , t h e m o n op o l i s t f a c e s t h e c o n t r o l p r ob l em

00

( -rt
m ax .) e [ B (x , R ) - c (x , q )] d t
q (t) , X (t ) 0

s .t. R = sx (q -R) .
R (o ) g iven .

T h e func t i on B (x , R ) i s t h e b en e f i t o r r ev enu e fun c t i o n ; i t is

xp , R ) wh e r e p , R ) i s t h e i nv e r s e demand cu rv e . Th e r e p ut a t i o n

a dj u s tm en t e qu a t i o n w a s d i s c u s s e d ab ov e i n t h e o n c e a n d f o r a l l

qua l i ty cho i c e m o d e l .

Th e cur r en t v a l u e Ham i l t on i an i s H (x , q , R , A ) = B (x , R ) - C (x , q ) +

A SX ( q - R ) . The n e c e s s a ry c ond i t i o n s f o r t h e o p t im a l T eg i m e i n c lu d e

(1) Bx (x , R ) -
c (x , q ) + A S (q- R) = 0
X

(2) - cq (x , q ) + A SX = 0
.

(3) B (x , R ) - A SX = r A A
R
(4 ) R = sx (q - R )

I c o ns i de r t h e c a s e w i th c o ns t a n t r e tu r n s t o s c a l e s o

c (x , q ) = x c ( q ) . Th e n c (x , q ) = x c ' ( q ) an d ( 2 ) g ive s us
q
c ' (q ) = A s .

We c a n t hu s e l i m i n a t e b o t h A and \ t o g e t
.

Th e r e f o r e c " ( q ) q = A S .
.
 57

(5) Bx ( x , r ) - c (q ) t c ' (q) ( q-R) = 0

"
r 1
·

(6) B ( x , R) - xc ' (q) = SC ' (q) - s c ( q ) q

R

Us (5) and ( 6 ) , i t i s e a s y to wri t e down t h e e q ua t i o n s whi c h

mu s t b e s a t i s f i ed by a s tea d y s t a t e quantity - qua l i t y pa i r ( x , q ) .

In s te a d y s t a t e q= R , R= O a n d we g e t

(7) B ( x, q) = c {q)
X
r
(8) . B q ( x , q ) = XC I (q) + s c I (q) •

I t i s ve r y in s t r u c t ive to c omp a r e t h e s e s t ea d y - s t a t e

e qu a t i on s to t h e f ir s t -o r d e r con d i t i o n s f o r q ua n t i ty a n d qual i t y

c ho i ce un d e r p e r f e c t i n f o rma t io n . W i th p e r f e c t i n f o rma t io n t h e

monopo l i s t s o l ve s

( ** ) ma x B (x, q) - xc ( q )
q,x

wi th f i r s t -order cond i t i o n s

(9) P (x, q) = c ( q)
X

(10) B ( X I q ) = XC I (q)
q

Theo r em 9 . S o l ong a s t h e r e i s a s o l ut io n t o t h e p e r f e c t

i n forma t i on pro b l em ( * * ) , t h e s te ady- s t a te q ua l i ty

l eve l i s s t r i c t l y l owe r and i s monoton i c a l l y

r
d e c re a s in g in s . T h a t i s , a s the spe e d o f l ea r n i n g

fa l l s or t h e i n t e r e s t ra t e r i s e s t he f i rm ' s s t e a d y

s t a t e qua l i t y l eve l d e c l in e s .

P re ) f : Tota l l y d i f fe re n t i a t e ( 7 ) and (8) to g e t

(ll) B dX + B dq = c • d q
XX xq
 58

r r
( 12 ) B dx + B dq = c ' d x + x c " d q + { i ) c " d q + c ' d ( i )
xq qq
/

Rewr i ti n g , we have

B
XX
B -c ' 0
xq

r
= C Id( )
8
-

B -c '
xq

T h e s e c on d- o r d e r cond i t io n f o r t h e p e r f e c t i n f o rma t i on p ro b l em

i s that the ma t r i x

B B -c '
XX xq

B -c ' B -c"x

xq qq

be nega t iv e de f i n ite . T h e add i t i o n o f t h e term ( £) c " w i l l pres erve

s
nega t iv e d e f i n itene s s s in c e c " > O . C a l l t he new ma t r ix A . So

