French savers in the “great recession” Preferences, financial expectations, and portfolio choice
Luc Arrondel & André Masson CNRS-PSE
2
Risky portfolio in France
Source: Tns-Sofia
3
Mars 2013 : 8,0% Mars 2014 : 7,2% Mars 2015 : 6,6% Mars 2016 : 6,2%
Risky portfolio in France 6000
20
Cac40
% ac4onnaire individuel
5000
3770
4000
3500
13.0
4385
4291
4067
3000
18
4917
16 14
3700
12
11.9
10
2581
9.1 2000
8.5
8
8.0
7.3
6.6
6.2
6 4
1000
2 0
0 Mars 2009
Mars 2010
Source: Tns-Sofia 4
Mars 2011
Mars 2012
Mars 2013
Mars 2014
Mars 2015
Mars 2016
Structure du portefeuille de valeurs mobilières et d’assurance vie par centile de richesse financière 3.4a. Taux de détention des valeurs mobilières et des assurances vie par niveau de richesse financière
Percentiles 0-25 25-50 50-70 70-90 90-99 99-100 Ensemble
Actions en direct
FCPActions
Obligation ou FCPObligation
0.004 0.041
0.002 0.010
0.000 0.001
0.105
0.028
0.011
0.256 0.492
0.085 0.184
0.038 0.100
0.690
0.374
0.195
0.135
0.001 0.003 0.024 0.029 0.078 0.115
0.046
0.021
0.020
Autres valeurs mobilières
Ass. Vie multisupport 0.010 0.047 0.096 0.215 0.427 0.610
Ass. Vie en euros 0.054 0.165 0.265 0.411 0.478 0.563
Bons de Capitalisation
0.121
0.239
0.004
2010
0.000 0.001 0.004 0.007 0.016 0.039
2014
5
Source: Insee, enquêtes « Patrimoine »
Safe portfolio in France
Septembre 2008
6
Related Literature The questions to measure RA : qualitative and lottery Risk aversion increases following the 2008 financial crisis
7
Related Literature The explanation : emotional response (FEAR)
8
Related Literature
9
Related Literature The question to measure RA : Likert scale Risk aversion increases following the 2008 financial crisis
10
Related Literature The question to measure RA : qualitative question about financial risk
11
Outline of the presentation § Data: May 2007-June 2009 -November 2011-December 2014 §
Panel dimension
1. Individual wealth behaviors: impact of the crisis §
Δ behaviors = g (Δ preferences, Δ present resources, Δ expectations)
2. Changes in future income and asset price expectations 3. Changes in individual risk, time & other preferences §
Usual measures: lottery choice, Likert scales
4. Measuring individual preferences: scores §
Method of scoring
§
Preference scores: determinants, correlations, wealth effects
6. Explaining changes : 12
§
Scores: only age effect on risk & time preferences
§
Changes in behaviors explained by changes in expectations
Data
Available surveys
Pat€r surveys
14
Cac40 and Pater survey
May 2007
Nov. 2011 Dec. 2014 June 2009
15
Example of decomposition of changes in behavior § Investment choice between: § a risky asset: expected return m & standard deviation σ § & a safe asset of return r
§ Expected utility: share of risky asset p è p = (m–r) / σ2γ . Change in p may depend on: § Changes in preferences: relative risk aversion γ § Changes in price expectations concerning the risky asset (m, σ) – or even the riskless asset (r)
§ If background risk on labor income added: share p lower if the saver is ‘temperant’ (substitution of risks) § Change in p may also come from changes in labor income risk § If present risk exposition in labor income increases, the share p decreases (if the saver is temperant: 4th derivative of the felicity function) 16
1. Changes in behaviors
« As regards financial investments, do you think that… » Pater 2007, 2009 2011 & 2014 (panel) : in % (N = 807) Question similar to Guiso et al. (2015)
18
2014: return to 2009
Risky portfolio in France (stocks and shares)
Source: Tns-Pater 19
2. Changes in expectations
Anticipated return on stock market within the next 5 years The question: distribute 100 points Five years from now, do you think that the stock market... (For each category write down the likelihood of occurrence assigning a value between 0 and 100. The sum of all your answers must be equal to 100) will have increased by more than 25%
/__/__/__/
will have increased by 10 to 25%
/__/__/__/
will have increased by less than 10%
/__/__/__/
will be the same
/__/__/__/
will have decreased by less than 10%
/__/__/__/
will have decreased by 10 to 25%
/__/__/__/
will have decreased by more than 25%
/__/__/__/
In your opinion, if you expect the stock market to increase within the next 5 years, which is the highest possible increase (as a percentage)?
