Determinants of Stock Market Participation Among Elderly U ...

34
1 | Page HAVERFORD COLLEGE Determinants of Stock Market Participation Among Elderly U.S. Households with Internet Access Christopher Leung Supervised by Biswajit Banerjee 5/2/2013 This study is an analysis of the determinants of stock market participation,. t looks broadly at the literature concerning stock market participation determinants and limits it to the elderly population in the United States with Internet access. In addition, it looks in depth and differentiates between reduced participation costs and the accessibility of those costs by examining the specific financial activities conducted on the Internet by households. Through econometric analysis, the use of the Internet for financial market research leads to an increase in probability of stock market participation by 21.8% while the use of an online brokerage shows an increase in probability by 38.7%. And so, finance-oriented Internet usage is seen to be important due to the widespread availability of the Internet, rendering regular Internet use less meaningful in the context of stock market participation.

Transcript of Determinants of Stock Market Participation Among Elderly U ...

Page 1: Determinants of Stock Market Participation Among Elderly U ...

1 | P a g e

HAVERFORD COLLEGE

Determinants of Stock Market Participation Among Elderly U.S. Households with Internet Access

Christopher Leung

Supervised by Biswajit Banerjee 5/2/2013

This study is an analysis of the determinants of stock market participation,. t looks broadly at the literature concerning stock market participation determinants and limits it to the elderly population in the United States with Internet access. In addition, it looks in depth and differentiates between reduced participation costs and the accessibility of those costs by examining the specific financial activities conducted on the Internet by households. Through econometric analysis, the use of the Internet for financial market research leads to an increase in probability of stock market participation by 21.8% while the use of an online brokerage shows an increase in probability by 38.7%. And so, finance-oriented Internet usage is seen to be important due to the widespread availability of the Internet, rendering regular Internet use less meaningful in the context of stock market participation.

Page 2: Determinants of Stock Market Participation Among Elderly U ...

2 | P a g e

I. Introduction

In 1985, Mehra and Prescott proposed that an alternative model to the traditional Arrow-

Debreu general would better model the observed historical differential returns of risky and

riskless assets (shown in Table 1) by incorporating additional frictional factors such as

transaction, information, and entry costs – soon after, an interest developed around the low stock

market participation rates themselves (i.e. stock market participation puzzle) to respond to the

equity premium puzzle and has consequently sparked numerous studies seeking important stock

market participation determinants that address an easing of the above mentioned market frictions.

The equity premium puzzle and stock market participation puzzle have become entangled

in a sub-strand of the literature due to the large societal shift that occurred during the late 1980s

and has defined the 1990s and early 2000s, i.e. the mass adoption of computers and subsequently

the Internet. Why may this be of large potential significance? If we look at the financial

participation rates from 1989 to the present, as shown in Table 2, it can be seen that there was a

steady positive trend in stock market participation, which started in 1989 and peaked in 2001.

After having peaked, the U.S. stock market participation rate has declined ever since. The current

trend can be traced to the collapse of the “Internet bubble” and extended by the 2008 Financial

Crisis. Participation in the other financial markets (bonds, transactional accounts, and certificates

of deposit) follows a similar trend and confirms the hypothesized primary cause. With the asset

bubble bursts, the effect of the Internet is hidden in these numbers, yet the Internet is here to stay

and it would be ill-informed to not take into account its effect on stock market participation rates

along with the determinants integral in modeling these participation rates.

Older studies have overlooked the trend towards universal access to the Internet for U.S.

households (from the 2010 U.S. Census, it has been reported that 60.1% of individuals sampled

Page 3: Determinants of Stock Market Participation Among Elderly U ...

3 | P a g e

had access to the Internet), and focused their examination on simple Internet usage. It is idealistic

to think that the reduction in participation costs are distributed evenly and fairly, however from a

more realistic stance, reduction in participation costs are more likely privileges, rather than rights

– i.e. as applies to most thing, it is only available to a select few, rather than to all. And so, I

believe the question I wish to address changes from how regular Internet use affects households

and their stock market participation decision to how specific households are able to take

advantage/exploit reduced costs with finance-oriented Internet usage. In the following study, I

examine various household characteristics, focusing on how households use the Internet to

access these reduced costs with my elderly U.S. household sample.