I A I >0 .
Now , u s i ng C r amer ' s Rul e ,

= B XX c '
< 0 s in c e B <0 .

d ( i)

TAl XX
dx

We ca nnot g e n e r a l l y s i g n d ( i ) b e c a u s e the r e l a t i o n s h i p

b e twe e n x and R i s amb i gu o u s even i n t h e p e r f e c t in f o rma t i on

case . But by ( 7) and ( 9 ) we c a n s e e that in s t e a dy - s t a t e t h e

s a l e s a s so c i a t e d w i th a g iven qua l i t y a re t he same that wou l d

pr eva i l un d e r p e r f e c t i n fo rma t io n .

T h e r e s u l t o f Theo rem 9 i s a ve ry i n t u i t ive : for high

spe e d s o f l ea r n ing th e f i rm ' s inc en t i v e s l oo k mu c h l ik e they

 59

d o und e r p e r f e c t i n forma t io n . L i k ewi s e , for l ow in t e r e s t r a t e s

t h e s ho r t - run g a i n s f r om c u t t i n g q ua l i t y a r e r e l a t i v e l y un impor t a n t

so q ua l i t y i n s t ea dy - s ta t e a g a in n e a r s i t s per f e c t in forma t i on

l ev e l .

Th i s l ea v e s u s o n l y w i th t h e que s t i o n o f whe th e r i t i s in

fact o p t ima l f o r the f i rm t o go to a s t e ady s t a t e . On e v e r y

spe c i a l c a s e whe r e i t i s opt ima l to do s o i s wh en c ( q ) = c q . In

th i s c a se w e g e t a bang-bang so l u t ion t o t h e c o n t r o l prob l em .

I n s uc h a c a s e o n e wo u l d wan t to put a n uppe r bound on q a s we l l

s
a s a l o we r bound o f 0 . Ca l l q the qua l i t y l ev e l wh i c h s a t i s f i e s
s
the s t eady- s t a t e e qua t i on s . Then i f R ( O ) > q t h e f i rm s e t s q= O
s s
unt i l R= q a t whi c h po i n t i t s e t s q=q f o reve r . Th e oppo s i t e
s
r e s u l t o c c u r s whe n R ( O ) < q . B y t h e t h eo rems a bo ve we know
s
q <q* .

Th e ban g-bang examp l e d o e s not g i ve a ny i n s ight i n to t h e

dynami c s . F o r t un a t e l y , s i n c e t h ere i s on l y o n e s t a t e va r i a bl e

i t i s po s s ib l e to e l iminate o s c i l l a t io n s a nd b e g u a r an t e ed

o f conv e r g e n c e to some s tea d y - s t a te . T h e k e y i s that t h e

H ami l to n i an i s j o in t l y conc ave i n t h e c o n t r o l va r i a bl e s :

H = B
XX XX

H = -xc " { q) and

qq

H = - c ' { q) + AS = 0 by ( 2 ) .
xq

So th e s econd d e r iva t i v e ma t r i x i s
 60

B 0
XX
/

0 - xc " ( q ) whi c h i s ne ga t ive

d e f in i t e . A s a r e s ul t , the opt ima l c o ntro l s a r e cont inuous

f unc t i o n s o f t h e s t a te v a r i a bl e : x * ( R} and q * ( R } .

Now there a r e two f un d ament a l l y d i f fe r e n t po s s i b i l i t i e s .

Ca se A i s when q * ( O ) > O ; C a s e B i s when q * ( O ) = O .