21
In your opinion, if you expect the stock market to decrease within the next 5 years, which is the lowest possible decrease (as a percentage)?
Anticipated return on stock market within the next 5 years Pater 2007, 2009 & 2011 (panel : ½ respondents = around 1,000 in each case) 60%
Average anticipated return : 2007 : 5.6%
56% 2007 (mean=5.6%)
50%
48%
2009 (mean=3.6%)
2009 : 3.6%
40%
2011 : 0.4%
31% 30% 30% 22% 20% )!"#
13%
%!!*#+,-./0&12"3#
10% (!"#
0% Negative
Zero
Positive
Panel 2007-2009 Panel 2009-2011
'(#$
%!$$#+,-./0!1'"3#
'!"#
%)#$ %$ %!#$
&!"#
%%#$
!"#$ %!"#
$!"#
22
!"#
*+,-./0+$
1234$
546/./0+$
Anticipated return on stock market within the next 5 years Pater 2011 & 2014 (panel : ½ respondents = around 1,500 in each case)
Average anticipated return : 2011 : 0.4% 2014 : 1.4%
Inversion: 2014 expectations more optimistic than in 2011 but still less than in 2009
23
Determinants of stock price expectations (Pater 2007) § Expected return § Men are more optimistic than women (+1,1%) § The amount of wealth: positive effects § ‘Information’ increases expectations => expected return rises with: § Age § Economic & financial press reading § Level of financial education § ‘Good in mathematics’, ‘luck in life’: positive effects, § Executives are more optimistic, self-employed are more pessimistic § Past capital gains have a positive effect
§ Expected risk § Lower among the elderly, those without a diploma, who do not read economic & financial press § Negative effect of past capital gains 24
24
3. Changes in preferences: usual measures
The hypothetical income lottery measure of risk aversion Barsky, Juster, Kimball and Shapiro (1997) Barsky, Juster, Kimball and Shapiro (1997) §
Suppose that you have a job which guarantees for life your household’s current income R. Other companies offer you various contracts which have one chance out of two (50%) to provide you with a higher income and one chance out of two (50%) to provide you with a lower income. Do you accept?
26
The income lottery (continued) Barsky, Juster, Kimball and Shapiro (1997)
Contract A no
yes
1/2
1/2 R R
1/2
R 0.5 R
27
R
R 1/2
4/5 R
R
The lottery measure of relative risk aversion Panel PATER 2007, 2009, 2011 and 2014 (N=807) The rational consumer chooses the contract if E u = u(2c) + 1/2 u(λc) ≥ u (c) Hypothesis: expected utility maximization, u is CRRA
28
Distribution across waves of the lottery measure Lottery measures are quite unstable over time K-S test (2007-2009) : 0.0841
29
(p=0.000)
Likert preference scales (self-assessed) Time preference : On a scale of zero to ten, where would you place yourself between the following two "extreme" descriptions ? 0 : persons who live day by day and take life as it comes, who don't think too much about tomorrow nor worry about the future 10 : persons who are preoccupied by their future (even their distant future) and whose mind is well set on what they want to be or do later in on life Live day by day
Preoccupied by the future
[___0___Í___1___Í___2___Í ___3___Í ___4___Í ___5___Í ___6___Í___7___Í___8___Í ___9___Í ___10___] Impatience : On a scale of zero to ten, where would you place yourself between the following two "extreme" descriptions ? 0 : persons who are impulsive or impatient, who can't stand waiting, who are quick to react, and who want everything right away 10: persons who are poised and thoughtful, who know how to grin and bear it, and who need time before forging an opinion or taking a decision Impatient
Patient
