Você está na página 1de 10

See the top panel o% *igure 4!

>e have x < z i% and only i% r < r


0
! >e need the pro)a)ility
o% r < r
0
conditional on

!
*or this, 'e must turn to player iGs posterior density over r conditional on

! "his posterior
density is uni%orm over the interval B

D C, as in the lo'er panel o% *igure 4! "his is


)ecause the ex ante distri)ution over r is uni%orm and the noise is uni%ormly distri)uted around
r! "he pro)a)ility that r < r
0
is then the area under the density to the le%t o% r
0
, 'hich is
r
0
/

2
2
F
/

D 2z2 /

2
2
F z /1&2
'here the second line %ollo's %rom su)stituting in /122! "hus, the pro)a)ility that x < z
conditional on

is exactly z! "he conditional c!d!%! G/z|

2 is the identity %unctionE


G/z|

2 F z /142
"he density over x is thus uni%orm! *inally, note that the uni%orm density over x does not
depend on the value o% ! *or any seHuence /
n
2 'here
n
0, the density over x is uni%orm!
"his proves Jemma 1!
;n the limit as 0, every investorGs signal converges to the true interest rate r! "hus,
%undamental uncertainty disappears, and it is 'ithout loss o% generality to 'rite the investorGs
strategy as )eing conditional on the true interest rate r! "hus, 'e search %or an eHuili)rium in
s'itching strategies o% the %ormE

Sell i% r > r

Hold i% r r

/152
*igure 4 reveals the intuition %or Jemma 1! 9s shrins, the dispersion o% signals shrins
'ith it, )ut so does the support o% the posterior density over r! "he region on the top panel
corresponding to z is the mirror image o% the region on the )ottom panel corresponding to
G/z|

2! +hanging stretches or sHuee.es these regions, )ut it does not alter the %act that the
t'o regions are eHual in si.e! "his identity is the ey to the result! "he uni%orm density over x
11
has )een du))ed #Japlacian )elie%s$ )y Morris and Shin /200&2, and entails that the strategic
uncertainty %aced )y players in the glo)al game is at its maximum, even 'hen the %undamental
uncertainty %aced )y players shrins to .ero!
!$ %olution
3iven Japlacian )elie%s, the s'itching point r

is the interest rate that maes each asset manager


indi?erent )et'een holding and selling! "hat is,

satis8es

1
0
Bu/sL r

2 w/sL r

2C ds F 0 /1<2
Ho'ever, since u/.2 and w/.2 are )oth linear %unctions o% x, the s'itching point r

can )e
o)tained )y solvingE
u

1
2

F w

1
2

/1I2
'hich can )e 'ritten asE
v

1
1
2

F /1 D r

p
1
2
c
1
2
A

/112
"here%ore, the threshold interest rate r

is given )y
1 D r

F
v /1 /22
p
1
4
cA
/102
F
v /1 /22
v cS D
3
4
cA
/202
;t remains to veri%y that 'hen asset managers strictly pre%er to sell 'hen r > r

, and strictly
pre%er to hold 'hen r < r

! ,oth propositions %ollo' %rom the monotonicity o% the payo?


di?erence u /x2 w/x2! >e have u /x2 > w/x2 to the le%t o% r

and u/x2 < w/x2 to the right


o% r

! "he s'itching strategy around r

is an eHuili)rium, and the only s'itching eHuili)rium!


"he monotonicity o% the payo? di?erence u/x2 w/x2 entails that the s'itching strategy
around r

is the uniHue dominance solva)le eHuili)rium in the sense that it is the only eHuili)rium
that survives the iterated deletion o% strictly dominated strategies /Morris and Shin /200&,
12
Section 222! "here%ore, the solution given )y /202 is the complete solution in that there is no
other eHuili)rium 5 'hether in s'itching strategies, or in any other strategies! >e summari.e
the solution as %ollo's!
Proposition There is a unique, dominance solvable equilibrium. In this equilibrium, all
asset managers use the switching strategy around r

