Professional Documents
Culture Documents
Risk Attitudes: Rdarrj (
Risk Attitudes: Rdarrj (
Risk Attitudes
r5: his chapter marks the beginning of our in<iepth study of preferences.
-L Before rve begin, let us revierv where rve have been and think about rvhere
lve are going. The frrst six chapters provided an introduction to the process of
structuring decision problems for decision analysis and an overview of the role
that probability and utiliry theorv play in making choices. Chapters 7 through 12
have focused on probability concerns: using probability in a rariety of ways to
model uncertainry in decision problems, including the modeling of informarion
sources in value-of-information problems.
At this point, we change directions and look at the preference side of deci-
sion analysis. How can we model a decision maker's preferences? This chapter
looks at the problems associated with risk and return tradeoffs. Chapter 14
briefly explores the axiomatic foundations of utility theory and discusses certain
paradoxes from cognitive psychology. These paradoxes generally indicate that
people do not make choices that are perfectly consistent with the axioms, even
though they may agree that the axioms are reasonable! Although such inconsis-
tencies generally do not have serious implications for most decisions, there are
certain occasions when they can cause difEculty.
The primary motivating example for rhis chapter comes from the history of
railrvays in the United States. Imagine whar rvas going through E. H. Harriman's
mind as he considered his strateg-y for acquiring the Northern Pacific Railroad
in March I901.
363
RdarrJ( (rttz)
t'.
l.r4 CHAPTER I3 RISK ATTITUDES
I have thc Illinois Central, the Union Pacific, the Central and Southern pacific,
not t() tncntion thc Oregon Railroad and Navigation Companr,. Isn't
that true?,.
"\'cs, sir'," re plicd his assistant.
"\4/cll, rve x'ill put the prcssure on Messrs. Hill and
Morgan. They rvill be sur-
priscd indeed to find out tltat I ltave acquired a controlling-interest in their orvn
railroad, the Northcrn Pacific. I may even be able to persuade them to let
me
ha'e the Brrrlington. By the way, horv are thc stc:k ptrrcrrases going?"
"Sir, wc have completed all of the purchases ttrat
you so far. you
may lrave noticed that our transactions have driven rhe price "r.rthorized
of Northern pacific
stock up to more than $100 per share.,'
Ilarriman cot'tsidered this information. If he bought too f;asr, he could force
the stock price up high enough and fast enough rhar Hitt might begin
ro suspecr
that Harriman was uP to something. Of courie, if Harri*ai .o,,li acquire the
shares quickly enough, rhere would be no problem. on the other hand, if
he
bought the shares slorvly, he would pay lower prices, and Hill might not norice
the acquisition until it rtas too late. His assistant's information,-horvever, sug-
gested that his situation was somewhat risky. If Harriman's plan were discovered,
Hill could persuade Morgan to purchase enough arlditional Northern pacific
shares to enable them to retain control. In that case, Harriman would have
paid
premium prices for the stock for nothing! On the other hand, if Hill did not
make the discovery immediately, the triumph would be that much sweerer.
'FIow many more shares do we need to have control?" asked Harriman.
'If you could purchase another 40,000 shares, sir, you would own 51 percent
-
of the company."
Another 40,000 shares. Harriman thought about giving Hill and Morgan or-
ders on how to run their own railroad. Horv enjoyable that would be! yes, he
rvould gladly increase his investment by that much.
'Of course," his assistant continued, 'if we try to purchase these shares im-
mediately, the pricc will rise very quickly. you wilr probably ehd up pa)rrng an ad-
ditional $15 per share above what you would pry ir rve were to proceed more
slorvly.'
'well, $600,000 is a lot of mone1,, and I certainly would not wanr ro pay
more. Bt't it would be worth the money to be sure that we lr,ould be able to
rvatch Hill and Morgan squirm! Send a telegram to my broker in New York right
alvay to place the order. And be quick! It's already Friday. [f rve are going to do
this, rve treed to do it today. I don't rvant Hill ro have the chance to think about
this over the weekend.'