The remainder of this paper reviews older literature in Section II, provides a data

description in Section III, and presents the econometric analysis and results in Section IV, V, and

VI. Results and impact of this study are summarized and enunciated in the final section, Section

VII.

II. Literature Review

In a bottom-up approach, much of the literature has analyzed household behavior through

the determinants of stock market participation rates, which is defined as the percentage of U.S.

households that own stocks. I intend to inform my study with their conclusions, ensuring that the

literature is applicable to my sample, which treats stockholding in elderly U.S. households.

In studies looking at the overall U.S. stock market participation rate, the models make use

of a dependent variable that is a binary variable for whether or not a household owns stock at a

particular time t. In the literature surveyed, seven out of ten papers use this binary variable. The

other three papers are less focused on the stockholding aspect of the stock market; Choi, Laibson,

and Metrick (2000) and Barber and Odean (2002) focus primarily on trading and the profile of

Page 4: Determinants of Stock Market Participation Among Elderly U ...

4 | P a g e

traders and as a result, they use dependent variables that treat trade frequency and usage of an

online broker. Cole and Shastry (2009) look at cognitive ability and its effect on financial market

participation, and therefore use amount of investment income as their dependent variable. Since

these three papers do not exactly focus on stock market participation determinants, but rather

trader profiling and education effects on financial markets, it is not unusual that the dependent

variables in these studies are different.

Returning to the binary variable used as a dependent variable by the majority of the

relevant literature, we find that it is built and conceptualized through the Consumption of

Capital Assets Pricing Model (CCAPM). It is the primary model used to predict the optimal level

of capital asset consumption and has led to the claim that on average, U.S. households do not

appreciate the equity premium present in the stock market, shown by their sub-optimal level of

stocks in their portfolio.

Bogan (2008) builds a calibrated CCAPM model from the standard frictionless CCAPM

by adding constraints, i.e. information costs and transaction costs, satisfying the necessary

frictional conditions suggested by Mehra and Prescott (1985). This model has its base in a

household’s utility optimization equation in regard to its allocation of disposable income to

capital assets. Because theory expects that individuals maximize their utility function,

economists are able to model the optimal allocation levels from an indirect utility derived from

the CCAPM. Following Bogan’s approach in defining the dependent variable, one differentiates

between the assumed linear indirect utility function of stockholders,𝑈𝑠𝑖 = 𝑋𝑠𝑖𝐵𝑠 + 𝑒𝑠𝑖, and that

of non-stockholders, 𝑈𝑛𝑠𝑖 = 𝑋𝑛𝑠𝑖𝐵𝑛𝑠 + 𝑒𝑛𝑠𝑖, where Xi are observable characteristics of household

i and ei is the error term. Because the indirect function is unobservable, one must turn to the

participation decision of household i, Di – to participate or to abstain. If we let 𝐷𝑖 = 1 when

Page 5: Determinants of Stock Market Participation Among Elderly U ...

5 | P a g e

𝑈𝑠𝑖 > 𝑈𝑛𝑠𝑖, we would say that a household participates in the stock market because their indirect

utility is greater when holding stocks than when they are not. Following suit, 𝐷𝑖 = 0 when

𝑈𝑠𝑖 < 𝑈𝑛𝑠𝑖 , since the household will abstain from the stock market since their utility is not

maximized when they hold stocks. These outputs, 𝐷𝑖 = 1 𝑎𝑛𝑑 𝐷𝑖 = 0, are the two outcomes of

this binary dependent variable.

Leaving the left side of the equation and moving towards the right side, we turn our focus

to the household characteristics that affect the probability that a household owns stock at time t.

To organize my approach of the significant and valid explanatory variables, I look to provide a

broad perspective by reviewing significant stock market participation determinants applicable to

the whole of the United States, which I classify into two categories: (1) general household

characteristics and (2) Internet-specific determinants. As it is hard to define such categories that

are mutually exclusive, I navigate the first section by addressing individual independent

variables found to be significant in the literature and group them under the general household

characteristics designation. In the second section, I focus on Internet-specific variables as they

address specifically the new environment in which we find ourselves.

