Down or Out: Assessing the Welfare Costs of Household Investment Mistakes

Transcription

1 Down or Out: Assessing te Welfare Costs of Houseold Investment Mistakes Laurent Calvet, Jon Y. Campbell and Paolo Sodini April Madrid

2 Down or Out Welfare cost of underdiversification nonparticipation Down Out New dataset containing detailed information on te finances of all te ouseolds in Sweden Compreensiveness main difference wit earlier work Guiso, Haliassos, Jappelli (2002) Vissing-Jorgensen (2002) Goetzmann and Kumar (2004)

3 Outline 1. Te measurement callenge and our data 2. Diversification Risk exposure Risk premium 3. Wo diversifies and ow? 4. Cost of Nonparticipation 5. Conclusions

4 Te Measurement Callenge

5 Existing Datasets Surveys (SCF) ouseolds do not like to sare financial information cannot be detailed, ig refusal rate, answers often implausible. Account providers (brokerage ouse, 401(k)) detailed but multiple accounts, non-representative sample Registers of ownersip (Sweden, Finland) detailed and representative sample but capture one asset class Tax records (US) estate tax returns

6 Our Data Panel covering four years ( ) and all resident ouseolds (4.8 million). For eac ouseold we ave: Demograpic variables Assets at security level Income flows Age, gender, marital status, education, birtplace, residence Holdings at yearend of stocks and mutual funds, bank accounts Labor income, welfare payments, capital income by assets, private pension savings Main Limitations Value of DC pension assets (but we see contributions to private pension) Asset allocation of capital insurance products (but we see total value)

7 Diversification

8 Participants 62% Swedis ouseolds eld risky assets in 2002 Complete Financial Portfolio $35,500 on average Cas $16,000 on average Risky Portfolio $19,500 on average Bank accounts Money market funds Stocks $9,200 Mutual funds $10,300

9 Two approaces Metodology Compare ouseold portfolios wit diversified indexes Bencmarks (83-04) MSCI All Country World Index edged for currency risk: S B = 45.2% unedged: S B = 34.5% Swedis Stock Index: S B = 27.4% Pooled Index: S B = 35.5% Idiosyncratic risk exposure model free statistical decomposition Risk premium loss CAPM and Fama Frenc factor models

10 Risk exposure Statistical Decomposition (risky portfolio) Excess Return e e r, t = α + βrb, t + ε, t Variance σ = β σ + σ B 2 i, Total risk Systematic risk Idiosyncratic risk σ β σ B Full Diversification σ = β σ B Plot σ on β

11 Volatility and Beta Of Risky Porfolios Standard Deviation (% per year) Beta TR vs. Beta Hedged World Index Unedged World Index Swedis Index in SEK Value Weigted Index

12 Measuring Mean Returns Historical average returns are noisy estimates Diversification value of ouseold portfolios Use asset pricing models to infer mean returns International CAPM Hedged world index is mean-variance efficient Fama-Frenc tree-factor model Market, size, and value factors

13 Scatter Plots of Houseold Portfolios Houseold Stock Portfolios Hedged World Index Unedged World Index Swedis Index in SEK Swedis T-Bill Stock Portfolios Return Mean Return Standard Deviation

14 Complete Portfolios Houseold Complete Portfolios Hedged World Index Unedged World Index Swedis Index in SEK Swedis T-Bill Return Mean Return Standard Deviation

15 Relative Sarpe Ratio Loss Mean Bencmark Houseold Sarpe ratios slopes of capital allocation lines Standard deviation Houseold relative Sarpe ratio loss wrt bencmark S B RSRL S S = 1, B S = μ σ

16 Relative Sarpe Ratio Loss Bencmark Hedged World Index Unedged World Index Swedis Index in SEK Maximal Sarpe Ratio Specification CAPM FF CAPM FF CAPM FF CAPM FF Sarpe Ratio (%) Complete Portfolios Mean Std Dev t Percentile t Percentile t Percentile t Percentile t Percentile t Percentile Stock Portfolios Mean Std Dev t Percentile t Percentile t Percentile t Percentile t Percentile t Percentile

17 Sarpe Ratio Losses Reasonable level of diversification especially wit respect to te unedged world index and Swedis index Most popular mutual funds are well diversified internationally and do not offer currency risk exposure Non trivial fraction of population quite undiversified

18 Return Loss Relative Sarpe ratio loss does not consider te amount of risk a ouseold is taking Average return te investor is giving up by coosing a suboptimal portfolio compared to a bencmark Mean e ( ) E ˆ r e E( r ) Bencmark Return Loss Houseold RL = S σ S σ B e e E( rˆ ) E( ) r σ Standard deviation complete portfolio

