04 - magi

30-05-2007

8:16

Pagina 93

Giornale degli Economisti e Annali di Economia Volume 66 - N. 1 (Marzo 2007) pp. 93-114

STOCK MARKET PARTICIPATION: NEW EMPIRICAL EVIDENCE FROM ITALIAN HOUSEHOLDS’ BEHAVIOR ATTILIO GARDINI

AND

ALESSANDRO MAGI*

Received: February 2007; accepted May 2007 This paper provides new and updated empirical evidence about the stockholding behavior of Italian households. By exploiting Bank of Italy SHIW data, we find that in the period 2000-2004 stock market participation rates declined markedly, in contrast with what happened in other European countries and in the U.S. In order to understand this fact, we propose some explanations based on the estimation results of crosssectional and panel data probit regressions. We stress the importance of irrationality and myopic behavior and their relationships with the level of investors’ education.

JEL codes: C01, C25, G11. Keywords: stock market participation; probit regression; behavioral finance.

1. INTRODUCTION Stock market non-participation is considered a puzzle in the literature on macroeconomics and finance because it is not easy to explain the reason why many households, in spite of high stock returns, do not own equities and allocate all their financial wealth to safe assets. Participation is quite low even among wealthy households: for example, at the 80th percentile of the wealth distribution, a typical U.S. household has about $200,000 in financial assets, but almost 20% of these households own no equity (Campbell, 2006). These findings have been emphasized in various contributions (Mankiw and Zeldes, 1991; Haliassos and Bertaut, 1995; Heaton and Lucas, 2000b) and are known in the finance literature as Stock Market (non)-Participation Puzzle (SMPP). In the Italian case, stock market participation was very low in the 1980s, but rose steadily during the 1990s. However, the latest micro data available suggest that there has been a marked drop at the beginning of this century. This paper documents this drop and relates it to the key de-

* Department of Statistical Sciences, University of Bologna - Via Belle Arti, 41 - 40126 Bologna. E-mails: [email protected]; [email protected] (corresponding author). We wish to thank Guglielmo Weber, Giuseppe Cavaliere, Luca Fanelli, an anonymous referee and participants to the 2nd ICEEE conference (Rimini, 25-26 January 2007) for helpful comments and suggestions. We are responsible for any remaining errors.

04 - magi

30-05-2007

8:16

Pagina 94

94

ATTILIO GARDINI

- ALESSANDRO MAGI

terminants of stockholding, such as age, wealth, education and other household characteristics. The relevance of the SMPP seems to be decreasing: the general trend of participation rates is increasing over time in U.S. and the rest of Europe (Ameriks and Zeldes, 2004; Gomes and Michaelides, 2005; Calvet, Campbell and Sodini, 2005; De Santis and Gérard, 2006), but in the last years this trend seems to be inverted for the Italian case. In the period 2000-2004 the stockholding behavior of Italian households shows a decline in participation rates, in contrast with what occurs in other European countries and in the U.S. This reversal trend could be the effect of unobserved individual heterogeneity or the result of a substitution between financial and real wealth in the composition of Italian households’ portfolios. Our econometric analysis shows that the impact of education variables on the probability of stockownership is decreasing over time: this could be due to the fact that the proportion of well educated investors who leave the stock market is bigger than the proportion of poor educated investors who leave it. Moreover, we find that concavity in age holds only in some cases: age seems to have no remarkable effect on stockholding, in contrast with previous studies and with “rational” life-cycle asset pricing models.1 As a possible answer, we suggest the existence of a myopic behavior which drives households’ stockholding choices and is consistent with the anomalies found in age coefficient estimates. The paper is organized as follows. In Section 2 we shortly discuss the prevailing literature on stock market participation. In the third section we focus our attention on the literature concerning the Italian case. In section 4 we describe the data used in our empirical analysis and we present the descriptive statistical evidence and the econometric framework. The most important empirical outcomes and implications are presented and discussed in section 5. Section 6 concludes the paper.

2. LITERATURE

REVIEW

Only 15 years ago, household stockownership in Europe was quite different than it is today. Few households invested in equities and most of their financial wealth was held in the form of risk-free low-return assets. Participation in the stock market was limited to a small part of the population, those in the very upper tail of the wealth distribution and relatively well educated. This picture changes markedly by the end of the 1990s: some dif-

1

See in particular Guiso, Haliassos and Jappelli (2002).

04 - magi

30-05-2007

8:16

Pagina 95

STOCK MARKET PARTICIPATION

95

ferences between countries remain, but a much larger proportion of investors begin to hold stocks in their portfolio. About 50 percent of households in the U.S. and Sweden, and over one third in the UK invest directly or indirectly (through mutual funds and other managed investment accounts) in the stock market. In the Netherlands, Italy, France and Germany the proportion is lower, between 15 and 25 percent, but in each of these countries it has increased quite significantly during the 1990s (Guiso, Haliassos and Jappelli, 2003). These changes were encouraged by a variety of institutional and socioeconomic developments. Some of them were transitory, such as the high stock returns in the 1990s, but many are permanent: the privatization of public utilities, the demographic trends and the growth of the mutual funds industry. As emphasized by Guiso, Haliassos and Jappelli (2003), “... such changes have brought into the stockholder pool many households with less financial sophistication [...]”. In fact, new entrants with the mentioned characteristics can induce greater volatility in stock markets by reacting excessively to market signals. Hence, as suggested by Guiso, Haliassos and Jappelli (2003), “if economic policy is to help avert stock market downturns and volatility in the new decade, it must address the needs of these newcomers and supervise the practices of mutual funds handling their accounts”. Haliassos (2002) remarks that, “resolving the stock market participation puzzle [...] can suggest important profit opportunities for financial institutions, practitioners, and even governments. If we understand what keeps people out of the stock market, we can expand the customer base by designing financial products that appeal to the average household”. Limited financial market participation has also important implications for individual welfare: Cocco, Gomes and Maenhout (2005) show that the welfare loss from nonparticipation in stock markets is between 1.5 and 2 percent of consumption. Trying to explain the stock market participation puzzle, some authors consider participation costs of investing in the stock market; other authors examine whether non-stockholders have background risk which is correlated with the equity market (Heaton and Lucas, 2000a, 2000b; Vissing-Jorgensen, 2002). While these approaches almost certainly explain part of the observed non-participation, they are not able to account for all of it. Polkovnichenko (2004) finds that even among wealthy households, for whom participation costs are relatively low, there is still substantial non-participation; and Curcuru, Heaton, Lucas and Moore (2004) argue that, while the presence of correlated background risk can lower the recommended allocation to equity, it cannot easily generate an allocation as low as zero. The emphasis on transaction and information costs as explanation of the SMPP is widespread in the literature (Vissing-Jorgensen, 2002, 2003), but the exact