General Household Characteristics As Determinants Of Stock Market Participation Age

Given the survey data often used in the literature, some identifying household

characteristics are significant in determining stock market participation. From the literature, age

is generally agreed to be an important determinant and can be explained by a variety of reasons.

However, there exists support of both statistically significant negative and positive coefficients

on the age variable. Explanations concerning the negative correlation between stockholding and

age are defended by the finite investment horizon of households as well as age-dependent risks,

such as health risks (Vissing-Joregenson 2002; Cole and Shastry 2009), while those that have

Page 6: Determinants of Stock Market Participation Among Elderly U ...

6 | P a g e

positive coefficients on their age variable look at the potential for exogenous information

revelation, i.e. households learn more about the stock market randomly over time, such that age

increases the probability of stock ownership (Bertaut 1998).

For those papers that do not find age to be significant such as Haliassos and Bertaut

(1995), the lack of significance in the age variable can be explained by the fact that the time-

horizon effect may have already been captured by a liquidity-preference variable, defined as how

willing one would is to give up liquidity for a higher return on investment. Combined with a

finite investment horizon and life-cycle, it is likely older households prefer higher liquidity over

higher returns.

Nonetheless, in the majority of the literature we find that the age variable maintains a

negative, yet significant relationship with stock market participation. (Choi, Laibson, and

Metrick 2000; Barber and Odean 2002; Vissing-Jorgenson 2002, Hong, Kubik, and Stein 2004;

Bogan 2008; Glaser and Klos 2012)

Labor Income and Financial Wealth

The finances of a household are naturally an important factor in a household’s decision to

participate in the stock market. Often used jointly and sometimes individually, labor income and

financial wealth are the main indicators of a household’s finances. In respect to the stock market

participation relationship, labor income’s effect is inconsistent in current studies, while financial

wealth has been found to be vital to the stock market participation decision.

Bertaut (1998) and Barber and Odean (2002) find that the labor income variable is

statistically insignificant but positive, whereas the rest of the papers either don’t use the labor

income variable (Vissing-Jorgenson 2002; Hong, Kubik, and Stein 2004) or find it positive and

significant in its relationship with stockholding (Haliassos and Bertaut 1995; Choi, Laibson, and

Page 7: Determinants of Stock Market Participation Among Elderly U ...

7 | P a g e

Metrick 2000) Labor income plays into the decision as it allows a household to plan their

consumption and follow the traditional consumption smoothing theory where households that

have excess income have the tendency to save/invest in order to maintain their same level of

consumption in the future. To shed additional light on labor income as a determinant of stock

market participation, Cole and Shastry (2009) note “the share of individuals who participate in

financial markets increases as at a decreasing rate with total income, reaching a peak of

approximately 60% for households with earned income levels of $150,000.” And so, we see that

there is ambiguity in the literature concerning the place of labor income in the stock market

participation decision.

Financial wealth, as it is defined, is an asset and can be drawn upon for consumption or

investment. In its relation to stock market participation, Vissing-Jorgenson (2002) finds that “a

per period stock market participation cost of just 50 dollars is sufficient to explain the choices of

half of stock market nonparticipants.” So, the literature finds, in effect, that wealthier households

have more to invest which renders insignificant the effect of the fixed costs to stock market

participation as a barrier to entry. The literature (Haliassos and Bertaut 1995; Bertaut 1998;

Vissing-Jorgenson 2002 and Hong, Kubik, and Stein 2004; Lusardi, Rooij, and Alessie 2007)

finds unanimously that wealth holds a positive and significant correlation with the stockholding

decision for households.

Education and Financial Literacy From the literature, the consensus is that education is a primary determinant of stock

market participation and is positively correlated in its relationship. The education variable is seen

in various forms, most often in dummy variable form, divided into less than high school, high

school, and college degree or the international equivalents. It is often interpreted that education

works through the cost channel, specifically the entry cost channel and information cost channel.

Page 8: Determinants of Stock Market Participation Among Elderly U ...