R.ISK
Basing decisions on exPected values (EMvs) is convenient, but it can lead to
decisions that may not seem intuitively appealing. For example, consider the fol-
lowing two 8ames. Imagine that you trave ttre opportunity io play one garne or
the othcr, but only one time. Which one would ytu prefer to ptayl youi choice
also is drawn in decision-rree form in Figure 13.i.
365
RISK
Garnelirasanexpectcdvaltteof$14'50'Game2'orttheotherhand'hasan
your choice on the basis of ex-
cxpccted valtre of ssd.oo. If you were ro make
Most of us' horvever' would con-
pectcd value , then 1'ou *ot'li choose Game 2'
itseems reasonable to suspect that
sider Game 2 to be riskier than Game 1, and
most peoPle actually would prefer Game I'
means that the decision maker is
Usir-rg .*p..,.i values io make decisions
If rve take a long-run frequency
considering only the average or exPected payoff.
rve would be likely to rvin
approach, the exfect d .'a"lue is the average amount
the range of possible values' After
over many plays of the game. But this ignoies
rr'e cotrld at in Game 1 is to lose $10'
all, if r*e play each game l0 times, the w"orst
Ontheotherhand,theworstwecoulddoinGame2isloseSl9'000!
Marryoftheexamplesandproblemsthatrvehaveconsideredsofarhave
value (Eltfv) ' EMV' however' does
been analyzed in terms of expecied monetar' example
the Texaco-Pennzoil
nor caprure risk attitudes. For example, coniider
in chapter 4 (pages 65{9). r Hujn Liedtke y..^.
rfr"id of the prosPect that
pennzoil could .ia ,-rp with nothi.ig .t the end of the court case' he
might be
the Eagle Airlines
rviling ro rake the g2 Uittio' that Teiaco offered' To considera much riskier alter-
the airplane is
case (chapter 5, pages 113-114), purchasing
were sensitive to risk' he
native than leavirig it . money in tt e bank. I1Carothers
someone like E' H' Harriman
might prefer to l.Iue the money in the bank. Even
he found himself' [n our
considered the riskiness of the situations in which
alternative (immediately pur-
example, Harriman weighed the value of a riskless
gain tontrol) against the risky
chasing the 40,000 shaies that were required to
-1 900
alternative of not purchasing the shares and the possible outcomes t6at mig6t
then follow. Evctr tlrough all of the dollar anr()urlrs were not specified, it is clear
that Harriman \{as not thinking in [ernts of EM\/.
Individualswho are afraid of risk or are sensitive [o risk are called risk-aua-se.
We can explain risk aversion if we think in rerms of a utitits-
lunction (Figure 13.2)
that is curved and opening downward (the tcchnical term for .u.*," wit6 this
shape is "concave"). This utility functio. represents z r'ay, to "tra.slate dollars
irlto "utilicy units." That is, if rve take some dollar atnounr (x), we can locare rhal
amounI on the horizontal axis. Read up to the iune and then horizontally
across to the vertical axis. From that point \ve carl read off the utilicyvalue U(.r-)
for the dollars we started with.
A utility function might be specified in terms of a graph, as in Figure i 3.2,
or given as a table, as in Table 13.1. A third form is a mathematical expiession. If
graphed, for example, all of the following expressions would have the same gen-
eral concave shape (opening downward) as the utility function graphed in hig-
ure 13.2:
U(x) = log(x)
U(x)=L-e-x/R
U(x) = af* (or U(x) = ;o.s;.
Of course, the utility and dollar values in Thble l3.l also could be graphed, as
could the functional forms shown abo..,e. Likewise, the graph in Figure i3.2
could be converted into a table of values. The point is that the utiliry funcrion
makes the translation from dollars to utiliq, regardless of its displayed form.
Utility
u(x)
Wealth
Utilitl
\\'caltli \/aluc
2i()() ) 7()
'l i(x) l2,r
.j()(x) () !]3
(i()(l (l (r5
I ()() () 1;
0 () )7
O RISK ATTITUDES
\{'e think of a rypical utiliq,curve as (l) uprr,ard sloping and (2) concave (the
curve opens dou'nrr'ard). An uprtard sloping utility curve makes fine sense; it
means that tnore rtealth is better than less s'ealth, everything else being equal.