04 - magi

30-05-2007

96

8:16

Pagina 96

ATTILIO GARDINI

- ALESSANDRO MAGI

nature of these costs is not well understood and their amount seems not high enough to prevent the majority of households from holding stocks. In order to provide new interpretations of stock market participation costs, some authors have started to think of these costs as an economist’s description of psychological factors that make equity ownership uncomfortable for some households. Hong, Kubik and Stein (2004), for example, find that households that interact more with other households in their community are more likely to own stocks, particularly if the participation rate is high in the community, suggesting that households are more comfortable following financial practices shared with others (Campbell, 2006). Guiso, Sapienza and Zingales (2005) find that households that express willingness to trust others are more likely to own stocks. In a recent contribution, Guiso and Jappelli (2006) find that the portfolio Sharpe ratio is negatively associated with investment in information: portfolio performance decreases as investment in financial information increases (in terms of time and money). This is in line with a recent developing strand of the behavioral finance literature (“overconfident” investor models) but it is at odds with standard “rational” models and with intuition. In fact, it is largely accepted that those who invest much time and money in acquiring information should get high portfolio performances (Verrecchia, 1982; Barlevy and Veronesi, 1999; Peress, 2004) and there is consensus about the fact that well educated individuals are often those who invest more in information. It is plausible to assume that more educated investors process information better than those less educated and hence the former could be more rational in making their financial choices.2 Such behavioral finance explanations (see Barberis, Huang and Thaler, 2006; Christelis, Jappelli and Padula, 2006; Guiso and Jappelli, 2006; Magi, 2005) assume that cognitive capabilites of individuals are bounded: this fact affects their ability to process information correctly, therefore they make sub-optimal financial decisions/portfolio choices.3

3. ITALIAN

HOUSEHOLDS’ STOCK MARKET PARTICIPATION

In Italy, in the 1980s direct stockholding accounted for about only 15% of households’ financial assets and indirect stockholding through mutual 2

About this line of research see Magi (2005) and, more recently, Alessie, Lusardi and van Rooij (2007), where the relationship between stock market participation and financial literacy is emphasized. 3 This approach also has some links with the recent literature on rational inattention (Sims, 2003; Reis, 2005).

04 - magi

30-05-2007

8:16

Pagina 97

STOCK MARKET PARTICIPATION

97

funds was almost absent. The high level of government debt and the high interest rates necessary to finance it made stocks relatively unattractive (Guiso and Jappelli, 2002). The thinness of the Italian stock market and its volatility discouraged stockholding, even after the introduction of mutual funds in 1984. Then, in the 1990s different factors contributed to increase stockholdings. We refer to the drop in Italian Treasury bill returns, the changes in the social security system, the privatization process of public companies and the Italian stock market growth. Moreover, the rise of competition among the financial firms offering investment services (reduction in entry costs and financial information costs) and the availability of new financial products increased significantly stock market participation. In the Italian case the prevailing literature on the determinants of stock market participation has reached some important empirical results:4 a) stock market participation rates are increasing over time; b) participation in risky assets is strongly increasing in wealth, exhibits a marked concave age profile and rises with education; c) participation depends on background risk (labour income uncertainty, entrepreneurial risk, etc.);5 d) participation costs6 act as a barrier to entry at low wealth and imply that participation increases with wealth. In this paper we update the econometric analysis and find some new results. The decrease in stock market participation between 2000 and 2004 may have been induced by the poor performance of the Italian Stock Exchange over the period. In particular, realised returns were strongly negative between 2000 and 2002 (-42.3% over the period) – only partly offset by the positive return realised between 2002 and 2004 (+34.6%).7 It is possible that many households revised their subjective expectations downwards as a result of the current low returns and decided to leave the stock market.

4 See in particular Guiso, Haliassos and Jappelli (2002); see also Guiso and Jappelli (2002) and Guiso, Haliassos and Jappelli (2003). See the same papers for a detailed descriptive statistical evidence with respect to wealth, education, age and other household characteristics. 5 Several important studies have found associations between stockholding and background income risk (Guiso, Jappelli and Terlizzese, 1996; Heaton and Lucas, 2000b). 6 These costs take several forms, from minimum investment requirements, to transaction costs in purchasing stocks and mutual funds, to information costs. 7 We have calculated equity returns by using the values of the MIBTEL index at the end of each year. Data drawn from www.borsaitaliana.it.

04 - magi

30-05-2007

98

8:16

Pagina 98

ATTILIO GARDINI

- ALESSANDRO MAGI

The drop in financial market participation could also be related to some macro events (“September the 11th” at the international level and the “Parmalat crash” in the domestic context) which weakened the trust of investors in equity markets and encouraged irrational choices (in a context of irrational preferences the investors’ trust is of crucial importance in determining stockholding behavior).