8 | P a g e

This may be due to the ability of more educated households to understand the stock market better,

open financial accounts more easily, etc. (Haliassos and Bertaut 1995; Hong, Kubik, and Stein

2004). Secondly, it may be seen to increase the ability of households to process information

and/or used as a proxy for future expected earnings. (Bertaut 1998) And so, education is

interpreted to be integral since it allows households to access the reduced information costs, if

there are any, and take advantage of the information as well as indicate a boost in future labor

income and investment income.

In a study dedicated to education and financial market participation, Cole and Shastry

(2009) use instrumental variable analysis to establish a causal link between the two. They

conclude that education is an integral factor in encouraging households to participate in various

financial markets (stocks, bonds, mutual funds, savings, IRAs, CDs, etc.) They estimate the

effect of an additional year of schooling on stock market participation to be around +1.5%.

Financial literacy is another interesting variable to consider; as it has been shown to

reduce information costs to a level that makes the stock market more attractive to households. It

works through a channel similar as education and can be explained by the ability or inability to

process this financial information. Working through the information cost channel, it is likely that

those with a higher level of financial literacy are more able to apply their knowledge to matters

concerning the financial markets, thus able to “access” the already reduced information costs

created by the expansion of financial resources and tools and media attention.

In addition, the financial literacy variable is gaining in importance as individuals are

claiming more responsibility over their financial futures with the advent of new financial

products and the increasing customizability of retirement plans. (Lusardi, Rooij, and Alessie

2007; Cole and Shastry 2009)

Page 9: Determinants of Stock Market Participation Among Elderly U ...

9 | P a g e

Lusardi, Rooij, and Alessie (2007) approach stockholding by way of financial literacy

and measure it with two special modules, which are composed of an “extensive list of questions

aimed at measuring and differentiating among different levels of literacy and financial

sophistications.”In the first module, they look at basic financial literacy, asking questions about

concepts such as inflation and interest rate. In the second, they look at more advanced concepts

such as valuation, difference between asset classes, and portfolio diversification strategies. As

theory expects them to, financial literacy increases the probability that a household owns stock

because stocks are relatively complex assets. Accordingly, it takes financial knowledge to

understand stock market processes and fundamental stock analysis – and so, financial literacy

reduces information costs by increasing the efficiency of financial information processing. From

their empirical analysis, Lusardi, Rooij, and Alessie (2007) conclude that with increased

financial literacy comes a higher probability that a household will participate in the stock market

through their ability to accumulate wealth, formulate retirement plans, as well as navigate the

stock market. Their finding is statistically significant.

Internet-Specific Variables as Determinants Of Stock Market Participation The literature on stock market participation and the Internet is very limited. In this

respect, Stock Market Participation and the Internet(2008) by Vicki Bogan and Causal Evidence

on Regular Internet Use and Stock Market Participation (2012) by Glaser and Klos are the only

papers that address this relationship directly to my knowledge. Bogan makes use of a probit

model, in which computer usage is a proxy for Internet usage. From her empirical results, she

posits that computer usage is significant in a household’s decision to own stock. However, her

conclusions are applied to a broad population though her sample does not seem to be powerful

enough to make such a statement. Furthermore, she assumes that all households experience the

reductions in information costs, which is questioned by Glaser and Klos (2012), who approach

Page 10: Determinants of Stock Market Participation Among Elderly U ...

10 | P a g e

stock market participation more realistically. Glaser and Klos attempt to establish a causal

relationship between the Internet and stock market participation rates through instrumental

variable analysis. Additionally, they pay particular attention to the interaction effect between

financial literacy and Internet usage, which has enlightened the conversation concerning the

accessibility of reduced frictional costs. This accessibility factor is central to my study.