dened by (20 and sell the ris!y bond when


r > r

and hold when r r

.
>e note some properties o% the solution! *irst, the threshold interest rate r

is decreasing
in ! "hus, the 'orse is the last5place aversion o% the asset managers, the more (ittery they
)ecome and the lo'er is the interest rate at 'hich they (ump %rom holding the risy )ond to
selling out!
Perhaps more important is the e?ect o% changes in A, the si.e o% the asset management
sector! >hen the asset management sector is large relative to the household investors, the
price impact o% concerted sales is large! "he strategic interaction )et'een asset managers is
thus heightened! "o use our analogy 'ith the musical chairs game, a larger asset management
sector means that the musical chairs game )ecomes more competitive! "here is more at stae
in coming last in the game, so that asset managers are 'illing to (ump ship at a lo'er threshold
interest rate!
"he impact o% the asset management sector can )e seen in several %eatures o% our solution!
*rom *igure 2 a)ove 'hich illustrates the maret clearing condition, the larger is A relative to
the total stoc S, the higher is the maret price p! "hus, as A increases, the ris premium
o% the risy )ond )ecomes more compressed! "he ris premium 'hen the si.e o% the asset
management sector is A is given )y
v
p
F
v
v c /S A2
/212
'hich is decreasing in A! "hus, a large asset management sector can )e used )y the central
)an to eep the ris premium compressed!
1&
Impact of larger asset
management sector
Risk premium
"igure 2. Risk premium and critical threshold r

as a function of the si!e of asset management sector


Ho'ever, there is a tradeo? that comes %rom the larger asset management sector! >e see
%rom our solution %or the threshold interest rate r

in /202 that the threshold interest rate is also


decreasing in A! "his means that the economy 'ill (ump to the high ris premium regime at
a lo'er value o% interest rates!
*igure 5 illustrates the e?ect o% a larger asset management sector! Jarge A entails a lo'er
ris premium in the lo' ris premium regime, )ut the (ump to the high ris premium regime
happens at a lo'er level o% the interest rate! "hus, 'hen the ris premium (umps at the trigger
point, the (ump 'ill )e larger!
"urning the comparison around, %or any given interest rate r, there is an upper )ound to the
si.e o% the asset managment sector that is consistent 'ith the lo' ris premium regime! *rom
the expression %or the critical threshold r

given )y /202, %or the economy to )e in the lo' ris


premium regime, 'e needE
1 D r < 1 D r

F
v /1 /22
v cS D
3
4
cA
/222
14
"his gives us an upper )ound %or A %or the lo' ris premium regime, namelyE
A <
4
&c

v /1 /22
1 D r
v D cS

/2&2
So %ar, 'e have assumed that A is exogenous! Ho'ever, i% 'e suppose that A is gro'ing in the
lo' ris premium regime, then /2&2 represents the relationship )et'een the %easi)le si.e o% the
asset management sector and the interest rate r! 9s A gro's, the central )an can maintain lo'
ris premiums )y eeping the interest rate lo'! Ho'ever, once the )ound is reached, the central
)an must reduce interest rates %urther in order to accommodate the gro'th in A! 7uring this
process, the ris premium continues to )ecome compressed!
Ho'ever, )y accommodating %urther increases in A, the central )an is )acing itsel% into a
corner, as sho'n in *igure 5! "he ris premium gets compressed as A gro's, )ut the threshold
point moves do'n! >hen, eventually, the central )an has to reverse course and raise interest
rates, the (ump 'ill happen at a lo'er interest rate, and the (ump in ris premium 'ill )e that
much larger!
$ Multi-Period Model
>e no' consider a multi5period version o% our model that is in the same spirit as our )enchmar
three period model! "ime indexed )y {0, 1, 2, , T, T D 1}, 'here date T D 1 is the terminal
date, interpreted as the #long run$!
>e have the same t'o groups o% investors 5 ris5neutral asset managers, and ris5averse
households 'ith mean5variance pre%erences! ;nvestors do not discount the %uture, and everyone
consumes only at date T D 1, the #long run$!
"he risy asset is a .ero coupon )ond that pays o? at the terminal date, and the expected
payout is v, as )e%ore! "here is late resolution o% uncertainty in that the uncertainty in the
payout o% the risy )ond is resolved only )et'een date T and T D 1! "his assumption o% the
late resolution o% uncertainty means that at the trading dates )et'een date 1 and date T, the
competitive demand curve o% the household sector remains unchanged!
15
9sset managers start out at date 0 holding A
0
units o% the risy )ond! >e denote )y A
t
the holding o% the risy )ond )y the asset management sector at date t! "he 6oating rate o%
interest )et'een date t to date t D 1 is denoted )y r
t
! 9t trading date t, asset managers have
precise in%ormation on the 6oating rate r
t
that rules )et'een t and t D 1, )ut r
t
is not common
no'ledge!
"he in%ormation structure %ollo's the )enchmar three period model! 9sset managers
o)serve a noisy signal o% the true interest rate %or the next interval o% time, and 'e examine the
case 'here the noise tends to .ero! "hus, 'e can condition the trading strategies o% the asset
managers on the true path o% interest rates /r
t
2
T
t=1
!
"hus %ar, the multi5period model is no di?erent %rom the three period model! Ho'ever,
there are t'o ne' elements that are introduced in the multi5period model! *irst, in order %or
asset managers to hold the risy )ond, the time path o% prices must )e such that the rate o%
price appreciation is at least as large as the expected return %rom the 6oating rate! "hat is, at
all trading dates t, 'e have
v
p
t
/1 D r
t
2 E
t
B/1 D r
t+1
2 /1 D r
T
2C /242
"his condition is reminiscent o% the Hotelling condition, and puts an upper )ound on the price
o% the risy )ond at each date! ;% /242 %ails to hold, then an asset manager has the incentive to
sell the risy )ond and roll over the proceeds o% the sale at the 6oating rate! ;n terms o% A
t
,
/242 imposes an upper )ound on the si.e o% the asset management sector at each date!
"he second ne' element in the possi)ility o% selling and then )uying )ac the risy )ond
a%ter one period! Just as selling entails #riding do'n$ the competitive demand curve o% the ris5
averse households, )uying )ac the risy )ond entails #riding )ac up$ the demand curve, 'hich
no' acts as the supply curve %or the risy )ond provided )y the ris5averse households! 3iven
the possi)ility o% )uying )ac the risy )ond, the value %unction incorporates the possi)ility o%
)uying )ac the asset!
7enote )y
t
/x
t
2 the value %unction %or an asset manager 'ho sells the risy )ond holding
at date t 'hen proportion x
t
o% asset managers also sell! "he expected proceeds o% the sale is
1<
p
t