4. DATA

DESCRIPTION AND ECONOMETRIC MODELS

In order to study recent trends in stock market participation and to analyse the process dynamics, we use data from four waves of the Bank of Italy Survey on Household Income and Wealth (SHIW).8 This is a biannual survey run by the Bank of Italy with the specific aim of providing information on household saving, income and wealth. The survey collects detailed information on the composition of household wealth and on demographic variables. The fact that the survey is repeated over time is very useful, because it allows the analysis of stockholding behavior of Italian households over time. In Tables 1-4 we report some data concerning Italian households’ participation rates: there is an evident and large decline in stock market participation (and in financial market participation as well), in particular since 2000. In 1998 the participation rate for direct stockholding was 8.4% (see Table 1); the total participation rate (direct and indirect stockholding) was 15.5% (Table 3) and the participation rate for all risky financial assets was 18.5% (Table 4).9 In 2000 there was a general increase in financial participation but in the next years participation rates declined dramatically and the reduction holds for direct stockholding, indirect stockholding, total stockholding and also for the broad category “risky financial assets”.10 This behavior could be transitory but it is nevertheless interesting because it is totally at odds with the recent prevailing trends in Europe and U.S. (Ameriks and Zeldes, 2004; Gomes and Michaelides, 2005). We have also analysed the 1998-2004 SHIW panel section (1,855 households and 7,420 observations), in order to find confirmation of the above de-

8

Bank of Italy (1998, 2000, 2002 and 2004). Risky financial assets include stocks, corporate bonds, managed investment accounts and mutual funds. 10 We point out that in 1998 and 2000 indirect stockholdings are partially over-estimated because they include not only equity mutual funds but also other funds. This problem is due to the nature of SHIW data. Instead, since 2002 SHIW wave, we have exact data on equity mutual funds. 9

04 - magi

30-05-2007

8:16

Pagina 99

STOCK MARKET PARTICIPATION

99

scribed tendency. We find that also in the balanced panel section there is an evident decline in stock market participation (see Tables 5-6-7): this holds for both direct, indirect and total participation. Hence, the peculiar behavior of the last years finds a further confirmation in the panel section analysis, showing that the reduction of stockholding is a clear feature of Italian investors’ behavior in the last period. 4.1 Cross-sectional analysis First, we estimate direct and indirect stock market participation in four different cross-sections, for a total of 31,171 observations, describing Italian households’ financial choices in 1998, 2000, 2002 and 2004. We estimate and test the following probit regression: y*i = xiβ + εi

i = 1, ...N

with yi = 1 if y*i > 0 (stockholding) and 0 otherwise. As regressors we use dummy variables (xi), including age, education, other household characteristics (married, male), real and financial wealth quartiles.11 Our aim is to analyse the main household characteristics which affect stock market participation and the new recent changes in Italian households’ portfolio choices. 4.2 Panel analysis Then, we run a balanced panel data probit regression with random-effects, in order to control for unobserved heterogeneity in the 1998-2004 SHIW panel section. For this purpose we estimate the following model: yit* = xit β + ε it i = 1, ...N ; t = 1, ...T ε it = α i + uit , uit ~ N(0,1), α i ~ N(0,σ α2 )

with yit = 1 if y*it > 0 (stockholding) and 0 otherwise, and where αi is the individual effect related to any unobservable individual characteristic assumed to be time-invariant, such as risk aversion, and uit is a time-varying effect (expectation errors, etc.). Both αi and uit are independent of observable characteristics xi, and uit are serially uncorrelated. In this framework the indi-

11

The econometric specification is similar to Guiso, Haliassos and Jappelli (2003), but we use real wealth quartiles and not income quartiles. If we add income quartiles as regressors, we find that their effect is negative and not significant. But without financial wealth quartiles as regressors, the effect of income quartiles becomes positive and significant. Probably, this is due to the high correlation between income and financial wealth.

04 - magi

30-05-2007

8:16

100

Pagina 100

ATTILIO GARDINI

- ALESSANDRO MAGI

vidual effects αi are assumed to be part of the stochastic component εit (see Miniaci and Weber, 2002).12 We control for unobserved heterogeneity by looking at the standard likelihood ratio-test used in panel data probit analyses. This test formally compares the pooled estimator with the panel estimator: the null hypothesis is that the two estimators are the same, i.e. that the panel-level variance comσ2 ponent σ α2 is unimportant. Formally we have H0 : ρ = 2 α 2 = 0 .

( )

σ α +σ u

4.3 Econometric estimation issues The econometric estimation procedure must face a basic identification problem concerning age and time effects (Heckman and Robb, 1985; Ameriks and Zeldes, 2004). At any time t a person born in year b is at years old, where at = t – b. Hence, it is impossible to separately identify age effects, time effects and cohort (birth-year) effects on participation and portfolio choices. Even if we have complete panel data on portfolios of households over time, any pattern in the data can be fit equally well by age and time effects, age and cohort effects, or time and cohort effects (Campbell, 2006). Theory suggests that there should be time effects on portfolio choice if households perceive changes over time in the risks or expected returns of risky assets. Theory also suggests that there should be age effects on portfolio choice if older investors have shorter horizons than younger investors and investment opportunities are time-varying, or if older investors have less human wealth relative to financial wealth than younger investors (Bodie, Merton and Samuelson, 1992; Campbell and Viceira, 2002). Cohort effects are more problematic. As suggested by Campbell (2006), such effects could be caused by different labor market experiences which affect the ratio of human to financial wealth held by a cohort at each age, but this effect is unlikely to be strong in modern economies. Cohort effects could also arise from differences in preferences, for example driven by different asset market experiences. Such effects cannot be identified without modelling them (or age or time effects) in some way. We will follow Heaton and Lucas (2000a) and most other studies by setting cohort effects to zero: under this assumption age effects can be estimated in cross-sectional data.13

12 We do not use the conditional fixed-effects logit estimator because it does not allow to estimate the effect of time-invariant characteristics and because the sample size drops dramatically (for example, in the direct stockholding case, we have 353 households and 1,412 observations). 13 For identification problems concerning the age effect, especially in short panel data, see also Guiso and Jappelli (2000) and Miniaci and Weber (2002).

04 - magi

30-05-2007

8:16

Pagina 101

STOCK MARKET PARTICIPATION

101

With regard to the coefficient estimates we recall that probit model coefficients cannot be directly interpreted as marginal effects. The following relation holds: ∂F( xi′β ) = β kφ ( xi′β ), ∂xik

where F (·) is the cumulative distribution function of the standard normal distribution, φ(·) is the density function of the same distribution and βk is the coefficient estimate. We report marginal effects in Tables 8-9, coefficient estimates in Tables 12-13.