Conclusions from both papers are similar and I will detail them here so that we can gain a

sense of what has already been explored in terms of Internet-specific variable. From Bogan’s

probit model, she estimates that computer usage increases a household’s probability of holding

stock by at least 3.4%. In a second empirical model, Bogan employs a true Internet dummy

variable (survey questions about Internet usage rather than computer usage) and estimates that

Internet usage increases a household’s probability of owning stock by about 31%. Her findings

indeed show significance in the relationship between stock ownership and Internet. Some

shortcomings of her model is seen in the broad generalizations she makes about all U.S.

households even though her sample has an average age of 65 years. She attempts to substantiate

her claim by citing the work of Ameriks and Zeldes (2001) who found that “equity portfolios

shares increased strongly with age.” However, it seems that Bogan simplifies the results of

Ameriks and Zeldes when we compare the claim with the data provided by Lusardi, Rooij,

Alessie (2007) who show that stock market participation rates do increase with age, but only up

until the age of 40 years; afterwards, the relationship levels off and remains fairly constant.

Therefore, Bogan’s assumption may be problematic when one interprets her results and leads to

a suspicion that Bogan’s results are truly not applicable to the whole of the United States.

Glaser and Klos (2012) search for a causal effect of regular Internet use on stock market

participation through the channel of reduced information costs and employ instrumental variable

Page 11: Determinants of Stock Market Participation Among Elderly U ...

11 | P a g e

analysis. The instrumental variables were created from the responses, based on a Likert scale

ranging from 1 (totally disagree) to 4 (totally agree), to two questions: (1) “Computers and other

modern electronic hardware are simply fun” and (2) “I fear that the technical progress will

destroy our lives.” Both are valid as instrumental variables since the statements are both highly

correlated to Internet usage (F-statistic is between 41 and 45; threshold value for strong

instruments is 10) and exogenous to stock market participation (computers used for fun and

attitude towards technology do not have much of a relationship with stockholding). And so, these

assessment questions were exploitable as instrumental variables for Glaser and Klos. In the

empirical model, each of the variables is split into three dummy variables, with the omitted

variable being the response, which was equal to 1 (totally disagree). The model yielded a

participation rate of about 43.6% in households that have regular Internet using households and

14.5% among non-regular Internet using households. This result is consistent with Bogan’s

empirical analysis and conclusion in 2008. In their second model, Glaser and Klos hypothesize

that the channel of causality is that of financial literacy. Glaser and Klos find that that lower

transaction costs have no effect on the probability of holding stocks for financially illiterate

households, though the effect was very significant among the financially literate households,

yielding an increase in probability of stockholding by 8.56%. This specific analysis of the

interaction effect between financial literacy and stockholding informs the construct of

accessibility of reduced market frictions and the approach that I assume.

III. Data

To consider the relationship between stock ownership by elderly U.S. households and

how they use the Internet, I will employ the 2009 HRS Internet Survey, a specialized study that

is conducted on the Internet and collects information on independent U.S. households that

Page 12: Determinants of Stock Market Participation Among Elderly U ...

12 | P a g e

includes financial decisions pertaining to the stock market. The 2009 Internet Survey included

4,433 observations. For a summary of how the sample was created, please refer to Table 3.

Although there are other various databases that have been represented in the literature,

the 2009 HRS Internet Survey is the first of its kind that provides specific questions that can be

used to create more direct variables, and is the driving force for my decision to use this specific

survey, even the HRS Internet Surveys earlier in the series lack the specificity in the questions

needed to look at the relevance of the Internet to the stock market.. Thus, the 2009 HRS Internet

survey is the first of its kind that has allowed such an approach to stock market participation

rates, when one considers Internet usage for financial activities as a determinant.

The data description for the two survey questions that are the basis for this study and

model are the following:

How often do you do each of the following activities on the Internet? Buy or sell stocks, mutual funds, or bonds online ............................................................................ 3692 1. Never 332 2. Rarely 238 3. Sometimes 118 4. Often 53 9. QUESTION SKIPPED ================================================================ How often do you do each of the following activities on the Internet? Get financial information online, such as stock quotes or mortgage interest rates ............................................................................ 1855 1. Never 898 2. Rarely 869 3. Sometimes 686 4. Often 125 9. QUESTION SKIPPED

The questions listed above are essential to this study since they provide a more direct

view in how the Internet affects stock market participation through these particular channels, an

online stock brokerage as well as financial research conducted on the Internet. This is where my

study finds its niche and brings us closer to our first hurdle of determining whether or not the

Page 13: Determinants of Stock Market Participation Among Elderly U ...

