``Wealth and Stock Market Participation: Estimating the Causal Effect from Swedish Lotteries by Briggs, Cesarini, Lindqvist and Ostling

Transcription

1 ``Wealth and Stock Market Participation: Estimating the Causal Effect from Swedish Lotteries by Briggs, Cesarini, Lindqvist and Ostling Discussant: Annette Vissing-Jorgensen, UC Berkeley Main finding: If you randomly assign (even large) lottery winnings to households, at most percent of stock market non-participants enter the stock market. Implication: Costs (pecuniary or non-pecuniary) of entering the stock market are likely part of the reason for non-participation, but not the main one. Not all of the relation between financial assets and stock market participation in the cross-section is causal.

2 The paper is thus a confirmation of the conclusion of Andersen and Nielsen (RFS, 2) who used wealth variation from unexpected inheritances: They found that only 35 pct of nonparticipants enter for large inheritances.

3 Background: Financial assets and stock market participation, SCF 23 This relation is documented in many papers I added a simple calculation showing that this relation means that most non-participants are not foregoing much by staying out of the stock market Fraction that owns stocks, directly or indirectly Decile of financial assets Label shows average financial assets in thousands (USD)

4 From the distribution of financial assets among non-participants, you can estimate the distribution of foregone expected dollar income from non-participation. 5% of non-participants have financial wealth of $,5 or less. 75% of non-participants have financial wealth of $, or less. Suppose they were to invest half in stocks upon entry. Suppose the equity premium is 5% after-tax. 5% of non-participants forego expected income of $38/year 75% of non-participants forego expected income of $25/year These numbers get smaller if you try to calculate a ``certainty equivalent number.

5 CDF of financial assets, by stock market participation, SCF Financial assets CDF, non-stockholders CDF, stockholders

6 CDF of expected dollar gain from participation, by stock market participation, SCF Expected dollar gain from participation CDF, non-stockholders CDF, stockholders

7 From this I concluded that even small pecuniary or time costs of entry seemed sufficient to explain most of non-participants decisions. It seems pretty rational for people with little financial wealth to pay little attention to their portfolio allocation. But for non-participants with lots of financial wealth entry costs cannot be the explanation for non-participation. The current paper argues that at most % of non-participants stayed out due to some type of entry cost (and participation costs per period are very small which seems plausible): I had suggested that entry costs may account for 5% or perhaps 75% of nonparticipation. I was very encouraged to see that up to % actually entered! There s clearly some truth to the fact that many non-participants don t enter because they don t have sufficient incentive to figure out that this is actually a good idea.

8 My comments on the paper: We need a bit more work to sort out if % is the right number. I think the true number for what fraction of non-participants stayed out due to some type of entry cost is likely higher than %. I really don t think most of the 6% who didn t enter did so because they firmly believe the equity premium is zero/negative as the paper argues: It s much more likely that they simply are unable to form precise views about the stock market return distribution. And there s evidence that if people think stock market return probabilities are ambiguous they don t enter. Implications for the AFA

9 Comment. Of those with large winnings how many didn t save much? On avg. initial non-participants saved less than half of winnings. What does this graph look like for non-participants who won a lot? Suppose /3 didn t save much. Then you d only expect 2/3 to enter. In that case the fraction of non-participants that stayed out due to some type of entry cost is %/.67=6%. So we need to know the distribution of savings rates, for non-part s w/large winnings.

10 Comment 2. Is it really likely that more than half of non-participants firmly think the equity premium is zero or negative? Authors estimate a CDF for the equity premium with 7 grid points between and 8.5%: Prob s picked so a lifecycle portfolio choice model generates regression coefficients (in simulated data) as close as possible to those in the empirical participation regression If over half think the equity premium is negative, that would rationalize why 6% don t enter even with winnings >2M krona.

11 They compare this to a CDF from Hurd, Van Rooij and Winter (2) which found that in fact over half of Dutch survey respondents appear to think the equity premium is negative. (Aside: More detail is needed on exactly which numbers from Hurd et al are graphed.) I don t think this is the right explanation for the 6% who didn t enter. Clearly, 6% didn t enter and thus must have thought it was a bad idea. But why? Did they firmly believe that the equity premium is zero or negative? I think no.

12 In the Hurd et al paper, the 27 respondents were asked to estimate the average historical stock return over the past 2 years. Only 7/27=5% of respondents were able to provide an answer. It would have been interesting to see the replies by participation, but presumably this means that at least half of non-participants have no view about the average return in the past (they are less informed than the participants). Of those who did provide an answer: 23.5% thought the avg. return was less than 6% 52.2% thought the avg. return was between 6% and 2% 2.3% thought the avg. return was above 2% So for those who had an opinion, there s no indication that many thought the average return in the past was negative -- most thought the avg. return in the past was over 6%.

13 So how do we get from this to Hurd et al concluding that over half of respondents think the equity premium is zero/negative? These same people are then asked to provide probabilities that the stock market return over the next year will be >%, >%, >2%, >3%, <%, <%, <2%, <3% 3% don t know About 2% thought that negative and positive returns were equally likely 27.8% (.6%+.2%) assigned more than 5 percent probability to the stock market return over the next year being positive. 6.% (.6%+.8%) assigned more than 5 percent probability to the stock market return over the next year being negative.