5. ESTIMATION

RESULTS

5.1 Cross-sectional estimates We analyse and discuss the estimation results of the four cross-sectional probit regressions, by comparing the estimated coefficients in different years. From Tables 8 and 9 we can see the impact of the different variables on direct (Table 8) and indirect (Table 9) stockholding: each of the coefficients indicates the marginal effect of the dummy on the probability to invest in stocks and in equity mutual funds. In the direct stockholding case, although some of age dummies are not statistical different from zero, the probit regressions indicate that participation is concave in age only in some cases (in 2000 and 2002), in contrast with previous studies where concavity always holds. This problem is particularly evident in 2004 but appears also in the previous years. Age seems to be a variable that does not affect stockholdings in Italy (there are only small differences between different age classes). This result is not consistent with “rational” life-cycle asset pricing models, but it could be explained assuming the presence of myopic investors, who behave in line with recent behavioral finance theories. The dummy variables high school and college have a positive but decreasing impact on the probability of investing in stocks and mutual funds.14 More educated households are not only more likely to hold stocks, but also to learn easily how to invest in stocks and to estimate more precisely the

14 In the context of behavioral finance models the level of education plays a crucial role in explaining investor heterogeneity and therefore stock market participation (see Magi, 2005).

04 - magi

30-05-2007

102

8:16

Pagina 102

ATTILIO GARDINI

- ALESSANDRO MAGI

costs and benefits of them. The decreasing participation of the last years could be due to a spin process which has reduced the average education level of Italian investors; this fact could also explain the decline in participation rates. This process could be related to institutional factors such as the lack of pension funds and the peculiar influence of Italian banks on household portfolio choices. Another possibility is that the “newcomers” do not trust the stock market and stay in it only a few time, going out very quickly. This lack of trust could depend on individuals’ irrational preferences and on their perception of participation costs. Guiso, Haliassos and Jappelli (2003) argue that the role of “new investors” (new entrants in the stock markets) may be important for understanding the volatility of equity markets, because of their low “financial sophistication” and hence their possible underreactions or overreactions to different stock markets signals (with relevant consequences for economic policy). Behind this consideration there is the idea that newcomers are less educated than incumbents. Our estimation results show that there is a decreasing effect over time of education variables on stock market participation: this fact may be interpreted as evidence that in the period 1998-2004 more educated investors left the stock market while those less educated entered it. Moreover, the descriptive statistical evidence reported in Tables 1011 shows that, among households who leave the stock market in the period 2000-2004, the major part is represented by the well educated ones (high school and college). In particular, by comparing the results of the two tables, we can see that the proportion of well educated investors who leave the stock market is larger than the proportion of poor educated investors who pull out from the stock market. In the period 2000-2004 the stock market loses (in percentage terms) 8.5 (12.6 minus 4.1) well educated investors and 3.6 (5.2 minus 1.6) poor educated investors in the direct participation case, and 12.8 (19.3 minus 6.5) well educated investors and 8 (11 minus 3) poor educated investors in the total participation case.15 The role of education recalls the relevance of information costs for participation decisions. Such costs, broadly interpreted, help explain the correlation between participation and education: poorly speaking, more educated households face lower information costs than others. But at the same time, information costs are not necessarily correlated with wealth and might therefore explain why so many wealthy households (which should face low information costs) do not invest in risky financial assets. However, in gen-

15 We have also calculated that the number of schooling years of the investors who leave the stock market in the period 2000-2004 is roughly more than twice of the number of schooling years of those who enter the stock market.

04 - magi

30-05-2007

8:16

Pagina 103

STOCK MARKET PARTICIPATION

103

eral, education strongly affects equity ownership, even controlling for age and wealth (see Tables 8 and 9).16 Therefore, education appears to improve the efficiency with which households make their financial decisions. This implies that education increases the welfare benefits of participation and may also reduce the psychological discomfort associated with activities for which households feel poorly prepared (Campbell, 2006).17 Demographic variables such as married and male have a limited (but significant) impact on direct stockholding; things are slightly different for indirect stockholding. In particular, the variable “male” has an important and significant impact (on stocks) in the last two years of our analysis (2002 and 2004), while in the previous years the impact is negligible. Financial wealth quartiles have a large and positive impact on the probability of direct and indirect stockholding; such impact is increasing in quartiles18 but decreasing in the last four years. The effect of financial wealth on direct and indirect stockholding is increasing from 1998 to 2000, but from 2000 to 2004 it is decreasing. This dynamic behavior is consistent with the drop in participation rates occurred in the period 2000-2004: investors reduced their participation to the stock market and this fact is also evident by observing the effect of financial wealth, which is an important determinant of the probability to invest in stocks.19 Real wealth quartiles have a decreasing impact on the probability to invest in stocks and the same thing holds for indirect stockholding. Moreover, the extent of the marginal effect is larger for stocks than for mutual funds. The trade-off between real and financial assets could be the main explanation of the reduction of stock market participation in recent years. Myopic 16 Campbell (2006) finds the same result in a slightly different setting. Moreover, the “trust effect” estimated by Guiso, Sapienza and Zingales (2005) is weaker for educated households. 17 Christelis, Jappelli and Padula (2006) study the importance of cognitive abilities for the decision to invest in stocks using data drawn by the recent Survey of Health, Ageing and Retirement in Europe (SHARE). The survey has detailed data on wealth and portfolio composition of individuals aged 50+ in 11 European countries and three indicators of cognitive abilities: mathematical, verbal fluency and recall skills. They find that the propensity to invest in stocks is strongly associated with cognitive abilities, for both direct stock market participation and indirect participation through mutual funds and investment accounts. 18 The relation between wealth and participation is standard in asset pricing models: individuals who participate to stock markets are in particular those in the upper tail of the wealth distribution. This fact probably depends on transaction and management costs related to stockholding (see Costa, Gardini and Iezzi, 2007). 19 Nevertheless, as a caution to the previous discussion, we observe that recent works on Italian households’ financial wealth (see D’Aurizio, Faiella, Iezzi and Neri, 2006) stress the fact that risky financial wealth is probably strongly underestimated. For example, in aggregate data, the stockownership seems to be underestimated by 70%.

