The Intergenerational Transmission of Risk and Trust Attitudes

1
The Intergenerational Transmission of Risk and
Trust Attitudes
Presented by David Huffman (IZA Bonn)

• Joint work with
• Thomas Dohmen (IZA Bonn)
• Armin Falk (IZA Bonn)
• Uwe Sunde (IZA Bonn)

2
Primary Research Question for this Paper

• Are the preferences and beliefs that drive
  economic decision making transmitted from parents
  to children?
• We focus on two particularly important elements
  of the decision makers endowment
• Willingess to take risks.
• Willingness to trust other people.
• We use the term attitude to refer collectively to
  the preferences and beliefs that go into
  decisions to take risks, or decisions to trust.

3
Importance of Risk and Trust Attitudes

• Almost every economic decision involves risk.
• E.g., financial investments, educational
  attainment, home ownership.
• Trust is similarly important, but in social
  interactions
• Most interactions involve the possibility of
  being taken advantage of.
• Trust determines behavior in these situations,
  and willingness to enter them in the first place.
• Virtually every commercial transaction has
  within itself an element of trust . (Arrow,
  1972, p.357)

4
Importance of Intergenerational Transmission

• Sheds light on where preferences and beliefs come
  from.
• Possible mechanism underlying persistent
  similarities in behavior across generations
• Culture Fertility and work practices,
  preferences regarding ethnicity/race of marriage
  partners.
• E.g., Bisin and Verdier (2000).
• Socioeconomic status Intergenerational
  correlation in education, wealth, etc.
• E.g., Solon (1992).

5
Importance of Intergenerational Transmission

• Macro implications persistent differences across
  countries slow reaction to institutional change.
• For instance, persistent differences in trust,
  which have an impact on economic growth.
• E.g., Knack and Keefer (1997).
• Cross-country differences in willingness to take
  risks.
• E.g. Fehr et al. (2006) document difference in
  risk attitudes between US and Germany.

6
Secondary Research Questions

• What determines the strength of parents
  influence on the child?
• Gender of child, personality, number of siblings,
  birth order?
• How fine-tuned is the transmission process?
• Are children like their parents only in terms of
  a general disposition, or are they similar in
  relatively detailed ways?
• Is there positive assortative mating of parents,
  based on risk and trust attitudes?
• I.e., do parents tend to marry partners with
  similar attitudes to their own?
• If so, this reinforces the impact of parents on
  children.

7
Methods

• Survey questions asking about willingness to
  take risks degree of trust in other people.
• Appropriate because can be administered to a
  large sample at relatively low cost.
• Our questions have been tested using large
  experiments, shown to predict actual behavior.

8
Data
9
German Socioeconomic Panel (GSOEP)

• GSOEP is a panel survey, starting in 1989.
• Annual wave 22,000 individuals (age 17), 12,000
  households
• Representative of the population
• Possible to link childrens information with
  parents.
• 3,595 children with both parents, n 10,785
• Oldest child 57 Ave. child 25.3 Ave. parent
  53.
• Household members are interviewed separately,
  ruling out collaboration of family members.
• Extensive socio-demographic information
• Individual characteristics
• Family and household background

10
Risk measures

• General risk question
• How do you see yourself Are you generally a
  person who is fully prepared to take risks or do
  you try to avoid taking risks? Please tick a box
  on the scale, where the value 0 means unwilling
  to take risks and the value 10 means fully
  prepared to take risks.
• Questions about willingness to take risks in
  specific contexts
• Health, car driving, financial matters, sports
  and leisure, career.
• Hypothetical lottery
• 50/50 chance to double investment or lose half.
• Possible to invest up to 100,000 Euros.

11
Experimentally validated (Dohmen et al. 2005)

• 450 subjects, representative sample of adult
  population
• Ss answer the general risk question
• Ss also take part in paid lottery experiment
• Lottery 300 or 0 Euros, with equal probability
• Safe option X Euros, where X varies across the
  20 choices
• One choice is randomly selected and implemented
• Incentive compatible measure of risk preference
  switching point from the lottery to the safe
  option
• Ordered Probit regressions
• Risk question predicts switching point, n450,
  extensive socio-economic controls

12
General Risk question and risk taking behavior
13
Additional Evidence on Validity

• General risk question predicts
• Holding stocks, self-employed, public sector,
  smoking, East/West Migration, active sports.
• Occupational sorting Mobility (Bonin et al.,
  2006a,2006b).
• Mincer wage regression for each point on
  11-point scale, 2 percent higher log monthly
  wages .
• Context-specific questions predict
• Health smoking (12.8 percent).
• Financial matters investment in stock (24.4
  percent).
• Sports and leisure practicing sports (18.4
  percent).
• Career self-employment (1.8 percent).
• Hypothetical lottery predicts choices in lab
  experiments.
• Student subjects design as in Dohmen et al.
  (2005).