[___0___Í___1___Í___2___Í ___3___Í ___4___Í ___5___Í ___6___Í___7___Í___8___Í ___9___Í___10___] Risk : As regards your attitude towards risk on the whole, where would you place yourself, using the same scale ? 0 : corresponds to very prudent persons, who seek to limit the risks of life as much as possible, and who aspire to a well-organised life, without surprises 10 : corresponds to persons drawn by adventure, who seek novelty and challenges, and who like to take risk and have got a lot at stake in their lives Prudent
Adventurous
[___0___Í___1___Í___2___Í ___3___Í ___4___Í ___5___Í ___6___Í___7___Í___8___Í ___9___Í ___10___] 30
Scales of preferences between 2007, 2009, 2011 & 2014 According to scales, people become (ref. 2007) … a little more patient
Panel PATER 200, 2009, 2011 & 2014 (N=807)
K-S test (2007-2009) : 0.0899
as farsighted or so less risk tolerant (until 2011) overall and in any domain of life 31
Risk by domain
(p=0.000)
Histogram of the preference scales Risk
Time preference
Defects: - anchor points (value 5) 32
- irregular spikes from one year to the other
4. Measuring preferences: individual ordinal scores
Our alternative approach (1): scores § Build in small touches the psychological profile of the saver § Multiply the number of simple questions : lotteries, opinions and intentions, real-life choices & actual behaviors, possible scenarios… § Since no one single question is satisfactory
§ …on different areas of life : consumption, leisure, health, investments, work, retirement, family… § considering different risks (big, small, gains, losses…) & time-horizons
§ Individual scores for preferences towards risk, time and family § A priori attribution of questions to preference parameters § Coding the questions (-1, 0, 1): scores are the sum of the answers given § as “averages” of the responses given, on a purely ordinal basis
§ Validation of ordinal measures § Internal consistency : Cronbach alpha, correlation of “sub-scores” in different domains Principal component analysis, etc. 34
Our alternative approach (2): virtues of aggregation Problem : no one question is fully satisfactory
§ Aggregation in synthetic, relative scores could be the answer § If a common component exists, such ‘aggregation’ of questions reduces ex ante framing effects, noise & measurement error, endogeneity biases § Score = context-free measure § Score = summary measure (risk : large & small risks, gains & losses…) § Score = partly a collection of natural instruments § Score = significant explanatory power of wealth amount & composition
§ Data choose the scores for each preference § Only one (internally consistent) risk score § Two preferences scores concerning the horizon
35
Ø
Short-term impatience
Ø
Time preference for the present over the life-cycle
§ One score for family altruism (children)
Risk aversion Ambiguity aversion Loss aversion, etc.
Top of the risk 1. French and weather
The question : “Do you take an umbrella with you when the weather forecast is uncertain?” ?
53% 45%
36
Yes
No
Top of the risk 2. French and their car
The question : “Do you park your car illegally?” Very Often “Do you park your car illegally?” 37
1
Often Sometime
2
10
Rarely
33
I have Never no car 42
12
Top of the risk « Cigarettes, whisky… » The question : “Do you think that it is worth depriving yourself of some of life’s pleasures to gain a few extra years of life? ”
Yes, quite
38
17%
Yes, rather
37%
Not really
37%
No at all
6%
DK
3%
George Best, a particular idea of money management ..