1
2
cx
t
A
t
, 'hich earns the 6oating rate 1Dr
t
)et'een dates t and tD1! 9t date tD1, ho'ever,
asset managers 'ho sold in the previous period then )uy )ac the risy )ond )y #riding )ac
up$ the household demand curve! "he value %unction
t
/x
t
2 is thus given )y

t
/x
t
2 F
p
t+1

p
t+1

1
2
cA
t+1
v
!
/1 D r
t
2

p
t

1
2
cx
t
A
t

d!
F /1 D r
t
2

p
t

1
2
cx
t
A
t

v ln

p
t+1
p
t+1

1
2
cA
t+1

/252

t
/x
t
2 is decreasing in x
t
!
"he value %unction at date t o% an asset manager 'ho holds the risy )ond 'hen proportion
x
t
sell the risy )ond is identical to that in the three period model and is given )y
"
t
/x
t
2 F v /1 x
t
2 /2<2
+learly, "
t
/x
t
2 is also decreasing in x
t
! Mote that )oth "
t
/x
t
2 and
t
/x
t
2 are linear in x
t
!
"hus, given Japlacian )elie%s, 'e can solve %or the threshold 6oating rate r

t
%rom
"
t

1
2

F
t

1
2

/2I2
so that r

t
satis8esE
1 D r

t
F
1

2

p
t

1
4
cA
t

ln

p
t+1
p
t+1

1
2
cA
t+1
/212
"he %ull solution entails solving out %or the prices recursively )ac'ards %rom date T D 1 in
terms o% the %undamentals! >hen the reali.ed 6oating rate is a)ove the threshold rate r

t
, there
is a sell5o? o% the risy )ond at date t, )ut the asset managers )uy )ac the risy )ond at date
t D 1!
"he typical time path o% asset manager holdings and the ris premium %ollo'ing the triggering
o% the (ump o% yield 'ill )e as in *igure <! ;nterest rate shocs then have persistent impact!
:ven 'hen the asset managers )uy )ac the risy )ond at t D 1, the %easi)le si.e o% the asset
managment sector is smaller at the higher interest rate r, so that the ris premium does not
revert )ac to its previously lo' level!
1I
Increase in
Partial rebound in
"igure 3. /ersistent impact of increase in interest rate above threshold r