19 Return Loss As a Fraction of Financial Wealt Bencmark Hedged World Index Unedged World Index Swedis Index in SEKMaximal Sarpe Ratio Specification CAPM FF CAPM FF CAPM FF CAPM FF Sarpe Ratio (%) Complete Portfolio Return Loss (%) Mean Std Dev t Percentile t Percentile t Percentile t Percentile t Percentile t Percentile Risky Portfolio Return Loss (%) Mean Std Dev t Percentile t Percentile t Percentile t Percentile t Percentile t Percentile

20 Return Loss Rescaled Measures Bencmark Hedged World Index Unedged World Index Swedis Index in SEK Maximal Sarpe Ratio Specification CAPM FF CAPM FF CAPM FF CAPM FF Return Loss in USD Mean Std Dev 2,243 1,954 1,127 1, ,648 1,756 1,978 25t Percentile t Percentile t Percentile t Percentile 1,190 1, ,056 95t Percentile 2,204 1, ,609 1,952 99t Percentile 7,565 6,341 3,244 2,892 1,129 1,521 5,640 6,429 Return Loss as a Fraction of Disposable Income (%) Mean Std Dev t Percentile t Percentile t Percentile t Percentile t Percentile t Percentile

21 Return Losses Modest return losses in complete portfolios especially wrt unedged world index and Swedis index Risky portfolios ave return losses tree times as large as complete portfolios 5% population loosing more tan 5% per year wrt edged world index Results robust to rescaling

22 Wo is undiversified and wy?

23 Return Loss Decomposition Decompose te complete portfolio return loss RL e complete, = E( rm ) w βrisky, RSRL 1 RSRL Risky asset sare (aggressiveness) Diversification loss Portfolio Inefficiency Systematic exposure of risky portfolio

24 Contributors to Complete Return Loss Return Loss ln(rl complete, ) Risky Sare ln(w ) Risky Portfolio Beta ln β Estimate Std Err Cange Estimate Std Err Cange Estimate Std Err Cange Estimate Std Err Cange Intercept Disposable Income % % % % Log of Financial Wealt % % % % Log of Real-Estate Wealt % % % % Age % % % % Hig-Scool Dummy % % % % Post Hig-Scool Dummy % % % % Missing Education % % % % Immigration Dummy % % % % Houseold Size % % % % Retired Dummy % % % % Unemployment Dummy % % % % Entrepreneur Dummy % % % % Student Dummy % % % % Private Pension Premia/Incom % % % % Log of Total Liabilities % % % % Adjusted R Diversification Loss ln RSRL /(1-RSRL )

25 Intuition Large loss Efficient bencmark Mean Small loss Financially sopisticated Financially unsopisticated Standard deviation

26 Wo Incurs Return Losses? Financially sopisticated ouseolds (ric, educated, ) invest efficiently but take more risk Retired and unemployed ouseolds invest inefficiently and take less risk Entrepreneurs and larger ouseolds invest conservatively Overall, financially sopisticated non-entrepreneurial ouseolds ave te greatest return losses Consistent wit te idea tat people know teir limitations

27 Nonparticipation

28 Wo Participates? 62% of Swedis ouseolds participate (in 2002) We look at demograpic determinants of participation Non participants are financially unsopisticated ouseolds (poorer, less educated ouseolds unfamiliar wit financial mrk) Consistent wit participation learning and setup costs Wat are te benefits of participation? Risk premium

29 Down or Out? Welfare cost of nonparticipation S B σ complete, p Mean Excess Return Bencmark Efficient Risk Premium Standard deviation

30 Down or Out? Welfare cost of nonparticipation S σ complete, Mean Excess Return Bencmark Imputed Nonparticipant Risk Premium S Impute and from non-participant caracteristics Standard deviation σ complete, If a ouseold will invest poorly ( down ) it may not be suc a bad mistake to stay out

31 Down or Out? Scenarios Sarpe Ratio: S Portfolio Risk: σ complete, Non Participation Cost Hedged World Index 45.2% Average Participant 9.50% 4.3% Unedged World Index 34.6% Average Participant 9.50% 3.3% Inputed on Average Non- Participant 26.9% Inputed on Average Non- Participant 8.40% 2.3% Inputed on Non-Participant wit dummies = % Inputed on Non-Participant wit dummies = % 2.1% Inputed on Immigrant Non- Participant 25.9% Inputed on Immigrant Non- Participant 8.50% 2.2% Inputed on Unemployed Non-Participant 26.8% Inputed on Unemployed Non-Participant 7.60% 2.0%

32 Conclusions Many Swedis ouseolds are quite well diversified mutual funds play a vital role in diversification A minority of ouseolds are undiversified Financial sopistication improves portfolio efficiency but also increases risk-taking resulting in iger return losses You migt want to be out if you are going to be down

33 Only in Sweden? Are ouseolds smart or is diversification te result of te Swedis specific institutional framework? Figt or Fligt? Portfolio Rebalancing by Individual Investors Active rebalancing wic offsets about ½ of idiosyncratic passive variations in te risky asset sare Sopisticated ouseolds tend to rebalance more aggressively Results seem to old in te US (SCF data wit imputed portfolio risk and expected return)