04 - magi

30-05-2007

8:16

Pagina 104

104

ATTILIO GARDINI

- ALESSANDRO MAGI

investors (maybe those less educated) have probably chosen to move their savings from financial to real assets, following the real estate bubble which affected the economies in recent years. 5.2 Panel estimates In Table 12 and 13 we have reported the maximum likelihood pooled and random-effects estimates for direct, indirect and total stockholding. The pooled estimation corresponds to a specification in which individual heterogeneity is ignored and observations are treated as independent. This implies that αi = 0 and that the explanatory variables are independent of the time varying component uit. Otherwise, in the random-effects probit estimation case, we consider heterogeneity, i.e. the unobserved individual effects. We highlight that our panel analysis is the first one updated with 20022004 SHIW waves. Following Miniaci and Weber (2002) we compare the two type of estimates in order to control for unobserved heterogeneity. The panel estimates always have the same sign of the pooled ones, with the exception of age classes. But cross-sectional analysis shows that age is not relevant for participation decisions in the Italian case. Otherwise, education variables and financial wealth have a positive and significant effect on participation, in line with previous studies. We have some important differences in the size of estimated coefficients: panel estimates are larger than pooled ones. Moreover, the likelihood-ratio test of ρ = 0 rejects the null hypothesis, showing that the panel-level variance

( )

component σ α2 is important and the panel (random-effects) estimator is different from the pooled estimator. This implies that the two type of estimates are not the same: the panel data estimation confirms the presence of unobserved individual heterogeneity (which could be related to risk aversion, information asymmetries, etc.). Moreover, year dummies have a relevant and significant effect on direct, indirect and total participation (see Table 13). This means that macroeconomic effects play an important role in explaining changes in stock market participation.20 In 2000 the impact of the year dummy is positive and relevant (with a lower but not significant effect for the indirect participation case); in 2002 there is an expected negative effect on participation (with the exception of stocks), while in 2004 the expected negative effect on participation is very high, in particular for indirect and total stockholding. We observe that, overall, the effect of time dummies is in accordance with the dynamics of participation rates. 20 The estimation has been carried out by dropping the dummy relative to 1998 because the period which is relevant to investigate decreasing participation is 2000-2004.

04 - magi

30-05-2007

8:16

Pagina 105

STOCK MARKET PARTICIPATION

6. CONCLUDING

105

REMARKS

This paper provides new and updated empirical evidence about the stockholding behavior of Italian households. By exploiting Bank of Italy SHIW data, we find results that confirm some basic well-established findings and we find some new evidence concerning the decline in stockownership. In fact, in Italy, totally at odds with the recent prevailing trends in Europe and U.S. and despite the growth in equity returns occurred since 20022003, participation rates declined strongly in the period 2000-2004. The role of financial and real wealth seems to be relevant in explaining the apparent instability of Italian households’ stock market participation. The impact of real wealth quartiles decreases over time and this can be a sign of a shift from risky financial assets to housing. A possible explanation is that myopic investors (probably those less educated) chose to move their savings from financial market to real estate one. Our estimation results show that concavity in age is valid only in some cases: age has no effect on stockholding, in contrast with previous studies. As a possible answer, we suggest the existence of a myopic behavior which drives households’ stockholding choices and is consistent with the anomalies found in age coefficient estimates. The results suggest the possible existence of a “turnover” process within the Italian stock market, which partially affects the decline in participation rates. Different reasons could explain such a behavior: a) the turnover process could be related to institutional factors such as the lack of pension funds and the peculiar structure of Italian financial sector (widespread influence of banks on household portfolio choices); b) another possibility is that the so-called “newcomers” do not trust the stock market and stay in it only a few time, going out very quickly. This lack of trust could depend on individuals’ irrational preferences, which heavily affect their risky choices and their perception of participation costs. REFERENCES ALESSIE, R. - LUSARDI, A. AND VAN ROOIJ, M. (2007), “Financial literacy and stock market participation”, working paper, Dartmouth College (IL), U.S. AMERIKS, J. AND ZELDES, S. (2004), “How do household portfolio shares vary with age?”, working paper, Columbia University. BANK OF ITALY (1998), Survey on Household Income and Wealth, Rome. BANK OF ITALY (2000), Survey on Household Income and Wealth, Rome. BANK OF ITALY (2002), Survey on Household Income and Wealth, Rome. BANK OF ITALY (2004), Survey on Household Income and Wealth, Rome. BARBERIS, N. - HUANG, M. AND THALER, R. (2006), “Individual preferences,

