BeFi Conference Topic 4

Cognitive Ability and Financial Preferences

by Shane FrederickMassachusetts Institute of Technology

© B

eFi F

orum

200

7

Cognitive Abilityand

Financial Preferences

Shane Frederick

Massachusetts Institute of Technology

1

2

smarter people …

Have bigger heads Earn more money Live longer React faster

Are more susceptible to visual illusions

3

“I have come to view g as one of the mostcentral phenomena in all of behavioralscience, with broad explanatory powers…[it] shows a more far reaching practicalvalidity than any other psychologicalconstruct yet discovered.”

# @ # #○ ○ ○ ○

4How far reaching?:What else does IQ predict?

or or

$100 today $100 for sure$140 next year 75% of $200or or

magazine preferences?fruit preferences?

time preferences? risk preferences?

A bat and a ball cost $1.10 in total. The bat costs a dollar morethan the ball. How much does the ball cost?

Cognitive Reflection Test (CRT)

If it takes 5 machines 5 minutes to make 5 widgets, how longwould it take 100 machines to make 100 widgets?

In a lake, there is a patch of lilypads. Every day, the patch doublesin size. If it takes 48 days for the patch to cover the entire lake, howmany days would it take for the patch to cover half of the lake?

5

0%

5%

10%

15%

20%

25%

30%

35%

0 1 2 3

# correct (out of 3)

% of respondents

Distribution of scores

3328

2317

6

0

1

2

3

MIT PFM PTON FWD CMU HARV UM WEB BGU UMD MSU UT

CRT score, by population 7

“You can’t afford to be old yet.”

Is CRT related to time preferences?

% choosing “patient” option

30%

35%

40%

45%

50%

55%

60%

0 1 2 3

$3400 this month $3800 in 2 months

p<0.0001

CRT score

9

$0.00

$1.00

$2.00

$3.00

$4.00

$5.00

$6.00

0 1 2 3

p<0.0001

10Willingness to pay for overnightshipping of chosen book

Patie

nce

11Relation stronger for women

CRT

Is CRT related to risk preferences?

75% chance of $200$100 for sure

0%

10%

20%

30%

40%

0 1 2 3

p<0.0001

% choosing “riskier” option 13

75% chance of $4000$1000 for sure

p<0.001

30%

35%

40%

45%

50%

55%

60%

0 1 2 3

14% choosing “riskier” option

3% chance of $7000$100 for sure

p<0.01

0%

5%

10%

15%

20%

25%

0 1 2 3

15% choosing “riskier” option

Ris

k To

lera

nce

16Relation stronger for men

CRT

What about gamblingin the domain of losses?

17

take a 75% chanceto lose $200lose $100 for sure

p<0.0001

20%

30%

40%

50%

60%

0 1 2 3

18% choosing “riskier” option

take a 3% chanceto lose $7000lose $100 for sure

p<0.0001

20%

30%

40%

50%

60%

70%

0 1 2 3

19% choosing “riskier” option

“…to characterize people with high SAT scores asthose who should set the norm for what is somehowtrue or right seems to be off target. People with highSAT scores have high levels of certain kinds ofcognitive abilities. They have no monopoly on qualityof thinking and certainly no monopoly on truth.”

Sternberg(2000)

Do the preferences of the “smart” establish anormative standard for the appropriate choice?

20

“There are no right or wrong answers.”

Really? Are our instructions sincere?

15% chance of $1,000,000$500 for sure or

15% chance of $1,000,000$500 for sure

0%

20%

40%

60%

80%

100%

0 1 2 3 4 5 6 7 8

p<0.0001

Score on 8 item CRT

or

consider this… 22

%choosinggamble

[Kevin] Unless it’s a misprint, I just have to wonder what kind ofmoron would take $500 over a 15% chance at a million bucks?

[A] Well those would be the “morons” who understand risk muchbetter than you. $500 for nothing is a much better return than an 85%chance at nothing. Keep buying those powerball tickets Kevin.

[Kevin] [Choosing the $500] makes sense only you’re dead broke anda goon with a baseball bat is coming for his money. For most of us,even those of us who don’t make much money, $500 is simply notenough when the alternative has an expected payout 300 timeshigher.

[B] Money means different things to different people. [Kevin] I believe that some people would take the $500 over the 15%

chance of $1M, but those here defending them aren’t convincing methat they aren’t morons. … This faux debate about the value of $500vs. a 15% chance at $1 million parallels the faux debate aboutevolution vs. intelligent design. In particular, there is no debatewithin the scientific community on the matter. The only debate isfrom people who don’t understand the science.

