Are Risk Aversion and Impatience Related to Cognitive Ability Thomas Dohmen, Armin Falk, David Huffm

1
Are Risk Aversion and Impatience Related to
Cognitive Ability? Thomas Dohmen, Armin Falk,
David Huffman, Uwe SundeIZA Bonn and University
of Bonn01-07-07ESA World Meeting 2007LUISS
Rome
2
Motivation

• Risk aversion and impatience are crucial
  determinants of behavior in economic models, as
  is cognitive ability.
• Their importance is confirmed empirically
• Measures of risk aversion/ impatience predict a
  wide range of important economic behaviors.
• Cognitive ability matters for wages and other
  labor market outcomes.
• Typically, risk aversion and impatience are also
  assumed to be independent of cognitive ability.
• This assumption, however, has received relatively
  little attention in the empirical literature.

3
Psychology Literature

• Several psychology studies find that higher
  cognitive ability is associated with greater
  patience, although some others find no
  relationship.
• There has been little work on cognitive ability
  and risk-taking.
• Limitations of psychological evidence
• Typically studies have reported raw correlations,
  without controls for, e.g., education, income.
• Studies have been restricted to small samples of
  young children, or students, as subjects.
• Measures of risk aversion or impatience are often
  (but not always) hypothetical.

4
Behavioral Economics Literature

• Two recent studies provide evidence on a link
  between impatience or willingness to take risk
  and achievement test scores, or other more direct
  measures of cognitive ability
• Benjamin et al. (2005) conduct paid choice
  experiments with Chilean high school students,
  and relate choices to SAT scores.
• High school students with higher math SAT scores
  are more likely to be risk neutral, and more
  likely to choose patiently.
• This is true controlling for parental
  income/education.
• Frederick (2006) relates various measures of
  cognitive ability to patience and risk taking
• Higher cognitive ability associated with greater
  patience, and willingness to take risks in the
  gain domain (less willing in lotteries with
  losses).
• Subjects are college students and adults.
• Results are correlations without controls.

5
A Representative Field Experiment

• This paper contributes to this literature, in a
  way that complements and extends previous
  research.
• We provide the first evidence based on a
  representative sample
• 1,012 adults living in Germany.
• Roughly half participate in paid choice
  experiments measuring risk aversion.
• Roughly half participate in paid experiments
  measuring impatience over an annual time horizon.
• Experiments involve relatively high stakes.
• All participants take two tests of cognitive
  ability, corresponding to sub-models of
  widely-used IQ test.
• We address confounds that could affect our study,
  as well as previous studies, in a series of
  robustness checks.

6
Why is it important to answer this question?

• Relevant for the specification of structural
  models that include both cognitive ability and
  preferences.
• E.g., Heckman et al. (2006) are unusual in
  allowing the discount rate to be related to
  cognitive ability.
• Our study can shed light on the value of this
  flexible approach, and suggest a direction of
  correlation.
• Interpretation of reduced form models with
  cognitive ability on right-hand side but not risk
  aversion or impatience.
• E.g., regressing schooling on cognitive ability,
  without patience, leads to omitted variable bias.

7
Why is it important to answer this question?

• Important for understanding intergenerational
  transmission of preferences and socio-economic
  status.
• IQ known to be transmitted from parents to
  children.
• Risk aversion and patience are strongly
  correlated between parents and children.
• Dohmen et al. (2006), Knowles and Postlewaite
  (2005).
• Could transmission of IQ be a mechanism behind
  similar choices over time or under uncertainty?
• Ramifications for policy interventions focused on
  improving child IQ.
• Potential for important feedback effects between
  child IQ and preferences.

8
Data
9
Design

• Data were collected between June 9th to July 4th,
  2005.
• Our sample was drawn so as to be representative
  of the adult population living in Germany.
• We used the same company that conducts the
  surveys for the German Socio-Economic Panel
  (SOEP).
• As for the SOEP, interviewers start at a randomly
  chosen address, and then knock on doors.
• Interviews conducted in subjects homes.
• In total, our data include 1,012 participants.
• The interview had two parts
• First, subjects answered a questionnaire.
• Second, subjects were asked to participate in
  either a lottery experiment or inter-temporal
  choice experiment.
• Answers were typed into a laptop computer.