• « I spent a lot of money on booze, birds and fast cars. The rest I just squandered. » •
39
« In 1969 I gave up
women and alcohol - it was the worst 20 minutes of my life. »
Top ten questions contributing to the risk score
Similar ranks of the questions in the risk score (ref. 2009)
40
Scores much better than other measures on… Scores correlated with all other survey & experimental measures
§ Important consistency (Cronbach=0.7) § Nicer histograms… èè § Explaining risk measures by standard covariates (2009) § Pseudo R2 : 7.4% score ; 1.3% lottery ; 0.9% general scale
§ Homogamy in couples, (rank) correlation (2007) § Risk : 0.52 score ; 0.19 general 0-10 scale ; 0.26 lottery (Kimball: 0.41) § Time preference : 0.45 score ; 0.25 for 0-10 scale
§ Intergenerational (rank) correlation (2002) § Risk : 0.22 score ; 0.08 n.s. lottery (Kimball: 0.20) § Time preference, altruism : 0.15 scores
§ Stability over time (see below) § Robustness of wealth effects & score correlations across dates 41
Histogram of the preference scores Risk
42
Time preference
Time correlations of preference measures across waves
43
5. Scores: overall stability of risk & time preferences
Risk score : histogram in 2007, 2009 & 2011 Pater 2007-2009 (Panel) & 2009-2011 (Panel) 2007 (mean=6.5)
140
More risk averse/prudent
2009 (mean=6.3) 120
100
140
More risk averse/prudent 80
2009 (mean=6,7) 2011 (mean=6,8)
120 60
More risk averse/prudent
100
40
K-S test :
20
0.0299
80
(p=0.269) 60
40
K-S test : 0.0163 (p= 0.956)
20
2009 : mean=6.7 median=7 2011 : mean=6.8 median=7 45
Source: Panel PATER 2009-2011 (N=1,970)
25
23
21
19
17
15
13
9
11
7
5
3
-1
-3
-5
-7
1
-9
3
-1
-1
5
0 -1
2007 : mean=6.5 median=7
1
Source: Panel PATER 2007-2009 (N=2,234)
26
22
24
20
18
16
12
14
8
10
6
2
4
0
-2
-4
-6
0
-8
2
-1
-1
6
4 -1
8
-1
-1
-2
0
0
Risk score : histogram in 2007, 2011 & 2014 Pater 2011-2014 (Panel) & 2007-2014 (Panel) More risk averse/prudent
K-S test : 0.0372 (p= 0.102)
Source: Panel PATER 2011-2014 (N=2204)
2011 : mean=6.3 median=6 2014 : mean=6.8 median=7 Age effects ? 46
K-S test : 0.0720 (p= 0.033)
Source: Panel PATER 2007-2014 (N=807)
2007 : mean=6.2 median=6 2014 : mean=7.0 median=7
Risk score : determinants (+ : more risk averse)
all observations
OLS
47
Covariates Wave 2009
Coef. 0.4317
Wave 2011
0.8321
6.43
Wave 2014
0.7951
4.51
Age
0.1334
30.81
Income Q2
1.0122
4.47
Income Q3
1.0270
4.47
Income Q4
1.1237
4.52
Income (missing)
1.7273
6.33
Sex: female
2.5938
18.69
Married
1.6539
10.75
Social origin : farmer
0.7909
3.81
Social origin : self-employ.
-0.5226
-2.55
Social origin : independant professionnal services
-0.1893
-0.49
Education : Bac.
-0.0270
-0.16
Education : >Bac.
-0.1378
-0.84
Children (at home)
-0.0788
-1.24
Children (out of home)
-0.2199
-3.47
Constant
-7.6059
-20.02
N
14895
n
8435
R2
0.195
Robust-t 3.96
Risk score : age & time (wave) effects Balanced sample
Covariates
Score Robust-t Coef.
Coef.
Scale Robust-t
Coef.
Robust-t
Coef.
Robust-t
Lotery Robust-t Coef.
Coef.