14
Trust measure

• Survey measure with three sub-statements
• In general, one can trust people (trust)
• In these times you cant rely on anybody else
  (unreliance)
• When dealing with strangers it is better to be
  careful before you trust them (caution)
• Four answer categories
• Agree fully, agree somewhat, disagree somewhat,
  disagree fully.
• We collapse an individuals responses to the
  three statements into a single indicator of
  trust.
• The method is principal component analysis.

15
Experimentally validated (Fehr et al. 2003)

• Trust game, representative sample, n429.
• First-mover decides how much of 10 Euros to pass.
• Passed points are doubled, given to second-mover.
• Second-mover decides how much to return (not
  enforceable).
• Subjects answer the same three trust questions
  that we use.
• Significant correlation between trust principal
  component and first-mover transfer amount
  (p-value lt 0.044).
• Trust PCA also predicts transfer amount in
  regressions, controlling for other observables.

16
Variation in Parents Risk Attitudes
17
Variation in Parents Trust
18
Gender and General Risk Attitude

19
Age and General Risk Attitude

20
Age and General Risk Attitude

21
Height and General Risk Attitude
22
Height and General Risk Attitude
23
Results
24
Outline of Results

• Transmission
• Risk attitudes
• Trust attitudes
• Heterogeneity in transmission
• Gender of child
• Family structure
• Receptive personality?
• Specificity of transmission
• Assortative mating

25
Childs Risk Attitude as a Function of Parents
Notes Weighted regression lines, taking into
account number of parents in a given category.
26
Childs Risk Attitude as a Function of Parents
Notes OLS estimates robust s.e. clustering on
household.
27
Childs Risk Attitude as a Function of Parents
Notes OLS estimates in (1) to (6) interval
regression (Tobit) for (7) robust s.e.
clustering on household.
28
Childs Trust as a Function of Parents
Notes Weighted regression lines, taking into
account number of parents in a given category.
29
Childs Trust as a Function of Parents
Notes OLS estimates robust s.e. clustering on
household.
30
Heterogeneity

• Overall, children are similar to their parents,
  but alignment is stronger for some than for
  others.
• We run regressions interacting parents attitudes
  with characteristics
• Gender of the child
• Number of siblings of the child
• Child is the first-born
• Gender of the child does not matter there is
  some evidence that family size/ birth order do
• Children in smaller families are more similar to
  parents in terms of risk attitudes first-born
  are more strongly influenced than later born
  (marginal).
• For trust, family size and birth order do not
  matter.

31
Heterogeneity

• Are some children especially influenced by
  parents, across the board?
• We find some evidence of such receptive types
  (or, influential parents)
• Children who are especially similar to their
  mother, in terms of the difference in attitudes,
  are also more similar to the father.
• Children who are especially similar to parents in
  terms of risk are also very similar in terms of
  trust.
• Results are based on correlations, e.g., child
  risk - parent risk and child trust - parent
  trust.
• Robust to controlling for observables.

32
Specificity of the Transmission Process

• Are children like their parents only in terms of
  a general disposition?
• Or are children similar to their parents in even
  more detailed ways?
• We use our context-specific risk questions, and
  individual trust questions
• Answers are highly correlated across questions.
• But not perfect e.g., a parent may be risk
  averse in car driving, but even less willing to
  take risks in financial matters.
• We regress a childs attitude in a given context
  on parents attitudes in all of the different
  contexts, simultaneously.