Case A

Con s i d e r the fun ct ion q * ( R) . We know that f o r R l a r g e

e n o u g h i t c annot p a y t o keep bu i l d ing up reputa t io n . Tha t

30
i s , fo r R l a r g e q * ( R ) < R . S in c e q * ( R) i s a c o n t inuous f un c t i o n ,

an d i t s t a r t s above q = R a t R= O a n d even tua l l y f a l l s b e l ow

s
q= R , a t some po int i t mu s t c ro s s q= R . That i s , for some q ,
s s
q * ( q ) =q . See Figure 1 8 .

ttJt( o)

'iS FI GU RE 1 8 ,*
s
CASE A : St e ad y - State a t q
 61

s
The steady- s t a t e a t q i s stable . As s um in g the p e r f e c t

i nforma t i on p ro bl em h a s a un ique s o l u t ion , the r e wil l be o n l y

one s uch s teady s t a t e . Theor em 9 ha s a n a l y z ed the po s i t io n

o f the s t e ady s t a t e .

Case B q* ( O ) =O

I t i s po s s ib l e th a t i t i s a lwa ys o p t ima l for t he f i rm to

run down i t s r ep u t a t i on . Thi s corre spond s to F i g u r e 1 9 b e l ow .

F I GURE 1 9

Ca se B: q* ( O ) = 0

Th i s woul d o c c u r i f l earn i n g wa s v e ry s low or the i n t er e s t

/

rate h i gh . The f i rm s imp ly run s down i ts reput a t i o n , eventua l ly

go i ng out o f bus i ne s s when R= O ( a s suming p ( x , O ) = 0 f o r a l l x }

I t i s i n t e r e s t i n g to not e that j u s t bec a u s e a f i rm i s i n t e nd ing

to run down i t s reputat ion doe s not imp l y it wil l neces sari ly

s e t q= O wh i le d o i n g so . A f t e r a l l , s in c e c ( q ) i s convex there

i s l i t t l e co s t t o imp ro v i ng qual i ty s l i ght l y from 0 , a nd i t

may s i gn i f i c a n t ly r e tard the d e t e r io r a t i o n o f reput a t i on .

 63

Con c l u s io n s

Th i s paper ha s / p r e sented a numbe r o f mod e l s wh i c h d e s c r ibe

a monopo l i s t ' beha v i o r whe n con s ume r s c a nn o t obs e rve a l l t h e

re l evant a t tr ibute s o f h i s produ c t p r i o r to purcha s e . In a

v e r y g e n e ra l s e t t i n g i t h a s b e e n s hown tha t the qua l i ty the

f i rm choo s e s to p roduc e is l ower under s u c h c i r c um s t a n c e s than

in a pe r f e c t i n forma t i on s e t t i n g . The we l fa r e con s equenc e s o f

t h i s reduc t io n i n qua l i ty are g e n e ra l ly amb i guo us be c a u s e we

a r e i n a s econd b e s t wor l d due to monopo l y power .

The o u t s tand i n g i s sue s a re many : How i s product qua l i ty

cho i c e re la ted to the i n fo rma t ion g a th e r i ng a c t i v i t i e s o f

i nd ividua l con s ume r s o r g e n e r a l ly the i n forma t i o n f lows i n the

mar ke t pl a c e ? D i r e c t l y r e l ated to t h i s is the que s t ion o f und e r

wha t c i rcums ta nc e s a f i rm wou l d f i n d i t pro f i t ab l e t o a l t e rna t e l y

r un u p and t hen m i l k i t s reputa t i o n . F inal ly , f i rms themse l v e s

e ng a g e i n a ho s t o f i n fo rma t i on - p rovid i n g a c t i v i t ie s . There

i s no rea son they wi l l provide i n fo rma t io n wi th s o c i a l l y

d e s i ra b l e content o r forma t . The r e l a t i on betwe en advert i s in g

a n d qua l i ty mu s t be exp l o red i n an i n f o rma t i o n envi ronmen t

o f t h e type p r e s e n t ed i n th i s paper .