13 | P a g e

Internet is a significant determinant of stock market participation and on our way to observing its

effect on overall stock market participation rates in elderly U.S. households. Having laid out the

determinants that have already been proven to be significant, I look to use these determinants

along with my Internet variables (the use of the Internet specific for financial research which

may reduce information costs and the use of an online brokerage which may reduce transaction

and entry costs) on the right side of a probit model which I will detail in the next section.

These two Internet variables are essential to my analysis since they embody the use of the

finance-oriented Internet usage in respect to stock market participation issues. The use of the

Internet for financial research looks at how households utilize the Internet specifically for

financially research, which may provide support for reduced information costs, due to increased

media attention to the stock market and an overall surge in financial news, for households that

are able to process the information. The second variable, the use of an online brokerage, obtains

its importance from its ability to reduce transaction and entry costs given the rise of online

brokerages, which has increased competition and reduced fees and commissions in the entire

industry as a whole. In addition, the participation costs are further reduced by the ease of simple

user interfaces of online brokerages and the increased efficacy in submitting orders. With these

two variables, we look singularly at how the Internet affects elderly U.S. households with

Internet access.

IV. Methodology

In my baseline specification, I use the dependent variable as defined above on the left

side of my equation in a probit model that introduces my two Internet-specific variables along

with control variables such as age, education, financial wealth, financial literacy variables, and a

proxy for income. It takes the form:

Page 14: Determinants of Stock Market Participation Among Elderly U ...

14 | P a g e

(1)𝑆𝑇𝑂𝐶𝐾𝑆𝑂𝑉𝐸𝑅𝐴𝐿𝐿= 𝛽0 + 𝛽1𝐴𝐺𝐸 + 𝛽2𝐸𝐷𝑈 + 𝛽3𝑊𝐸𝐴𝐿𝑇𝐻 + 𝛽4𝐹𝐼𝑁𝑅𝐸𝑆𝐸𝐴𝑅𝐶𝐻+ 𝛽5𝑂𝑁𝐿𝐼𝑁𝐸𝐵𝑅𝑂𝐾𝐸𝑅𝐴𝐺𝐸 + 𝛽6𝐹𝑂𝐿𝐿𝑂𝑊𝑆𝑇𝑂𝐶𝐾 + 𝛽7𝑈𝑁𝐷𝐸𝑅𝑆𝑇𝐴𝑁𝐷𝑆𝑇𝑂𝐶𝐾+ 𝛽8𝑊𝑂𝑅𝐾𝑆_𝑃𝐴𝐼𝐷 + 𝜀𝑖

The variables that I have employed in this model are listed below: DEPENDENT VARIABLE DESCRIPTION STOCKSOVERALL Binary variable for owning stock (1 – yes, 0 – no) INDEPENDENT VARIABLES DESCRIPTION AGE Continuous Variable that denotes age of household head EDU Continuous Variable that denotes years of education for household head WEALTH Continuous Variable that denotes Household Financial Net Worth (in $100,000) FINRESEARCH Dummy variable for using the Internet for online research (1 – yes, 0 – no) ONLINEBROKERAGE Dummy variable for using an online brokerage (1 – yes, 0 – no)

FOLLOWSTOCK Dummy variable for self-assessment rating of how closely one follows stock market (1 – yes, 0 – no); proxy for financial literacy

UNDERSTANDSTOCK Dummy variable for self-assessment rating of how much one understands stock market (1 – yes, 0 – no); proxy for financial literacy

WORKS_PAID Dummy variable for working for pay (1 – yes, 0 – no) ; proxy for income My dependent variable is a binary variable that indicates whether a household owns stock

or not. I have defined stockholding to be the holding of any stock in any form, including directly

or indirectly. Direct holdings are those that are held outside of a managed asset and are most

likely held in the form of shares in individual public corporations. On the right side, I have eight

explanatory variable, with two model-specific variables and six control variables: continuous

variable in age (measured in years), continuous variable in education (measured in years),

continuous variable in wealth (measured in $100,000), binary variable for self-assessment on

how closely one follows the stock market and how well one understands the stock market

(proxies for financial literacy), as well as a binary variable for works for pay. The two model-

specific variables are Internet-related and account for the usage of the Internet specifically for

financial research as well as the use of an online brokerage for executing trades online. Both

indicate frequent use of the Internet for financial matters.