33
Specificity of the Risk Transmission Process
Notes OLS estimates robust s.e. clustering on
household.
34
Specificity of the Trust Transmission Process
Notes OLS estimates robust s.e. clustering on
household.
35
Assortative mating

• Is there positive assortative mating based on
  risk or trust attitudes?
• If so, this reinforces the impact of parents on
  children.
• A priori, not clear whether optimal to marry
  someone with the same, or opposite attitude.
• Depends on whether these attitudes are
  substitutes or complements in household
  production.
• There is already evidence that partners tend to
  be similar in terms of various behaviors
  (plausibly related to risk).
• E.g., similar preferences regarding smoking,
  leisure activities.
• No evidence at the level of fundamental attitudes.

36
Assortative Mating Risk and Trust Attitudes
Notes OLS estimates robust s.e. clustering on
household.
37
Summary

• We document a robust intergenerational
  correlation in willingness to take risks and
  trust.
• Our evidence is based on experimentally-validated
  measures that predict actual risk taking and
  trusting behavior.
• There is some evidence that family size and birth
  order matter for the strength of transmission,
  for risk.
• Evidence of receptive types.
• Transmission process is relatively fine-tuned.
• Reinforced by positive assortative mating.

38
Concluding Remark

• An important question that we have not answered
  what is the underlying mechanism?
• Genetics.
• Intentional effort of parents to shape the
  attitudes of the child.
• Unintentional transmission (imitation by
  children) .
• Combination of all of these may interact.
• To provide decisive evidence we would like data
  with a large number of adopted children, or
  twins
• Regardless of the precise mechanism, it is
  important that children end up like parents in
  terms of willingness to take risks, and
  willingness to trust.

39
Ongoing Research on Preferences

• Example 1 Cognitive ability and time/risk
  preferences
• Cognitive ability and risk and time preference
  are usually thought of as distinct traits.
• But this assumption has received relatively
  little attention in the empirical literature.
• Low cognitive ability could lead to limited
  ability to integrate small-stakes risks with
  lifetime income think about the future.
• Low cognitive ability could indicate limited
  resources for supressing emotional impulses urge
  for immediate gratification fear of risks.
• A link between cognitive ability and time/risk
  preferneces has far-reaching implications
• Specification of structural models.
• Interpretation of reduced form models.
• Changing child IQ also has an impact on
  preferences?

40
(No Transcript)
41
Ongoing Research on Preferences

• Example 2 Hyperbolic discounting and
  self-control problems in a representative sample.
• Most evidence of hyperbolic discounting comes
  from experiments with small samples of college
  students.
• In principle, HD can lead to self-control
  problems, in the sense of making and breaking
  resolutions regarding own behavior.
• No-one has measured hyperbolic discounting in a
  representative sample, and tested whether it
  predicts a self-reported tendency to make and
  break resolutions.
• We use a representative sample of 500 German
  adults
• Incentive compatible measures of discounting
  type, and Survey questions about making and
  breaking resolutions.
• How prevalent is hyperbolic discounting?
• Does hyperbolic discounting in fact predict a
  pattern of making and breading resolutions in
  daily life?

42
Prevalence of HD in the Population
43
HD and Making/Breaking Resolutions
Notes Censored observations excluded. Robust
s.e. in parentheses.
44
Larger Research Agenda on Preferences

• We are interested in measuring and understanding
  fundamental preferences (risk, time, fairness,
  etc).
• Heterogeneity
• Determinants
• Impact of preferences on behavior
• Preference are usually taken as given
• Part of the residual in our models
• Or backed-out using functional form assumptions
• We are using a combination of methods to measure
  preferences
• Survey questions asking about preferences
• Paid choice experiments
• Combination

45
Interaction Between Risk and Trust
Notes OLS estimates robust s.e. clustering on
household.
46
Example Falk/Zehnder (2006)

• Trust game with 986 Subjects, representative
  sample of Zurich.
• Profit first mover 20 investment back
  transfer
• Profit second mover 20 3investment back
  transfer

Mean profit (net the endowment)
47
Correlation in partners observable
characteristics
48
Similarity in Schooling Degrees
49
Religion of partners
50
Characteristics of Residence Location up to age 15
51
Summary of robustness checks on restricted samples
52
Data

• German Socio-Economic Panel
• 7,272 couples with non-missing information on
  preferences
• General risk and trust measure as before

53
Concluding Remarks

• Tip of the iceberg parents may be important for
  other fundamental determinants of behavior as
  well
• Time preferences?
• Reciprocity we find some initial evidence of an
  intergenerational correlation, using (not yet
  validated) SOEP questions.