Cross-Cultural factors and Portfolio Choice
Daniel Egan, Greg Davies, Peter Brooks
Barclays Wealth Behavioural Analytics
FUR Conference
01/07/2008
2
Building Individual Portfolios
Irrational
Cultural
Individual
Circumstantial
How do we interpret and use variation in behaviour and preferences?
?
3
Survey and Dataset
UK SG TW HK CN INN= 2,468 353 398 415 443 389 468
Male % 63 51 69 63 71 57Age- mean (sd) 41 (10) 38 (9) 39 (9) 40 (8) 34 (7) 33 (10)
Preliminary filterearned equal to or greater than the top 10th local percentile of income earners reported investable wealth equal to twice the top 10th local percentile of income
SurveySection I - DemographicsSection II - 80 questions relating to investment and personality traits.
All questions in Section II were randomised.English in the UK, Singapore, and IndiaLocal script in Hong Kong, Taiwan, and China (PRC)
Validation Filter-- Must have taken reasonable time to complete survey -- Must have answered a few questions consistently
-not preferring a 5% return over a 20% return or a 15% return. - The total effect was to remove an average 10% of each locations sampled
Key Variables
4
Risk Tolerance
5
Have to fix order to be consistent!!
Risk Tolerance Score Risk Tolerance Profile
Portfolio Choice
£0
£5,000
£10,000
£15,000
£20,000
£25,000
£30,000
£35,000
£40,000
Portfolio1 Portfolio2 Portfolio3 Portfolio4 Portfolio5
High
Most Likely
Low
6
“The chart shows the high, low and most likely final values of £12,500 invested in 5 different portfolios for 5 years. For example, in Portfolio 1 you will get £13,500 and in Portfolio 5 you end up getting anything between £7,500 and £34,000, but the most likely amount is £19,000. Which portfolio would you prefer?”
Which portfolio would you prefer? No large differences across nations
Risk Perception
£0
£5,000
£10,000
£15,000
£20,000
£25,000
£30,000
£35,000
£40,000
High
Most Likely
Low
7
The chart shows the high, low and most likely final values of a £12,500 investment in 5 years. Please rate how risky you think this investment is.”
How risky is this investment? That depends...
Monthly returns Jan 2000 – June 2008
8
Return Expectations
9
Classifying expectations Shows what “good” returns are
Excellent
Good
Neither bad
nor good
Poor
Terrible
Bivariate analysis
Predicting Portfolio Choice
10
Risk tolerance
11
Portfolio Choices by Risk Tolerance Score
Least risky
Most Risky
** Statistically significant differences in mean portfolio choice across Risk Tolerance Profiles
Risk Perception
12
Mean Portfolio Choice by Risk Perception
Higher return expectations do drive riskier choice
13
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Portfolio 1 Portfolio 2 Portfolio 3 Portfolio 4 Portfolio 5
High Expectations
Low Expectations
Responsive Expectations
Middle Expectations
Returns Expectations and Portfolio ChoiceCumulative Choice
Cumulative distribution of each returns expectation profileacross choices of portfolio
Comparative explanatory power
Predicting Portfolio Choice
14
Baseline Difference in Portfolio Choice
15
UK SG TW HK CN IN
Difference from UK
Allow effect of Risk Tolerance to vary
16
UK SG TW HK CN IN SG TW HK CN INUK
Intercepts
Difference from UK Difference from UK
Allow effect of Risk Perception to vary
17
UK SG TW HK CN IN SG TW HK CN INUK
Intercepts
Difference from UK Difference from UK
Intercepts
Allow different Returns Expectations
18
UK SG TW HK CN IN
Difference from UK
Mod High Resp
All together now!
19
RT
Baseline Main Effects Interactions
Conclusions
Controlling for: Risk Tolerance; Risk Perception; Returns Expectations; Country-specific effects;
Only China maintains base-level difference in portfolio choice
Risk Tolerance always predictive Singaporeans more sensitive
20
Risk Perception not predictive
Except China – negative!
Returns Expectations are! Especially when interacted
with Risk Tolerance
Fin.
21
dd month year
22
Social support network effects perceptions, not choices.
dd month year
23
Risk perception and social support Portfolio choice and social support
"If things went wrong financially, I could rely upon the support of my family and friends."
"If things went wrong financially, I could rely upon the support of my family and friends."
dd month year
24
dd month year
25
A progression of cultural differences
A Similarity index (UK based) reveals an intuitive progression
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
SG TW HK IN CN
The more mature the economy and market, the more it resembles the UK
The more dislike the UK, the higher the variance of its measurements
Indicates lack of stability in market expectations and understanding
Two models of the investment decision
multi-attribute representation
allows better comparison of constituent attributes
requires reconciliation into single decision node
dd month year
26
Source:
Bottom-upPercieved RiskSocial SupportRecency Bias
Subjective Benefit* Returns Expectations* E([U(X)]
Risk/Return Risk ToleranceRisk PreferenceAffective Appraisal
Portfolio Choice
Objective Risk•Possibility of loss•Mass of loss•Maximum loss* Variance
Objective Benefit* E(X)?•E[U(X)]?•Max(X)?
ProspectivePortfolios
Top-down
affective reaction
non-compensatory
results in transitive preferences
27
What and why are there cultural differences?
Risk Preferences Risk Perception Probabilistic Calibration Trading Behaviour
Overconfidence
Social network support (Weber and Hsee)
Calibration Base-level risk Market Maturity (risk/return
trade-off) Utility of Wealth
Source:
Documented Differences Explanation
Income Risk
28
Income Risk? Income Risk
29
0%
10%
20%
30%
40%
50%
60%
70%
Risk Tolerance
62% of the sample had no change in their allocated Risk Profile over 4 months
No changes of more than 1 Profile
The majority who changed were within 3 points of the profile boundary
The downturn in markets allowed us to test the scales market sensitivity
There was only a minor increase of 1.9% in the “Low” and “Medium-Low” Risk” profiles from June, 2007
Stability
Good Risk Tolerance scales are also stable
over time, and through market conditions. Risk
Tolerance measurements may have some natural
variance in them, but not a large amount
Testing the Scale
Our scales performed extremely well, being
stable both across time and changing market conditions.
This implies we were measuring a real underlying trait,
rather than mood or sentiment
Stable Over Time
Immune to Market Conditions
0%
5%
10%
15%
20%
25%
30%
35%
40%
June 2007
March 2008
Down One No Change Up One
Low Med-Low Moderate Med-High High
Return / Volatility Environment
dd month year
30
Social network support
dd month year
31
Risk Tolerance across all locations
dd month year
32
UK SG TW HK CN IN
UK SG TW HK CN IN RTS
Intercepts
Intercepts
Risk Perception
dd month year
33
UK SG TW HK CN IN Risk Perception