The des i rabi l i ty o f publ i c i n f o rma t ion prov i s ion c a n be

31
e va l ua ted u s i n g the r e su l t s above . Improved i n fo rma t i on

i n c r ea s e s the speed o f l ea r n i n g by con sume r s ; the e f f e c t th i s

ha s o n qua l i ty has been t r e a ted above . Equ i pped with th i s

r e l a t i o n s h i p and an e s t imate o f con sume r s ' p r e f e ren c e s over

qua l i t y as we l l as qua n t i t y , a we l f a r e an a ly s i s of i n forma t io n

p ro v i s ion c a n be made .
 64

Note s

1 . See , f o r exampl e , " S a lop- S t i g l i t z , o r Wi lde- Schwa r t z .

2 . S e e Lanc a s te r [ 1 9 7 5 ] . Thi s que s t ion i s o n l y i n te r e s ti n g

whe n the v a r i ety o f p rodu c t s i s l im i te d by t h e ex i s te n c e o f

some f ix e d c o s t s .

3 . Wi th o n e d imen s iona l qua l i ty q , th e u t i l i ty d e rived f r om

c o n s um i ng one u n i t c an be wr i t te n a s e q whe r e e me a s u r e s

t h e i n t en s i ty o f p r e f e rence f o r qua l i ty . So a n y two

c o n s ume r s e 1 and e wou l d ag r e e that h igher q w a s be tter .

2
f 8 l > e 2 , con sumer 1 wou ld be w i l l i ng to pay mor e f o r
n cr e a s ed qua l i ty than consumer 2 .

Wi th two ( or mor e ) a ttr i bu te s , consume r s can d i s ag r e e wh i c h

o f two produ c t s i s pr e f e r r e d : one may p re f er t h e r e s taur ant
w i th good f ood but s l ow s e r v i c e ( se n s i t ive t a s te but not i n
a r u s h ) wh i l e anoth e r may p r e f e r a f a s t food o p t i o n .

4 . Imag ine tha t there i s a m i n ima l qu a l i ty b e l ow wh i c h con sum e r s

c an te l l t h e product i s shoddy .

5 . See Darby and Karni and Ne l son f o r a f u l l e r d i s c u s s io n o f

the s e no t i o n s .

6 . A tre a tment a l o ng the s e l i ne s i s g iven i n Gro s sman , Kih l s trom ,

and M i rm a n . ·

7 . And i n f ac t the mod e l s men t ioned tak e p r i c e a s f ixed and

exogogenou s .

8. I n tere s t i ng l y enoug h , th ere do e s ex i s t a t l e a s t o ne r e gu l ar

pub l i c a t io n prov i d i ng i nf o rma tion on the l o c a l l ev e l abou t
services : Wash ington Checkboo k .

I t i s y e t not po s s ib l e to prov ide such a s erv i c e w i thout

ou t s ide fund i ng , however , du e to pub l i c goods p r ob l em s
di s c u s sed be low .

9 . For v e r y l ow qu a l i t ie s m a l p r a c t i c e s u i t s may become r e l evan t .

I n mo s t ca s e s , howev e r , con sume r s r e spond to l ow qua l i ty by
s imp l y no t buy i ng the produ c t ag a i n or they c an l ower the i r
r e serva t io n p r i c e f o r i t . I ignore produ c t l i ab i l i ty be low .

10 . I n a f u l ly g e n er a l formu l a t ion one wou l d c e r ta i n l y want to

i nc lude adve r t i s i ng .

11 . Al though they have a compu t a t i o n to s how the f i rm ' s be s t

c ho i c e o f qu a l i ty , I be l i eve i t i s i n e r ro r . Th i s i s
b e c au s e they ignore the e f f ec t qu a l i ty h a s o n the in f l ow
o f n ew con sume r s throug h i t s e f f e c t on m a r k e t s h a r e .
 65

12 . T h i s i s i n c on tr a s t to the approach w i th x ( t ) a s the s ta te

v a r i a bl e . Fo l l owing the ru l e : " Swi tch br and s i f the pro­
d u c t f a i l s " i s a u s e fu l ru l e of thumb but hardly optima l .

13 . S e e , for exampl e , B e t tm a n [ ] or the cons umer behavior o r

mar k e t ing l i t e r a tur e .

14 . I am a s s um i n g that under impe r fe c t i n forma t io n c o n s ume r s

a c t a s though the i r expe c t a t i o n s wer e known w i t h c e r ta i n t y .
Tha t i s , Ro i s a po i n t exp e c ta t i o n .