& Implications for monetary policy


Monetary policy is a po'er%ul tool %or in6uencing 8nancial conditions! ;n particular, the
commitment to lo'er interest rates into the %uture raises the prices o% 8nancial assets and
compress ris premiums, 'ith conseHuences %or real economic activity! ;n this respect, our
analysis shares the conclusions %rom orthodox monetary analyses on the impact o% %or'ard
guidance, and especially on the commitment to lo'er policy rates into the %uture /see, %or
instance, >ood%ord /20122 %or a %orce%ul statement o% this argument2!
Ho'ever, our analysis parts company 'ith orthodox monetary analysis on 'hether %or'ard
guidance and commitment to %uture rates is a policy that can )e 8ne5tuned, or )e reversed
smoothly 'hen the time comes to change tac! "he #maret$ is not a person, and maret
prices need not correspond to the )elie%s o% that person! ;n our glo)al game analysis, monetary
policy 'ors through the #ris5taing channel$ through the ris5taing )ehavior o% di?erent
11
sections o% the maret! Monetary policy a?ects ris premiums directly, so that the impact on
real economic activity 6o's through shi%ts in ris premiums, as 'ell as shi%ts in the actuarially
%air long term rates!
=ne lesson %rom our analysis is that Huantities matter! "he si.e o% the asset management
sector, as encapsulated )y the holding o% risy )onds A
t
, determines the ris premium ruling
at date t, as 'ell as the threshold point %or the 6oating interest rate r
t
'hen a sell5o? occurs!
"o the extent that Huantities matter, the lesson is similar to the one %rom the 2001 8nancial
crisis! Just as 'e 'ould )e concerned 'ith a )uild5up o% leverage and in the si.e o% )an )alance
sheets, 'e should similarly )e interested in the gro'th o% holdings o% 8xed income securities o%
)uy5side investors! "he central )an can compress ris premiums %urther )y committing to lo'
%uture interest rates, and accommodating an increase in the si.e o% the asset management sector!
Ho'ever, there is a trade o?! ,y accommodating %urther gro'th o% the asset management sector,
the central )an is trading o? a lo'er ris premium today %or a more disruptive un'inding at
a lo'er threshold interest rates 'hen, eventually, the central )an has to reverse course!
=n the empirical %ront, our model suggests that o)serving the (oint movements o% price
changes and Huantity changes is in%ormative a)out the ris5taing o% maret participants! ;n
particular, the model predicts the (oint occurrence o% price declines and sales o% the risy )ond!
"hus, rather than the demand response )eing to cushion shocs, the demand response tends to
ampli%y shocs!
*eroli, 3reenla', 4ashyap, Schoenholt. and Shin /20142 8nd in their N9- analysis o% price
and valuation changes %or risy 8xed income categories /such as mortgage )aced securities,
corporate )onds and %or emerging maret )onds2 that price declines are %ollo'ed )y sales, and
sales are %ollo'ed )y %urther price declines! ;n this 'ay, the accumulated impulse responses o%
price and Huantity shocs are large!
9n implication %or the conduct o% monetary policy is that the separation )et'een monetary
policy and policies to'ard 8nancial sta)ility is much harder to accomplish than is o%ten sug5
gested! Under the ris5taing channel, monetary policy impacts the economy through shi%ts in
the ris5taing )ehavior o% maret participants! 9s such, any monetary policy shoc is also a
10
shoc to ris5taing and hence is insepara)le %rom the concern %or 8nancial sta)ility!
7iscussions o% 8nancial sta)ility a%ter the crisis have )een conditioned )y the experience o%
the crisis itsel%! "hus, having neglected the dangers o% excessive leverage and maturity mismatch
)e%ore the crisis, policy maers have given them central importance since the crisis! 9s is o%ten
the case, accounta)ility exercises usually address no'n past 'eanesses, rather than asing
'here the ne' dangers are!
Ho'ever, our analysis suggests that the ris5taing channel may operate through 8nancial
institutions that are not leveraged! 9sset managers typically have very lo' e?ective leverage,
and hence do not )ecome insolvent in the 'ay that )ans or highly leveraged hedge %unds
do! Ho'ever, this does not mean that they do not have an impact on the economy! 9s the
protagonists in 8nancial maret dynamics shi%t %rom )ans to asset managers, more attention is
needed on the part o% researchers on the maret5'ide impact o% institutional investors! Nayanos
and >oolley /201&2 sho' ho' momentum and reversals result %rom small agency %rictions, even
'ith long5only investors!
"he ris5taing channel o% monetary policy a?ects ris premiums directly, 'ith e?ects on
corporate investment and household consumption! "hese shocs could have a direct impact on
37P gro'th through su)dued investment and consumption! "he potential real economy impact
is real, even though no institutions %ail, and no 8nancial institutions are )ailed out using pu)lic
%unds! 9sset managers are not #systemic$ in the sense de8ned in the 7odd5*ran 9ct as they
are not #too5)ig5to5%ail$/","*2! Mor are there easy regulatory solutions that 'ould su)stitute
%or central )an interest rate policy in a?ecting ris5taing!
"hus, the most important implication o% our analysis is that monetary policy and policies
to'ard 8nancial sta)ility cannot )e separated! "hey are, e?ectively, the same thing!
20

Você também pode gostar