04 - magi

30-05-2007

106

8:16

Pagina 106

ATTILIO GARDINI

- ALESSANDRO MAGI

monetary gambles and stock market participation”, American Economic Review, forthcoming. BARLEVY, G. AND VERONESI, P. (1999), “Information acquisition in financial markets”, Review of Economic Studies, 67, 79-90. BODIE, Z. - MERTON, R. AND SAMUELSON, W. (1992), “Labor supply flexibility and portfolio choice in a life cycle model”, Journal of Economic Dynamics and Control, 16, 427-449. CALVET, L. - CAMPBELL, J. AND SODINI, P. (2005), “Down or out: Assessing the welfare costs of household investment mistakes”, unpublished paper, HEC Paris, Harvard University, and Stockholm School of Economics. CAMPBELL, J. (2006), “Household Finance”, Journal of Finance, 61, 1553-1604. CAMPBELL, J. AND VICEIRA, L. (2002), Strategic Asset Allocation: Portfolio Choice for Long-Term Investors, Oxford University Press. CHRISTELIS, D. - JAPPELLI, T. AND PADULA, M. (2006), “Cognitive abilities and portfolio choice”, CSEF working paper n. 157. COCCO, J. - GOMES, F. AND MAENHOUT, (2005), “Consumption and Portfolio Choice over the Life Cycle”, Review of Financial Studies, 18, 491-533. COSTA, M. - GARDINI, A. AND IEZZI, S. (2007), “Latent class models in financial data analysis: an application to the measurement of the equity home bias”, Statistica, forthcoming. CURCURU, S. - HEATON, J. - LUCAS, D. AND MOORE, D. (2005), “Heterogeneity and Portfolio Choice: Theory and Evidence”, in Handbook of Financial Econometrics, Ait-Sahalia, Y. and Hansen, L.P. (eds), Elsevier Science, Amsterdam. D’AURIZIO, L. - FAIELLA, I. - IEZZI, S. AND NERI, A. (2006), “L’under-reporting della ricchezza finanziaria nell’indagine sui bilanci delle famiglie (in Italian)”, Temi di Discussione n. 610, Bank of Italy, Rome. DE SANTIS, R. AND GÉRARD, B. (2006), “Financial Integration, International Portfolio Choice and the European Monetary Union”, ECB working paper series n. 626, Frankfurt. GOMES, F. AND MICHAELIDES, A. (2005), “Optimal Life-Cycle Asset Allocation: Understanding the Empirical Evidence”, Journal of Finance, 60, 869-904. GUISO, L. - HALIASSOS, M. AND JAPPELLI, T. (2002) (eds), Household Portfolios, Mit Press. GUISO, L. - HALIASSOS, M. AND JAPPELLI, T. (2003), “Household Stockholding in Europe: Where Do We Stand and Where Do We Go?”, Economic Policy, 36, 123-170. GUISO, L. AND JAPPELLI, T. (2000), “Household Portfolios in Italy”, CSEF working paper n. 43, University of Salerno. GUISO, L. AND JAPPELLI, T. (2002), “Stockholding in Italy”, CSEF working paper n. 82, University of Salerno. GUISO, L. AND JAPPELLI, T. (2006), “Information Acquisition and Portfolio Per-

04 - magi

30-05-2007

8:16

Pagina 107

STOCK MARKET PARTICIPATION

107

formance”, CSEF working paper n. 167, University of Salerno. GUISO, L. - JAPPELLI, T. AND TERLIZZESE, D. (1996), “Income risk, borrowing constraints and portfolio choice”, American Economic Review, 86, 158-172. GUISO, L. - SAPIENZA, P. AND ZINGALES, L. (2005), “Trusting the stock market”, working paper, University of Chicago and Northwestern University. HALIASSOS, M. (2002), “Stockholding: Recent Lessons from Theory and Computations”, working paper, University of Cyprus. HALIASSOS AND BERTAUT (1995), “Why do so few hold stocks?”, Economic Journal, 105, 1110-1129. HEATON, J. AND LUCAS, D. (2000a), “Portfolio choice and asset prices: the importance of entrepreneurial risk”, Journal of Finance 55, 1163-1198. HEATON, J. AND LUCAS, D. (2000b), “Portfolio Choice in the Presence of Background Risk”, Economic Journal, 110, 1-26. HECKMAN, J. AND ROBB, R. (1985), “Using longitudinal data to estimate age, period, and cohort effects in earnings equations”, in Mason, W. and Fienberg, S. (eds), Cohort Analysis in Social Research: Beyond the Identification Problem, 137-150, Springer-Verlag, New York. HONG, H. - KUBIK, J. AND STEIN, J. (2004), “Social Interaction and Stock Market Participation”, Journal of Finance, 59, 137-163. MAGI, A. (2005), “Financial Decision-Making and Portfolio Choice under Behavioral Preferences: Implications for the Equity Home Bias Puzzle”, Note e Ricerche, 3/2005, University of Bologna. MANKIW, G. AND ZELDES, S. (1991), “The consumption of stockholders and non-stockholders”, Journal of Financial Economics, 29, 97-112. MINIACI, R. AND WEBER, G. (2002), “Econometric issues in the estimation of household portfolio models”, in Household Portfolios, Guiso, L. - Jappelli, T. and Haliassos, M. (eds), Mit Press. PERESS, J. (2004), “Wealth, information acquisition and portfolio choice”, Review of Financial Studies, 17, 879-914. REIS, R. (2005), “Monetary policy for inattentive economies”, Journal of Monetary Economics, 52, 703-725. SIMS, C. (2003), “Implications of rational inattention”, Journal of Monetary Economics, 50, 665-690. VERRECCHIA, R. (1982), “Information acquisition in a noisy rational expectations economy”, Econometrica, 50, 1415-1430. VISSING-JORGENSEN, A. (2002), “Towards an explanation of household portfolio choice heterogeneity: nonfinancial income and participation cost structures”, NBER working paper n.8884. VISSING-JORGENSEN, A. (2003), “Perspectives on behavioral finance: does “irrationality” disappear with wealth? Evidence from expectations and actions”, in M. Gertler and K. Rogoff (eds), NBER Macroeconomics Annual 2003, MIT Press.

04 - magi

30-05-2007

8:16

108

Pagina 108

ATTILIO GARDINI

- ALESSANDRO MAGI

TABLES

TABLE 1 – Participation rates for stocks YEARS

HOUSEHOLDS

PARTICIPANTS

PART

1998

7.147

599

8,4

2000

8.001

820

10,2

2002

8.011

672

8,4

2004

8.012

433

5,4

RATE

(%)

The table reports the number of participants and participation rates for stocks. Our elaboration based on data drawn from the 1998, 2000, 2002 and 2004 SHIW.

TABLE 2 – Participation rates for equity mutual funds YEARS

HOUSEHOLDS

PARTICIPANTS

PART

1998

7.147

776

10,8

2000

8.001

970

12,1

2002

8.011

588

7,3

2004

8.012

384

4,8

RATE

(%)

The table reports the number of participants and participation rates for equity mutual funds. Our elaboration based on data drawn from the 1998, 2000, 2002 and 2004 SHIW.

TABLE 3 – Participation rates for total stockholding YEARS

HOUSEHOLDS

PARTICIPANTS

PART

1998

7.147

1.110

15.5

2000

8.001

1.451

18.1

2002

8.011

1.069

13.3

2004

8.012

711

8.9

RATE

(%)

The table reports the number of participants and participation rates for total stockholding (stocks plus equity mutual funds). Our elaboration based on data drawn from the 1998, 2000, 2002 and 2004 SHIW.