'15 . I have r u l e d out the po s s i b i l i ty that the f irm a lwa y s

want t o p roduc e m i n ima l qua l i t y by a s sum i n g V ( O , R ) = 0 .
Th i s requ i r e s some ab i l i t y o f c o n s ume r s to o b s e rve 0
qua l i ty i f i t i s very l ow . Wi thout t h i s a s sumpt ion i t
could wel l be the ca s e that t h e c ho i c e f o r t he f i rm i s
" m i n imal qua l i t y " , wha t ever t ha t may mean . Thi s wou l d ho l d
f o r a t tr i bu t e s whi c h a r e comp l et e l y uno b s e r ab l e ( e . g . ,
automobi l e s a f e ty f e a tu r e s i f no o n e too k t he t r o ub l e
t o run t e s t s o n n e w mod e l s ) . Theorem 2 s t i l l ho l d s wi thout
a s s um i ng that V ( o , R · ) =o s o l o ng as v 1 ( o , o } > o .

16 . I n d i vi du a l con s ume r s c a nnot " g ame " w i th the f ir m by c ha n g i n g

the i r expe c t a t i o n s beca u s e the r e a r e man y con s ume r s so a n y
o n e con s um e r cannot i n f l u e n c e the f i rm ' s cho i c e .

1 7 . Such soph i s t i c a tion wou l d not be l og i c a l l y i ncon s i s te n t w i th

the type o f l e a r n i n g embo d i ed i n V , but i s imp l au s ib l e .

18 . Thi s s to r y a l so g o e s through i f F i g u r e 3 i s r e l evan t , but the

qua l i ty u l t ima t e l y produced then d epends on q .

19 . Obvio u s l y , t h i s i s a very spec i a l type o f l ea r n i n g . I t may

appl y , f o r examp l e , to automob i l e r epa i r r e cord s or o t h e r
examp l e s o f durab i l i ty .

20 . Th i s i s o f cour s e a very spec i a l c a s e o f l ea r n i n g . I t may

we l l ho l d fo r a t t r ibut e s whi ch no one c a n o b s erve for some
t ime ( e . g . , durabi l i t y ) but wh i c h then be come p ub l i c
knowl e d g e , p e rha p s through pub l i c a t i o n .

21. Thi s i s r e l a x e d be l ow.

2 2 . I s t ick to { 0 , 1 } dema n d s fo r s imp l i c i t y .

2 3 . P e rm i t t i n g d i v e r s e expe c t a t i o n among c o n s ume r s f u r t he r

comp l i ca t e d t h e a n a l J s i s a n d c a n l e a d t o we l f a r e g a i n s o r
l o s s e s f rom t h e mi s in forma t ion a s we l l .

24 . A s d i s c u s sed above , i n g e n e ra l , the j o in t d i s tr ibution o f

expe c ta t i o n s and va l u a t i o n s o f qua l i ty d e t ermine dema n d .
The upd a t ing wh i c h o c c ur s a s c o rr s um e r s buy and l earn in
tha t mo r e g e n e r a l c on t ext i s co n s i d e r a b l y mo r e compl ex .
I ma ke some r ema r k s abo u t i t b e l ow .

25 Aga i n , co n s ume r s have { 0 , 1 } d emand s . Th i s i s not e s se n t i a l

f o r the a n a l y s i s , but prov i d e s a sub s ta n t i a l s imp l i f i c a t i o n .

 66
/

26. I t r e a t con s umer s a s hav ing po i n t e s t im a t e s o f qua l it y ,

thereby a s s um i n g away t he po s s ib i l i t y o f the c o n s ume r ' s
buy i n g a p roduct he exp e c t s wi l l not be wo rth the pr i c e
i n o r d e r to l e a r n about i t s qua l i ty .

27. As in Ca s e I , the monopo l i st mu s t eventua l l y s e t t l e down

to a s t e a dy- s t a te l ev e l o f s a l e s .