10
Design (continued)

• A random device in CAPI software determined
  whether the subject was asked to participate in a
  lottery experiment or an inter-temporal choice
  experiment.
• We deliberately had subjects participate in only
  one experiment to avoid order effects.

11
Cognitive Ability Measures

• Each of our two measures is designed to
  correspond to a different sub-module of the
  Wechsler Adult Intelligence Scale (German
  version).
• The WAIS is one of the most widely used IQ tests,
  with 6 verbal and 5 non-verbal sub-modules.
• One sub-module involves matching numbers and
  symbols with a 90 second time limit.
• Another sub-module involves a timed vocabulary
  test.
• Previous research shows that our tests correlate
  well with these sub-modules, other WAIS modules,
  and other prominent IQ measures
• (Lang et al., 2003 and 2007)

12
Symbol Correspondence Test

• Subjects were posed with 9 unfamiliar symbols,
  each paired with one of the digits 1 to 9.
• After brief instructions, subjects saw a new
  screen
• Mapping from numbers to symbols at the top.
• In the center of the screen, one of the symbols,
  and a box underneath in which the subject could
  type a number.
• After a subject entered a number, a new screen
  would appear, with another symbol.
• Subjects had 90 seconds to match as many as
  possible.
• 105 declined to participate, and procedural
  problems arose in a couple of cases, leaving 905
  subjects with non-missing scores

13
Word Fluency Test

• Subjects named as many animal names as they could
  in 90 seconds.
• After each naming, the experimenter pressed a key
  indicating
• Correct name
• Repeated name
• Incorrect name.
• Pilots with taped interviews indicate that
  experimenter error rates are relatively low.
• 87 declined to participate, a few more started
  and then decided to stop. Procedural problems
  arose in some cases, leaving 848 subjects with
  non-missing scores.

14
Distribution of Symbol Correspondence Scores
15
Distribution of Word Fluency Scores
16
Lottery Experiment

• Subjects made 20 choices in a choice table.
• The choice in each row was between a lottery and
  a safe option.
• Lottery 300 or 0 Euros with equal probability.
• Safe option X Euros, where X varies across 20
  choices.
• Subjects knew that at the end, one row would be
  randomly selected, choice in that row
  implemented.
• Switching row in the choice table is an incentive
  compatible measure of certainty equivalent.
• Switching later in the table indicates a greater
  willingness to take risks.

17
Choices in the Lottery Experiment
18
Inter-Temporal Choice Experiment

• Subjects made 20 choices in a choice table.
• The choice was always between a payment available
  Today and a larger amount available in one
  year.
• Early payment 100 Euros.
• Delayed payment 100X Euros, where X varies
  across 20 choices.
• Subjects knew that at the end one row would be
  randomly selected, choice in that row
  implemented.
• The payment would arrive in the mail as a check,
  cashable either immediately or in one year.
• Switching row in the table measures impatience
  the rate of return needed to induce waiting one
  year.
• Switching later indicates greater impatience.

19
Choices in the Inter-Temporal Choice Experiment
20
Results
21
Behavior and Symbol Correspondence Scores
22
Behavior and Word Fluency Scores
23
Baseline Regressions Willingness to Take Risks
Notes Interval regression coefficients, marginal
effects. Robust s.e..
24
Baseline Regressions Willingness to Take Risks
Notes Interval regression coefficients, marginal
effects. Robust s.e..
25
Baseline Regressions Impatience
Notes Interval regression coefficients, marginal
effects. Robust s.e..
26
Baseline Regressions Impatience
Notes Interval regression coefficients, marginal
effects. Robust s.e..
27
Robustness
28
Education

• Cognitive ability is associated with achievement
  of more advanced schooling degrees.
• This is true in our data, and in the literature.
• Education, in turn, might affect risk aversion or
  impatience.
• Thus, it is interesting to see whether cognitive
  ability has a direct effect, or works only
  indirectly through education.
• Another issue arises because education could
  affect cognitive ability (e.g., Cascio and Lewis,
  2006).
• Is cognitive ability simply a proxy for education
  in the baseline results?