04 - magi

30-05-2007

8:16

Pagina 109

109

STOCK MARKET PARTICIPATION

TABLE 4 – Participation rates for risky financial assets YEARS

HOUSEHOLDS

PARTICIPANTS

PART

1998

7.147

1.323

18,5

2000

8.001

1.717

21,5

2002

8.011

1.500

18,7

2004

8.012

1.062

13,3

RATE

(%)

The table reports the number of participants and participation rates for the category “risky financial assets”. Risky financial assets include stocks, corporate bonds, managed investment accounts and mutual funds. Our elaboration based on data drawn from the 1998, 2000, 2002 and 2004 SHIW.

TABLE 5 – Participation rates for stocks (panel section) YEARS

HOUSEHOLDS

PARTICIPANTS

PART

1998

1.855

180

9.7

2000

1.855

252

13.6

2002

1.855

212

11.4

2004

1.855

135

7.3

RATE

(%)

The table reports the number of participants and participation rates for stocks. Our elaboration based on data drawn from the 1998-00-02-04 SHIW panel section.

TABLE 6 – Participation rates for equity mutual funds (panel section) YEARS

HOUSEHOLDS

PARTICIPANTS

PART

1998

1.855

251

13.5

2000

1.855

285

15.4

2002

1.855

190

10.2

2004

1.855

128

6.9

RATE

(%)

The table reports the number of participants and participation rates for equity mutual funds. Our elaboration based on data drawn from the 1998-00-02-04 SHIW panel section.

TABLE 7 – Participation rates for total stockholding (panel section) YEARS

HOUSEHOLDS

PARTICIPANTS

PART

1998

1.855

353

19.0

2000

1.855

438

23.6

2002

1.855

339

18.3

2004

1.855

231

12.5

RATE

(%)

The table reports the number of participants and participation rates for total stockholding. Our elaboration based on data drawn from the 1998-00-02-04 SHIW panel section.

04 - magi

30-05-2007

8:16

110

Pagina 110

ATTILIO GARDINI

- ALESSANDRO MAGI

TABLE 8 – Cross-sectional probit regressions for direct stockholding STOCKS

VARIABLE 1998

2000

2002

2004

Age 30-39

0.0004 (0.06)

0.0073 (0.84)

0.016 (1.77)

0.033 (2.34)

Age 40-49

-0.0043 (-0.62)

0.019 (2.02)

0.021 (2.29)

0.032 (2.43)

Age 50-59

0.0037 (0.49)

0.023 (2.41)

0.028 (2.84)

0.033 (2.55)

Age 60-69

-0.0004 (-0.06)

0.0084 (0.93)

0.024 (2.51)

0.031 (2.35)

Age 70+

-0.007 (-0.97)

-0.012 (-1.59)

0.0085 (1.01)

0.022 (1.91)

High School

0.041 (9.33)

0.036 (8.86)

0.036 (9.94)

0.023 (7.77)

College

0.085 (10.56)

0.058 (8.60)

0.051 (8.42)

0.042 (8.00)

Married

0.010 (2.87)

0.008 (2.32)

0.010 (3.23)

0.003 (1.26)

Male

0.0073 (1.91)

0.006 (1.58)

0.012 (3.86)

0.011 (5.04)

II fin wealth quartile

0.097 (4.05)

0.16 (4.08)

0.12 (3.79)

0.079 (3.38)

III fin wealth quartile

0.21 (6.76)

0.36 (6.69)

0.24 (5.67)

0.16 (5.06)

IV fin wealth quartile

0.38 (9.43)

0.51 (8.25)

0.39 (7.43)

0.25 (6.46)

II real wealth quartile

0.04 (5.10)

0.034 (5.14)

0.010 (1.94)

0.015 (3.20)

III real wealth quartile

0.03 (4.28)

0.024 (4.04)

0.013 (2.64)

0.012 (2.75)

IV real wealth quartile

0.024 (3.78)

0.016 (2.98)

0.023 (4.38)

0.016 (3.62)

Observations

7,147

8,001

8,011

8,012

The coefficients in the table indicate the marginal effect of the regressor on the probability of stockownership. Each of the regressors is a dummy variable. z-statistics are reported in parenthesis.

04 - magi

30-05-2007

8:16

Pagina 111

111

STOCK MARKET PARTICIPATION

TABLE 9 – Cross-sectional probit regressions for indirect stockholding EQUITY

VARIABLE

MUTUAL FUNDS

1998

2000

2002

2004

Age 30-39

0.008 (0.70)

0.010 (1.02)

0.017 (1.56)

0.025 (2.06)

Age 40-49

0.0043 (0.38)

0.023 (2.14)

0.012 (1.22)

0.031 (2.50)

Age 50-59

0.013 (1.10)

0.035 (3.01)

0.025 (2.18)

0.032 (2.60)

Age 60-69

0.015 (1.26)

0.017 (1.55)

0.007 (0.65)

0.033 (2.60)

Age 70+

-0.015 (-1.43)

-0.003 (-0.32)

-0.0002 (-0.02)

0.008 (0.86)

High School

0.046 (7.70)

0.039 (8.44)

0.030 (6.65)

0.017 (6.51)

College

0.090 (8.95)

0.064 (8.37)

0.043 (6.09)

0.022 (5.23)

Married

0.016 (2.83)

0.0075 (1.72)

0.010 (2.47)

-0.0006 (-0.26)

Male

0.003 (0.52)

0.006 (1.31)

0.0012 (0.28)

0.0077 (3.60)

II fin wealth quartile

0.099 (4.80)

0.19 (4.23)

0.0075 (1.88)

0.047 (2.66)

III fin wealth quartile

0.26 (9.95)

0.42 (7.16)

0.098 (10.84)

0.13 (5.10)

IV fin wealth quartile

0.43 (14.05)

0.62 (9.39)

0.21 (18.49)

0.24 (6.80)

II real wealth quartile

0.043 (4.52)

0.028 (4.14)

0.0095 (1.41)

0.0054 (1.41)

III real wealth quartile

0.026 (3.14)

0.0075 (1.29)

0.012 (1.88)

0.0060 (1.71)

IV real wealth quartile

0.012 (1.52)

-0.0063 (-1.18)

0.013 (2.09)

0.0028 (0.84)

Observations

7,147

8,001

8,011

8,012

The coefficients in the table indicate the marginal effect of the regressor on the probability to invest in equity mutual funds. Each of the regressors is a dummy variable. z-statistics are reported in parenthesis.