V. Results

Page 15: Determinants of Stock Market Participation Among Elderly U ...

15 | P a g e

From my baseline probit model on the stockholding decision, age and education are both

shown to be significant explanatory variables. The negative coefficient on the age variable

reflects the lower propensity for households to participate in the stock market as the household

grows older; the coefficient is statistically significant and this can be attributed to the fact that the

age variable captures the time horizon-effects; in Haliassos and Bertaut (1995), their age

variables are rendered insignificant due to the inclusion of a liquidity preference variable which

has already captures the horizon effects. With no liquidity variable, age is significant at the 10%

level and reduces the probability that a household owns stock, suggesting that households tend to

dis-save later in the life-cycle in order to maintain consumption levels. In addition, it would seem

that the lack of a bequest motive variable in this model also adds to the significance of the age

variable since bequests would in essence increase the time-horizon for households, encouraging

households to maintain their capital asset consumption level. The coefficient on the education

variable is positive and significant, showing that an increased access to information and more

efficient information processing increases the probability that a household owns stock. With

lower stock market participation costs, these households find the market more attractive since

they are more likely able to demystify the stock market.

Financial wealth is found to be positive and significant at the 1% level. As the availability

of excess funds increases, capital asset consumption increases (i.e. stockholding) since

households are not financially constrained in the short-term; thus they are more forward thinking

and more inclined to invest these excess assets so that they may smooth their consumption. The

income variable is negative, but insignificant. As measured, the income variable indicates solely

whether or not the household receives labor income. Given the roughness of the measurement,

the insignificance may be attributed to the variability within the income variable itself. The lack

Page 16: Determinants of Stock Market Participation Among Elderly U ...

16 | P a g e

of a true income variable is a large drawback of model, but in a way it agrees with the literature’s

confounded thoughts on the variable as well as its statistical insignificance.

The self-assessment variables of how closely one follows the stock market and how well

one understands the stock market are both positively correlated, though the former is significant

and the latter is not. Increased familiarization with the stock market reduces information costs

through the same channel as education. In respect to the latter variable, there is potential for

response bias to be present, since the question considers an individual’s ability. This may have

led to an inflated perception of the respondent’s own ability.

As anticipated, both the Internet variables I used were both significant and positive.

Given the widespread access of the Internet, the ability to focus singularly on how people utilize

the Internet, in respect to their finances, becomes the more interesting question. Already, the

Internet has reduced entry costs (lower fees and commissions with the rise of online brokerages)

and information costs (increased media coverage of the financial markets).

With the positive correlation between the use of the Internet for financial research and

stockholding, we see that households who make use of the Internet specifically for financial

research are more inclined to hold stocks. This is important to note because it means that the

Internet itself is not a major factor in the stock market participation decision, but rather the

manner in which a household employs the Internet as a tool. For the online brokerage variable,

the model provides a positive coefficient that is significant at the 5% level and allows us to

explore the Internet’s role in increasing access to the stock market, both through the simplicity of

the interface as well as the availability of various brokerages that target certain households. With

both the reduction in information costs and transaction costs, we can see that the Internet

Page 17: Determinants of Stock Market Participation Among Elderly U ...

17 | P a g e

variables that I have used in the model are valuable additions to the literature since they are not

proxies but rather direct measurements of the information/transaction channel itself.

In Table 6, I have detailed the marginal effects of model 1 which allows us to translate

roughly the effect of the independent variables in percentage terms from the probit model. The

results yield that the use of Internet for financial research increases the probability that a

household owns stock by about 21.8% while the use of an online brokerage improves it by

38.7%

In addition, I would like to highlight my empirical results by building a model for

comparison, which I will create from my sample. This model looks at the effect of regular

Internet usage on stock market participation.