14 So many respondents must have reported probabilities that didn t sum to. And many reported probabilities that didn t change monotonically in the return thresholds or that summed to >. Even keeping the respondents whose answers sum to < (by scaling scaled the responses that sum to < to the sum to ), Hurd et al only report results on belief distributions for about % of the sample (their appendix). Focusing on those, even among stockholders in 2 who stayed in the market until 26, the average probability that the stock market return would be >% was only 5.8: Many stockholders must have said <5%, seemingly contradicting their own choice to participate. Suggests that asking for probability answers lead many people to give answers that are not meaningful and even for those with meaningful answers the beliefs look inconsistent with their own estimates of past average historical returns and their own participation decision: The answers look systematically too pessimistic.

15 Related evidence on the problem of asking for probabilities: Dominitz and Manski (27). Asking HRS respondents in 2, they found that: 87% responded to a question about the probability of a positive nominal stock return. The average probability answer is 53.7%, 5% for non-stockholders, 6% for stockholders. But even among stockholders, at least 25% thought the probability of a positive stock return was 5% or less, again contradicting their own choice to participate.

16 In surveys that ask for mean returns, answers seem more meaningful, with the qualifier that none of the existing US data sets that ask for means have survey data on stock market expected returns for non-stockholders with low wealth: Amromin and Sharpe (22) using the Michigan Survey of Consumer Attitudes. 2-25, people with at least $5, in stocks/stock mutual funds. Looking forward, what is the annual rate of return that you would expect a broadly diversified portfolio of U.S. stocks to earn, on average, over the next three years? At most 2% could have E(r). Much lower than Dominitz and Manski s 25%

17 Using UBS/Gallup data from Vissing-Jorgensen (23): Density UBS/Gallup data, Expected Year Stock Market Return Only 8% answer E(r). Around 2% in Around 2% in 22 after the stock market crash.

18 Comment 3. If non-entry is not about firmly held negative beliefs, it s probably more about inability to form firmly held beliefs after winning and unwillingness to invest in the face of high perceived ambiguity about the return distribution Dimmock, Kouwenberg, Mitchell and Peijnenburg (2): Measure ambiguity aversion using survey questions about preferences between drawing a ball from a box with known odds vs. a box with unknown odds. + std. dev. ambiguity aversion 2 pct points decrease in prob. of participating (in a sample with mean participation of 23 pct). This is equivalent to a change in wealth of. standard deviations ($238,). The effect is coming from households with low self-assessed stock market competence, i.e. from those who are ambiguity averse and think the stock market is an ambiguous bet. Some people think the stock market is ambiguous (those with low stock market competence) and if they are ambiguity averse they don t enter.

19

20 Comment. Is it likely that some people are unable to form reasonable beliefs even when they suddenly get a large incentive to do so? Why? Remember: Almost % of the high-wealth individuals in the SCF hold stocks. At most % of lucky Swedish non-participants enter the stock market even with large lottery winnings. So what s the difference between someone who saved up $2K and a nonparticipant who won it in a lottery or unexpectedly inherited it? The lucky non-participant has not been paying attention to portfolio choice. He rationally decided to not have an opinion about the stock market in the past, given low financial wealth. Out guys like him, are able to figure out that they should enter, given sufficient incentive (in terms of large wealth) to put in some effort to sort this out. These tend to be college educated with low initial wealth: They can figure it out, they just never had an incentive to.

21

22 But what about the other 6 lucky non-participants who don t enter? Why can t they figure out that they should enter? Keep in mind that we re now down to understanding 2% of the population: - 67% of the Swedish lottery winners already participated in the stock market before. - Of the other 33%, 3 pct points enter given sufficient wealth (% of 33%). - So, we re down to 2% of the population. They are disproportionately less educated. Why is it so hard to figure out that the equity premium is (95% likely) to be positive? Why couldn t they just download some data and calculate that the equity premium is positive.

23 That s not how people form expectations! People over-weight recent stock market returns People over-weight returns they have lived through People over-weight their own portfolio returns It seems like people really value the information they themselves have ``seen So it seems likely that some of the entry cost is that people start paying attention to how to invest as they accumulate wealth. After paying attention for a while, they are informed enough and comfortable enough to enter. In that sense someone who earned wealth and someone who won it are not the same and you wouldn t expect similar participation rates.

24 (from Vissing-Jorgesen (23))

25 (from Greenwood and Shleifer (23))

26

27 (from current paper)

28 Of course an alternative explanation is that those who earned the wealth are otherwise different in terms of preferences, expected income, or IQ. But the effect of financial wealth on stock market participation is largely unaffected by controls for risk aversion and current income (and current house size as a measure of expected future income). And there s plenty of wealth effect left after controlling for IQ in Grinblatt, Kelohargju and Linnainmaa (2). So it seems too early to say that most of the effect of financial wealth on stock market participation is not causal. My main worry is that we don t sufficiently understand people s choice between financial wealth and non-financial wealth (mainly home equity). - For example, I suspect that a lot of people have low financial wealth because they don t like having mortgage debt. - In that case, low financial wealth still may cause non-participation but was itself caused by aversion to debt.

29 Comment 5. Implications for the American Finance Association Going forward, a central question is why it s so hard for some people to understand stock returns and, in particular, why they cannot just ask an advisor. Because most give bad advice -- people should be skeptical about advice. Example: Mullainathan, Noeth and Schoar (22) o ``The advisers encouraged the client to invest in index funds in only 7.5% of the advice sessions (2 visits). In contrast, in 5% (or 2) of the visits, the adviser suggested investing in actively managed funds. Is it time for the American Finance Association to step up!? o If the American Medical Association can give detailed recommendation about how to treat a host of illnesses, why can the American Finance Association not give basic portfolio advice? o I have asked this before. The educated wealthy males generally laugh and change the topic, but why? o Perhaps time for a field experiment? What s the treatment effect of trustworthy information?

30 (from Christiansen, Joensen and Rangvid (28))