23

Which preferences do we want tothink of as “mistakes”

24

or

or ?

NO

Which preferences do we want tothink of as “mistakes”

25

Extreme Impatience (Excessive Discounting)

$3400 this month $3800 in 2 months

Which preferences do we want tothink of as “mistakes”

26

Overweighting very small probabilities

Underweighting small probabilities

3% chance of $7000$100 for sure

Which preferences do we want tothink of as “mistakes”

27

Excessive Risk Aversion

15% chance of $1,000,000$500 for sure

0

5

10

15

20

25

30

35

40

45

50

1 2 3 4 5 6 7 8 9 10

(Benjamin, Brown, and Shapiro, 2007)

AFQT

part

icip

atio

n in

finan

cial

mar

kets

28If we conceive of some preferences as mistakes,how do we reduce them?

Raise everyone’s IQ (NO)

Arrange environment so 0’s act like 3’s (PERHAPS) By formulating the problem differently

“when you are 64”; “when you are 15 years older”

By educating people

about the concept of “Expected Value”

about the law of large numbers

about the observed correlations between IQ and preferences

29What would Johnny do? (Nisan and Koriat, 1977)

OR

Yesterday in another classroom, I met Johnny. Ioffered him a choice between one candy now or twocandies tomorrow. Johnny is a really smart kid.Everybody thinks so.

What do you think Johnny chose?

now tomorrow

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

$100vs.

50% of$300

$60 vs.

1% of$5,000

Tunavs.

Salmon

$3,400this

monthvs.

$3,800next

month

$500vs.

15% of$1,000,000

30

Fishpref.

timepreferences

riskpreferences

Do adults want to “be like Johnny”?

%making“smart”choice

cont

rol

info

rmed

cont

rol

info

rmed

cont

rol

info

rmed

c info

rm

c info

rm

Didn’t know it;& don’t care

Already knew it

Didn’t know it& care!

Two remaining questions

31

32(1) Does the pattern shown in stylized laboratorychoices hold for real world behaviors?

0%

5%

10%

15%

20%

25%

30%

35%

0 1 2 3

33(2) Would these facts have “normative force”?

Suppose a portfolio manager presents the true facts…. Our clients who scored a “0” preferred “D”. Our clients who scored a “1” preferred “C”. Our clients who scored a “2” preferred “B” Our clients who scored a “3” preferred “A”

D C B A

- + - + - + - +

The End

5 ¢ $1.05

34

0

5

10

15

20

25

30

35

40

45

50

1 2 3 4 5 6 7 8 9 10

0%

20%

40%

60%

80%

100%

0 1 2 3 4 5 6 7 8

(1) Does the pattern shown in stylized laboratorychoices hold for real world behaviors?

Answer: It sure looks like it.

CRT8 AFQT

36What explains the relation between CRT scores & risk preferences?

3% chance of $7000$100 for sure

0%

5%

10%

15%

20%

25%

0 1 2 3

Lower scoring people make more errors computing EV?

… and either

(1) systematically under estimate EV (2) dislike choosing option with uncomputed EV

People who find probabilities confusing adopt a riskaverse approach to decisions ?» “If you are bad at doing such risk calculations, you learn to be

conservative really fast, or you go broke.”

– But for losses the low CRT group is not more conservative; theyare more risk seeking

37

take a 3% chanceto lose $7000lose $100 for sure

20%

30%

40%

50%

60%

70%

0 1 2 3

%choosing

risky

CRT is related to wealth, and wealth, not “cognitivereflection” drives risk preferences» “It looks like most of the high CRT group came from MIT and Princeton and most of

the low CRT group came from Toledo and Michigan State. You can make someassumptions about the difference in opportunity costs of [foregoing the sure thing]if [one group] is going to spend it on keeping power on in their apartment, while theother is going to spend it on apple martinis for Muffy.”

– But if low CRT people are poorer, why are they more risk seeking inlosses?

– Benjamin, Brown, & Shapiro (2007) find correlation between AFQTscores & ownership of stocks (controlling for wealth).

– In a multiple regression with CRT, wealth has NO incrementalpredictive validity

38

What could “explain” the data?39

Probability

Decisionweight High CRT

Low CRT

0 1

1

0

P R E S E N T E D B Y

Shlomo BenartziCo-Founder, BeFiAssociate Professor Co-chair of theBehavioral Decision Making GroupThe Anderson School at UCLA

Warren CormierCo-Founder, BeFiPresident, Boston Research Group

© B

eFi F

orum

200

7