04 - magi

30-05-2007

8:16

112

Pagina 112

ATTILIO GARDINI

- ALESSANDRO MAGI

TABLE 10 – Transition matrix for stock market participation: high school and college DIRECT

TOTAL

DIRECT

PARTICIPATION

IN

PARTICIPATION IN

2004

2000 IN

OUT

TOTAL

IN

9.8

12.6

22.4

OUT

4.1

73.5

77.6

13.9

86.1

100

TOTAL

PARTICIPATION IN

PARTICIPATION IN

2004

2000 IN

OUT

TOTAL

IN

16.2

19.3

35.5

OUT

6.5

58

64.5

22.7

77.3

100

Our elaboration based on data drawn from the 2000-02-04 SHIW panel section (2,522 households).

TABLE 11 – Transition matrix for stock market participation: first grade and lower secondary education DIRECT

TOTAL

DIRECT

PARTICIPATION

IN

PARTICIPATION IN

2004

2000 IN

OUT

TOTAL

IN

1.5

5.2

6.7

OUT

1.6

91.7

93.3

3.1

96.9

100

TOTAL

PARTICIPATION IN

PARTICIPATION IN

2004

2000 IN

OUT

TOTAL

IN

3.2

11

14.2

OUT

3

82.8

85.8

6.2

93.8

100

Our elaboration based on data drawn from the 2000-02-04 SHIW panel section (2,522 households).

04 - magi

30-05-2007

8:16

Pagina 113

113

STOCK MARKET PARTICIPATION

TABLE 12 – Pooled regressions for participation STOCKS

FUNDS

TOTAL

Age 30-39

-0.011 (-0.08)

0.14 (1.07)

0.16 (1.27)

Age 40-49

0.10 (0.78)

0.09 (0.68)

0.21 (1.75)

Age 50-59

0.06 (0.45)

0.12 (0.91)

0.24 (2.00)

Age 60-69

-0.09 (-0.65)

0.012 (0.09)

0.06 (0.55)

Age 70+

-0.13 (-0.95)

-0.23 (-1.65)

-0.14 (-1.10)

High School

0.52 (10.26)

0.38 (7.96)

0.50 (11.40)

College

0.59 (8.68)

0.40 (5.95)

0.52 (8.46)

Married

0.15 (2.48)

0.15 (2.63)

0.14 (2.71)

Male

0.24 (3.98)

0.05 (1.00)

0.15 (3.10)

II fin wealth quartile

0.99 (4.09)

0.78 (4.37)

0.94 (6.08)

III fin wealth quartile

1.66 (7.00)

1.49 (8.63)

1.69 (11.19)

IV fin wealth quartile

2.10 (8.88)

2.07 (11.98)

2.29 (15.11)

II real wealth quartile

0.51 (5.40)

0.28 (3.51)

0.35 (4.84)

III real wealth quartile

0.47 (5.36)

0.19 (2.58)

0.29 (4.35)

IV real wealth quartile

0.48 (5.53)

0.06 (0.77)

0.21 (3.22)

-3.90 (-14.06)

-3.25 (-14.93)

-3.40 (-17.53)

7,420

7,420

7,420

VARIABLE

Constant Observations

The table reports pooled Probit regressions for direct, indirect and total stockholding. The sample uses the pooled data of the 1998-00-02-04 SHIW panel section. We have a total of 7,420 observations. zstatistics are reported in parenthesis.

04 - magi

30-05-2007

8:16

114

Pagina 114

ATTILIO GARDINI

- ALESSANDRO MAGI

TABLE 13 – Panel regressions for participation STOCKS

FUNDS

TOTAL

Age 30-39

-0.27 (-1.20)

0.15 (0.82)

0.093 (0.51)

Age 40-49

-0.004 (-0.02)

0.16 (0.85)

0.24 (1.35)

Age 50-59

-0.034 (-0.15)

0.26 (1.42)

0.33 (1.81)

Age 60-69

-0.24 (-1.02)

0.10 (0.54)

0.08 (0.44)

Age 70+

-0.27 (-1.13)

-0.18 (-0.91)

-0.15 (-0.80)

High School

0.79 (7.66)

0.52 (6.60)

0.72 (9.02)

College

0.84 (5.78)

0.56 (5.02)

0.72 (6.23)

Married

0.34 (2.80)

0.16 (1.75)

0.20 (2.15)

Male

0.24 (2.41)

0.05 (0.62)

0.19 (2.38)

II fin wealth quartile

1.30 (3.76)

1.04 (4.21)

1.29 (5.73)

III fin wealth quartile

2.25 (6.54)

1.95 (8.08)

2.26 (10.13)

IV fin wealth quartile

2.90 (8.34)

2.70 (11.00)

3.08 (13.50)

II real wealth quartile

0.72 (4.32)

0.28 (2.43)

0.39 (3.41)

III real wealth quartile

0.61 (3.79)

0.25 (2.28)

0.35 (3.15)

IV real wealth quartile

0.66 (4.09)

0.085 (0.75)

0.25 (2.27)

Year 00

0.35 (4.12)

0.05 (0.70)

0.21 (3.15)

Year 02

0.093 (1.08)

-0.40 (-5.44)

-0.21 (-2.98)

Year 04

-0.35 (-3.69)

-0.71 (-8.70)

-0.61 (-8.09)

Constant

-5.69 (-13.11)

-4.12 (-13.10)

-4.53 (-15.21)

LR-test: ρ = 0

536.50

275.71

513.19

VARIABLE

The table reports panel data Probit regressions with random effects for direct, indirect and total stockholding. We include year dummies (1998 has been dropped). The sample uses the balanced panel section of the 1998-00-02-04 SHIW data. We have a total of 7,420 observations for 1,855 households; T = 4. z-statistics are reported in parenthesis.