Ferr' people will argue rtith this. Concavitt' in a utility crlrve implies that an indi-
vidual is risk-averse.
Imagine that you are forced to play the follor.r'ing game:
\Alin $500 rvith probabiliry 0.5.
Lose 5500 u,ith pr6bability 0.5. -
llbuld ),ou pay to ger our of this situation? Horr' much? The game has a zero ex-
Pectell ralue, so if you rtould pat' something to get our, you are avoiding a risky
situation *i4, zero expected rzlue. Generally,,lf you \oplq trade a ble for a
sute a!qogg1|-Ulr! is less than the expected ralue oFThe
ge: Purchasing insurance is an example of risk-averse behavior. lnsurance
companies analyze a lot of dau in order to understand the probabiliry distribu-
tions associated rvith clainrs for different kinds of policies. Of course, this rvork is
costly. To make up these costs and still have an expecr.ed profit, an insurance
company must charge more for its insurance policy, than the policy can be ex-
pected to produce in claims. Thus, unless vou havr: some reason to believe that
)'ou are more likely than otl-rers in your riskgroup to make a claim, you probablv
are paying more itr insurance premiums than the expected amount you rvould
claim.
Not everyone displal's risk-averse behavior all the Lime, and so utility curves
need not be concave. A convex (opening uprvard) uriliry curve indicates risk-
seeking behavior (Figure 13.3). The risk seeker might be eager ro enter into
a gamble; for example, he or she might pa), ro pla1, the game just described.
An individual who plays a state lottery exhibits risk*eeking behavior. State lot-
tery tickets typically cost $1.00 and have ah expected value of approximately
50 cents.
---+
Flisk;Seeking
Risk-Averse
Wealth
le INVESTINGINTHESTOCKMARKET'REVISITED
IIrr.clrar,cautility[urrctiotttlrattranslatcs[i.omdollarsl-out,ilir}',lr<rwslrouldrr,e
trsc iti' Tltc rt'hole idca qt a utility fttnctiotr is thdt it should help to choose
fror-n
expecterl
atn()11g alte rnativcs that have
t,nc"rtait' payoffs' Instlad.3f maximiz'ing
tree or
t,c dccision ,nakerlshortlci n''axinii'e cxpected^utiliqL In a decision
b}' the
'alrrc,
infltre ncc-diagram paYoff talrl<:, thc
uct dollar payoffs rvould be replacecl
corl.csPot}dirrgutilitl.r.alttcsartcltlrcarrall,sispcrfo.rmcd.usingtlroser'alues.Tlre
best choice then should be thc act'ion
rfith the highest expected uriliq"
tl-re stock ma-rket-iuvestment
example fr-om
As an examplc, Ie I tts reconsiclcr
Chapters5andl2.Yourtillrecalltlratarlinvestorhasfurrdsthathervishestoin-
vest.Helrastl.rreeclroices:ahiglr-riskstock,alow-riskstock,orasavingsaccoullt
rhar rr,ould pa1, $500. If he invisG
in the stocks, he must Pay a $200 brokerage
fee. . r happens to the market' If
n'hat-L^-^.
\\,ith the rrvo srocks his pavoff depends on high-risk-stock and $1200 from
fr"m the
the marker goes .rp, r,.-**iiiJ"rn srzoo the high-
at the same livel' his payoffs for
the lolv-risk stock. If the market sta\s mar-
resPecdvely- Finally' if the stock
and lorv-risk stocks will be $300 and $400' earn $100 frorn
ket goes donn, he rvill tose $g0o
with the iigt -ritt stoc.l.<, but still
that thl market will go uP' s.,y the same' or
the lou.risk stock. Th.';;;;;Uiti,i.,
go dorvn are 0'5, 0'3' a-n-d q'2'.r.esP:cl]tl"'
Figurel3.4shorvshisdecisiontree,includingthebrokeragefeeandthepay-
Note that the values at
nro stocks under different market.o-,,di.io'.u.
offs for the
theendsofthebranchesaretherulpayoff$dlsintoaccountboththebro-
I3.2 gives his utility furrction'
kerage fee and .h. i";:;;;;nt pa1,off.'raute in Chap
values oritre three investments
\Are alread),."r.,rrJrlJ;;;#g.ted
ter 12' TheY are:
EMV(High-Risk Stock) = 580.'