28 . So thi s V ( q , R ) fun c t ion i s a t r un c a t e d ve r s ion o f the o n e

u s e d in the once and for a l l qua l i ty c ho i c e probl em .
L

29. In g en e r a l , I c a n p erm i t R = t A q Th i s h a s no e f f e c t

t h e qua l i ta tive r e su l t s t k=O k t-k .

k
The f a c tor t p A i s the r e l ev a n t s p e e d o f ex p e c t a t i o n f o rm a t i o n .
k

30. I t c a nnot pay t o c o n t in u a l l y bu i l d up h i g he r a n d h i g he r

reputa t ion . H e r e i s why : suppo s e q * ( R) > R for a l l R .
Then R ( t ) wou l d be s t r i c t l y i n c r ea s ing . Bu t when R> q *
-
( th e pe r fe c t i n fo rma t ion l ev e l o f qua l i ty ) we have
B ( x , R) - c ( x , R) < B (x , q * ) - c ( x , q* ) .
Furthermo re , i f q > R
B ( x , R ) - c ( x , q ) < B ( x , R ) - c ( x , R) .
There f o r e , So long a s q > R> q * we know the f low o f p ro f i t s
i s l e s s than the pe r fe c t in f o rma t i on p ro f i t f l ow . B u t whe n
R> q * t h e f i rm cou l d a c h i ev e a t l e a s t B ( x , q * ) - c ( x , q * )
foreve r . H e n c e cont inua l bu i l d in g o f reputa t i o n c annot
be o p t ima l .

31. The a t tr a c t i vene s s A min imum qua l i ty s t a nda r d s , howeve r ,

shou l d not be t r e a t e d i n a monopo l y mod e l . T h i s i s b ec a u s e
on l y o n e qua l i t y i s pro v ided , so the po l i cyrna k e r cou l d
j u s t s pe c i fy the s oc i a l l y opt ima l qu a l i ty . I n stead ,
qua l i ty s tandard s mu st f a c e the i s sue o f d i ve r s i ty o f
p r e f e r e n c e s i n t h e p r e s en c e o f many suppl i er s . See
S ha p i ro [ 1 9 8 0 ] .
 67

Append i x

We beg in f r om

(A. l ) MR ( x , R ) - c (q) =
A S ( R - q)

( A. 2 )

sx ( q - R )
.

(A. 3) R =

F r om A. l , when q R

=
MR ( x , R) - c ( g)
(A. 4) 7\
s ( R - q)
.

q)
.

(R MR X + MR R ] - [ MR - c ] R
1 X
-
.
R
So A = -

s (R - q }

S ub s t i t u t i n g for A in t o ( A . 2 ) g iv e s

MR - c r MR - c

(A.5) x p ( x , R)
R
- R - q X =
-

s R - q - A
.

Mu l t i p l y b y s ( R- q) and su bs t i tu t e f or \ to ge t

+
(A. 6) sx [ ( R - q) p - ( MR - c ) J = r [MR - c ] - [ MR MR R]
R É x R

+ MR - c •

R - q R •

F r om ( A . 3 ) the l a s t t e rm i s j u s t -sx ( MR - c ) a n d t h u s
c a nc e l s t h e l a s t t e r m on t he left ha nd- s i d e t o gi ve-

. .
{ A. 7 ) sx ( R - q) p
R
= r ( MR - c ) - MR X - MR R
x R

t o s u b s t i t u t e f o r R, g i v e s
.

S ol v i ng f or x , u s ing (A . 3)
 68

sx(R : q ) (M - pR ) + r(MR - c )
(A. 8) x =
MR
X

This can be rewritten as

sx

r(MR - c )

X = [ -MR J ( q - R ) [ MRR - p R )
0

+
• x( R - q )
X

Finally, MR = p + p so
X
or

r (MR - c ( q )
- sx ] ( q - R ) [ xp ) ]
X =
[ MR xR + sx R -
q •

(A. 9 )

 69

B i b l i o g r aph y

Ak e r l o f , G. , " T h e Ma r k e t for Lemon s , " Qua r t e r ly Jou r n a l o f E c o n om i c s ,

197 0 .