Robust-t
Wave 2009
0.1995
1.24
-0.1014
-0.61
-0.4068
-4.23
-0.4525
-4.67
0.1653
4.05
0.1648
3.98
Wave 2011
0.4053
2.35
-0.1311
-0.70
-0.6737
-7.15
-0.7571
-7.84
0.2526
6.17
0.2517
5.99
Wave 2014
0.8081
4.37
-0.1479
-0.68
-0.5591
-5.69
-0.7060
-6.76
0.1460
3.39
0.1445
3.20
0.1178
8.12
0.0184
4.17
0.0002
0.11
-0.1273
-0.15
4.6687
18.49
3.1755
32.28
Age Constant
6.2285
26.76
5.6606
65.72
3.1860
85.87
N
3168
3168
3083
3083
2883
2883
n
792
792
790
790
790
790
0.002
0.063
0.014
0.026
0.010
0.010
R2
48
Risky asset portfolio : stockholding all observations probit (marginal effect) Covariates Score RA Score TP Age Age squared Income Q2 Income Q3 Income Q4 Income (missing) Gross wealth Q2 Gross wealth Q3 Gross wealth Q4 Gross wealth (missing) Married Education (Bac.) Education (>Bac.) Children (at home) Children (out of home) Wave 2009 Wave 2011 Wave 2014 Parents own stocks ? : Yes Parents own stocks ? : Dnk Expected return on stocks Expected variance on future income Constant
49
N n
probit (marginal effect)
Panel (2007+at least one wave) RE probit
Coef.
Robust t
Coef.
Robust t
Coef.
Robust t
-0.0037 -0.0106 0.0056 0.0000 0.0650 0.0929
-5.16 -8.28 3.71 -1.45 3.66 5.31
-0.0033 -0.0095 0.0072 0.0000 0.0798 0.1161
-3.27 -5.37 3.34 -1.33 2.96 4.45
-0.0268 -0.0865 0.0382 -0.0001 0.5498 0.9309
-2.65 -5.14 1.44 -0.47 2.22 3.77
0.1610
8.17
0.1725
6.22
1.3951
5.32
0.1389 0.1025 0.1355 0.3001 0.0322 -0.0142 0.0503 0.1077 -0.0161 -0.0097 -0.0674 -0.1199 -0.1791 0.1487 0.0068
6.04 6.81 9.54 17.25 2.26 -1.42 4.22 9.65 -4.24 -2.35 -9.99 -15.91 -18.13 13.76 0.75
-1.8493
-12.17
0.1463 0.1263 0.1418 0.3121 0.0804 -0.0166 0.0514 0.1142 -0.0189 -0.0103 -0.0839 -0.1307 -0.2098 0.1674 0.0786 0.3856 0.0209 -2.3750
4.58 6.13 7.56 14.34 3.32 -1.20 3.12 7.71 -3.34 -1.77 -8.16 -11.36 -14.72 12.08 0.18 8.34 1.59 -12.85
1.1818 0.6656 0.8414 1.5774 0.3841 -0.0517 0.4740 0.8490 -0.0744 -0.1138 -0.6677 -1.0688 -1.8833 0.7135 1.8434 1.9320 0.0779 -4.8172
3.68 3.68 4.74 7.93 1.64 -0.34 3.00 5.19 -1.23 -1.82 -7.22 -8.27 -8.94 5.66 0.45 4.45 0.66 -6.42
14895 8435
9446 6003
4022 1830
Panel (2007+at least one wave)
Stockholding (in differences)
Covariates
Coef.
Robust t
Coef.
Robust t
Score RA Score TP Expected return on stocks Expected variance on future income Wave 2011 Wave 2014
-0.002 -0.002 0.092 -0.193 -0.021 0.023
-1.12 -0.73 1.67 -0.41 -1.06 0.94
-0.0022 -0.0020 0.0886 -0.2232 -0.0321 0.0152
-1.05 -0.72 1.66 -0.47 -1.44 0.57
Age
-0.001
-2.99
-0.0019
-3.28
0.0780 0.0776 -0.0927 -0.0063 0.0064 0.0255 0.0018 -0.0094 -0.0184 2023 1231
2.39 2.38 -0.69 -0.36 0.4 1.69 0.25 -1.11 -0.40
Affected by the crisis less than average Affected by the crisis as the average Affected by the crisis (DK) Education (Bac.) Education (>Bac.) Married Children (at home) Children (out of home) Constant N n 50
0.037 2023 1231
1.11
Panel (2007+at least one wave)
Risky asset portfolio : Propensity to take risk Ordered probit
Covariates Score RA Score TP Age Age squared Income Q2 Income Q3 Income Q4 Income (missing) Gross wealth Q2 Gross wealth Q3 Gross wealth Q4 Gross wealth (missing) Married Education (Bac.) Education (>Bac.) Children (at home) Children (out of home) Wave 2009 Wave 2011 Wave 2014 Parents own stocks ? : Yes Parents own stocks ? : Dk Expected return on stocks Expected variance on future income Constant
51
N n
Coef.