29
Risk Results are Robust to Adding Education
Notes Interval regression coefficients, marginal
effects. Robust s.e..
30
Impatience Results Robust to Adding Education
Notes Interval regression coefficients, marginal
effects. Robust s.e..
31
Income and Credit Constraints

• Cognitive ability is associated with higher
  income.
• Income could have an impact on risk-taking
  behavior, or potentially patience.
• In the lottery experiment, income cushions the
  impact of losing a lottery.
• In the inter-temporal choice experiment, low
  income could imply credit constraints, making the
  choice100 Euros today a necessity.
• It is interesting to see whether cognitive
  ability only works through the indirect channel
  of income.
• Also, we use a survey measure of credit
  constraints
• If you suddenly encountered an unforeseen
  situation, and had to pay an expense of 1,000
  Euros in the next two weeks, would you be able to
  borrow the money?

32
Risk Results are Robust to Adding Income
Notes Interval regression coefficients, marginal
effects. Robust s.e..
33
Impatience Results are Robust to Adding Income
Notes Interval regression coefficients, marginal
effects. Robust s.e..
34
Confusion about Incentives?

• One potential confound could arise if low
  cognitive ability caused people to be confused
  about incentives.
• This would be a problem if confusion lead to
  behavior that happened to look like, e.g.,
  greater risk aversion.
• For risk attitudes we have a way to address this
  issue explicitly.
• We use a very simple survey measure that asks
  about willingness to take risks
• In general, are you a person who is fully
  prepared to take risks, or do you try to avoid
  taking risks?
• Respondents answer on a scale for 0 to 10, where
  0 indicates completely unwilling and 10
  completely willing

35
Cognitive Ability also Related to Survey Measure
Notes Interval regression coefficients, marginal
effects. Robust s.e..
36
Other Robustness Checks

• Could Test Scores Proxy for Personality?
• Arbitrage between experiment and market rates of
  return.
• Risk aversion, impatience, and test-taking
  strategy.
• Time preference vs. concavity of utility (risk
  aversion) as determinants of impatient behavior.

37
Conclusions

• We find a significant and robust relationship
  between cognitive ability, risk aversion, and
  impatience.
• People with higher cognitive ability are more
  patient, and more willing to take risks.
• Regardless of the precise mechanism, it is
  important to know that cognitive ability, risk
  aversion, and impatience are systematically
  related.
• Implications for specifying econometric models.
• Interpretation of reduced form models.
• Intergenerational transmission.
• Inequality and IQ.
• Implications for policy interventions focused on
  improving child IQ.

38
Conclusions

• For future research establishing underlying
  mechanisms.
• Bounded rationality, in the sense of choice
  bracketing?
• Cognitive ability could affect whether people
  bracket narrowly, making risky choices in
  isolation, or recognize that they form part of a
  larger portfolio.
• Cognitive ability could also affect whether
  people integrate present and future
  considerations.
• Two-system explanation?
• Cognitive system in the brain can be overpowered
  by affective system affective system is the
  source of urges for immediate consumption, fear
  of losses.
• Cognitive ability could proxy for the resources
  an individual has to suppress emotional urges.
• Feedback from preferences to development of
  cognitive skills?
• Although cognitive ability may affect risk
  aversion and impatience, there could also be
  important feedback effects.
• E.g., patience conducive to developing cognitive
  skills.
• Evolution
• Adaptive for low-cognitive ability to go with
  conservative strategies of risk aversion and
  grabbing immediate rewards?

39
Thomas Dohmen IZA, P.O. Box 7240, 53072 Bonn,
Germany Phone 49 (0) 228 - 38 94 -204 Fax 49
(0) 228 - 38 94 180 E-mail dohmen_at_iza.org http/
/www.iza.org