EN{\''(t-$LRisk Stock) = 540
EMV(Savings Account) = 500'
rvouldthoose the high-risk st'ock'
As a result, an expected-r'alue ntaximizer
1 500
High-Risk 1rl0
Stock
(-200) -1 000
(EMV = s8o)
1 000
200
-1 00
500
DoUar
Uriliq,
\raluc
l'alue
I 500 1.00
1000
0.86
. 500 0.65
200 0.52
t00 0.46
-100 0.33
-1000 0.00
EU(High_Risk Stock)
= 0.63g
' Eu(l.orv_Risk Stock)
= 0.652
EU(Savings Accounr)
= 0.650.
Now the preferred acdon is
to invest in the low-risk stock
highest expected udlity, because it pro*ides the
lngs accounl You can see "r,h;"dl;'d".; ;;;;itd, ,o''.r.n from thai for rhe sar,_
trotithe expected utilities
these invesrmenm in order make it possibre ro rank
"i p..i..1ce;Accolding to the utiriq, funcrion r*e
o,,gr, to n n d &r e h ish-ri, k,,"J'
;:il:::? i| i:,: H::': :f:'.*L:,:t1,, e r e as,;
1.00
High-Risk 0.46
Stock 0.00
(EU =
0.86
Stock 0.52
0.33
0.65
tltility 1y1
Expecled
Utility
Certainty
Equivalent
(cE)
Expecled
Value
(El"M f
FIGURE t3.6 Gaphicat Representation of Risk kemium
EXPECTED IJTILITY. CERTAINTY EQUIVALENTs, AND
RISK PREMIUMs 373
1.00
o.B0
0.60
0.40
0.20
-3000 i
4CO0
[i{iH1ffi?{t;#j[*rl[,ff .'.'t':]II.Tr"'dti;n::
t"it'ty.q"i*r.",i, ts.s, *..-;r,:;'Lr1in.
*pp.o*imatery .".-
Figure
*rjbl'"-
il';::T::ilri[rr:ifj;ir*;:ll:ffi
H/c can
. gcr trrc first h1,o points of
i,orrr uririry functjon by arbirariry, ser_
ilL.,J', $?"' lol"l F,H"t i,ll,,;''- :: :'-T I'#', stranse''b;; ;;''*i e x-
L o,,, w" .*
r
1',
u'i,,ing ,o
se,, vour oppor-
iffi,i"t"T';l,TJ*:l#j;:;Il:'LlH
(CE) for this reference
gamble' A ;^',1:"
x11i3;
rs to nnd yoLrr ceraincy
equivalent
decision tree for your choice
Figure I3.g. is shorrn in
. Finding your certainq,equirarent is r*here'our subjecd'e judgmenI
rnto play. The cE undoubieary
this reference gambr.
*iii*r.1-from p.rro' io p.r*.,. Suppose comes
yor. ..ioirf .quirarent is S30. that for
That is, for $3r you rvourd
F
t:..:c tire rnone)', but for $29 you would rather play the lortery; $30 must be your
u ue indifference point.
The key to tlte rest of thc analysis is this: Because you are ind.ir(fcrcnr berween $30
nnd llu nslq gantbb, tfu utility oI $30 must equal tltz expecicd utiLity of tlrc gambb. \\te
knorv the utilities of $10 and $100, so we can figure our rhe expected utitiry of
thc garnblc:
U(30) = 0.5 U(100) + 0.5 U(10)
= 0.5 (t) + 0.5 (0)
= 0.5.
We have found a chird point on your utility curve. To 6nd anorher, take a differ-
. ent reference lottery:
We now have five points on your utility curve, and we can graph and draw a
curve through them. The graph is shown in Figure 13.10. A smooth curve drawn
through the assessed poins should be an adequate representation of your utiliry
function for use in solving your decision problem.
Utility
1 .00
0.7 5
i
0.50
0.25
20 40 60 80 100 Weatth
$100
$10
l;
fusk aversion and insurance