RE (linear)
RE (balanced)
Coef.
t
Coef.
t
-0.0136 -0.0105 0.0038 -0.0001 -0.0077 0.0224 0.0652 0.0500 0.0696 0.0571 0.1949 0.0975 0.0331 0.0613 0.0924 -0.0138 -0.0177 -0.0819 -0.0943 -0.0755 0.0985 -0.0159 0.5530 1.2699 1.3113
-7.90 -3.69 0.88 -1.69 -0.20 0.59 1.61 1.00 2.35 1.99 5.95 2.55 1.32 2.35 3.50 -1.37 -1.70 -5.02 -4.31 -2.25 4.41 -0.76 7.42 1.84 11.43
-0.0110 -0.0035 0.0040 -0.0001 -0.0129 0.0011 0.0509 0.0400 0.0938 0.0915 0.1795 0.0995 0.0437 0.0440 0.0782 -0.0060 -0.0111 -0.0721 -0.0981 -0.0662 0.0608 -0.0433 0.5430 1.6570 1.2664
-4.58 -0.88 0.54 -0.89 -0.26 0.02 0.93 0.66 2.22 2.18 3.81 1.71 1.21 1.24 2.06 -0.46 -0.80 -2.72 -3.56 -1.72 1.88 -1.62 5.12 1.67 6.17
3795 1744
1936 649
Panel (2007+at least one wave)
Propensity to take risk (in differences) Covariates
Score RA Score TP Expected return on stocks Expected variance on future income Wave 2009 Wave 2011 Wave 2014 Age Affected by the crisis less than average Affected by the crisis as the average Affected by the crisis (DK) Education (Bac.) Education (>Bac.) Married Children (at home) Children (out of home) N n 52
Coef.
t
Coef.
t
-0.006 -0.001 0.293 0.457 -0.080 -0.019 0.028
-1.55 -0.24 2.88 0.48 -4.53 -0.79 1.13
-0.006 0.000 0.295 0.444 -0.107 -0.041 0.001 0.000 0.049 0.049 0.112 -0.012 0.033 -0.007 -0.010 -0.006
-1.53 -0.07 2.92 0.47 -1.41 -0.51 0.01 -0.17 0.76 0.75 1.57 -0.42 1.39 -0.32 -0.95 -0.44
1892
1892
1164
1164
Conclusions (1): Has France become more prudent ? § From mid 2007 to end of 2014, overall: yes, all measures § Panel data, same individuals § Score: only age effects, older people being more prudent § Others: age & period effects: the crisis makes people more prudent
§ Score: France globally more prudent due to composition effects § Population aging: more old people => significant but limited effect § Young newcomers: more risk averse over time: great recession babies? § Preferences are heterogeneous between households & have a significant explanatory power of wealth & portfolio choices § Other measures polluted: may be OK for unobserved individual heterogeneity (better than nothing), but not for understanding changes in financial behaviours
§ Preference formation: early in life (scores) § Environment: “depression babies”. Social origin, education, gender & 53
§ Especially parents’ own preferences (correlation = 0.2)
Conclusions (2): the declining rate of stock ownership § Composition effects for the whole population? § Lower willingness to take risks in portfolio choices § 2 dependent variables: panel data, at least in 2007 § No effect of (variation) of preference scores § Due to higher exposure to (background) income risk (‘hit by the crisis’)
§ And due to more pessimistic asset price expectations § Δ expected return on stocks § Δ ‘lagged’ (previous wave) expected return on stocks Ø Means that stock ownership should not decrease any more after 2014?
§ But stock ownership continues to decrease until march 2016... (puzzle) 54