B u t t e r s , G . , " E qu i l i b r i um D i s t r i bu t i on s o f S a l e s a n d Adve r t i s in g
P r i c e s , " Re v i ew o f E c onom i c S t ud i e s , 1 977 .

Da rby , M . , a n d E . K a rn i , " Fr e e Comp e t i t i o n a n d t h e Opt ima l Amo un t

o f F r a ud , " J o u r n a l o f Law and E c o nomi c s , 1 97 3 .

Farrel l , J. , " P r i c e s a s S i gn a l s o f Qua l i t y , " 1979 .

" A Mod e l o f S t o c ha s t i c Qua l i ty , " 1979 .

Go u l d , J . , " D i f f u s i o n P ro c e s s e s a n d Opt ima l Adve r t i s i n g P o l i c y , "

i n P he l ps e t a l . , M i c r o e c o n om i c Founda t i on s o f Emp l o ym e n t
a n d I n f l a t i o n T h e o ry , 197 0 .

G r o s sm n , S . , K ih l s t r om , R . , and M i rma n , L . , " A B a y e s i an App r o a c h

t o t h e P r o d u c t io n o f I n f o rm a t i o n a n d L e a r n i n g by Do i n g " , Rev i ew
o f E c o n om i c S tud i e s ,

F.c t.owi t z , Y . , a n d F . Ma t h e w s o n , " A d v e r t i s i n g , C o n s um e r I n f o r ma t io n ,

a n d P rodu c t Qua l i t y , " B e l l J o u r n a l o f E c o nomi c s , F a l l 1979 .

L a n c a s t e r , K . , " A N e w Appro a c h t o C o n s um e r T h e o ry , " J o u r n a l o f

P o l i t i c a l E c o no my , 19 6 6 .

------ , S o c i a l l y Opt i ma l P r o d u c t D i f f e r e n t i a t i on , " Ame r i c a n

E c o n om i c Re v i e w , 197 5 .

Ne l s o n , P . , " Ad ve r t i s i ng a s I n f o rma t i o n , " J o u r na l o f P o l i t i c a l

E c on omy , 1974 .

, " I n fo rma t i o n a n d C o n s um e r B e ha v i o r , " J o u rn a l o f

-- -.
P o l i t i c a l E c onomy , 197 0 .

S a l o p , s . , and J . S t i g l i t z , " B a r ga i n s a n d R i po f f s , " Re v i ew o f

E c o n om i c S t u d i e s , 1977 .

S c hma l en s e e R . , " A d v e r t i s i n g a n d P r od u c t Qu a l i t y , Journal o f Po l i t i c a _

E c onomy , 1978 .

S hap i r o , C . , " P r e m i um s f o r H i g h Qua l i t y P r o d u c t s a s Ren t s to

Repu t a t i o n , " M . I . T . , 19 8 0 .

Sh e s h i n s k i , E . , " P r i c e , Qu a l i t y a nd Qu a n t i ty R e g u l a t i o n i n Mon opo l y

S i t u a t io n s , " E co n om i c a , 1 9 7 6

Sma l l wood , D . , a n d J . C o n l i s k , " P r o d u c t Qua l i ty i n Ma r k e t s Wh e r e

Con s ume r s A r e I mpe r f e c t l y I n f o rme d , " Q u a r t e r l y J o u r n a l o f
E c o no m i c s , 1979 .

S p e n c e , M . , " Mo n opo l y , Qu a l i t y , and R e r; u l :;;, 1: i :-:. , :::_�__::_:::_ _ _ _____

o f E c o nom i c s , 197 5 .
 70

Spence , M . , " Consumer i sperceptio n s , Produc t Fai l ur e , and

Produc e r L i ab i l i ty , " Rev i ew o f Econom i c Stud i e s , 1 9 7 7 b .

W i l d e , L . , and S c hwar t z , A . , " Equ i l ibruirn Compar i son S ho pp i n g " ,

Rev i ew o f Econom i c S tud i e s , 1 9 7 9 .