(0)𝑆𝑇𝑂𝐶𝐾𝑆𝑂𝑉𝐸𝑅𝐴𝐿𝐿= 𝛽0 + 𝛽1𝐴𝐺𝐸 + 𝛽2𝐸𝐷𝑈 + 𝛽3𝑊𝐸𝐴𝐿𝑇𝐻 + 𝛽4𝐸𝑀𝐴𝐼𝐿 + 𝛽5𝐹𝑂𝐿𝐿𝑂𝑊𝑆𝑇𝑂𝐶𝐾+ 𝛽6𝑈𝑁𝐷𝐸𝑅𝑆𝑇𝐴𝑁𝐷𝑆𝑇𝑂𝐶𝐾 + 𝛽7𝑊𝑂𝑅𝐾𝑆_𝑃𝐴𝐼𝐷 + 𝜀𝑖

The model employs the same dependent variable as used in the baseline specification, but

deviates from the primary specification in its explanatory variables. Excluding the finance-

oriented Internet usage variables (use of internet for financial research and use of an online

brokerage) and replacing them with a proxy for internet usage (defined as usage of e-mail), I

recreate a model that ignores the accessibility factor of reduced participation costs. This serves as

a point of comparison for the results of my baseline specification. The results of this model are

motivating because the Internet variable (usage of e-mail) is found to be insignificant in

explaining stock market participation, lending credibility to the new question that I have posed:

is Internet access or how a household utilizes the Internet more important when considering their

stock market participation decision.

VI. Robustness Specifications

Page 18: Determinants of Stock Market Participation Among Elderly U ...

18 | P a g e

In the next specifications, I will deconstruct my dependent variable so that we may see

the effect of the Internet on the various types of stockholding, thus changing my fairly liberal and

generous original definition to one that is more specific; we will look at stocks in trust funds,

IRA/KEOGH accounts, 401k plans, mutual funds, and individual public corporations.

(2 − 7)𝑆𝑇𝑂𝐶𝐾𝑆𝑆𝑃𝐸𝐶𝐼𝐹𝐼𝐶𝑇𝑌𝑃𝐸= 𝛽0 + 𝛽1𝐴𝐺𝐸 + 𝛽2𝐸𝐷𝑈 + 𝛽3𝑊𝐸𝐴𝐿𝑇𝐻 + 𝛽4𝐹𝐼𝑁𝑅𝐸𝑆𝐸𝐴𝑅𝐶𝐻+ 𝛽5𝑂𝑁𝐿𝐼𝑁𝐸𝐵𝑅𝑂𝐾𝐸𝑅𝐴𝐺𝐸 + 𝛽6𝐹𝑂𝐿𝐿𝑂𝑊𝑆𝑇𝑂𝐶𝐾 + 𝛽7𝑈𝑁𝐷𝐸𝑅𝑆𝑇𝐴𝑁𝐷𝑆𝑇𝑂𝐶𝐾+ 𝛽8𝑊𝑂𝑅𝐾𝑆_𝑃𝐴𝐼𝐷 + 𝜀𝑖

Looking at the individual investment accounts and thus redefining our dependent variable

to be whether or not a household holds stocks in a specific account type, we achieve a more

narrow approach. Similar results are produced from these robustness regressions; however there

are few notable deviations from the baseline specification. Education maintains its positive and

significant relationship with stockholding in all account types, except in trust funds, which may

make sense since beneficiaries do not have administrative access to the portfolio. When we look

at the age variable, we see that there is some contradictory evidence to what has been shown in

the baseline specification in terms of significance and sign of the coefficient. All account types

besides the 401K plan show a positive and significant correlation between age and stockholding.

It is interesting to note though that the age variable in the 401K plan specification is negative and

is the only specification to reflect the same result as the baseline specification.

In all specifications, financial wealth is positive and significant, confirming the findings

in the baseline specification. The income variable is statistically insignificant in all specifications,

however is negative in specifications 2 to 4 and positive in specifications 5-7.

The coefficient on the variable, how closely one follows the stock market, is generally

significant and positive, however we see a deviation (in that it is negative and insignificant) in

the trust fund specification, which once again potentially explained by the lack of administrative

Page 19: Determinants of Stock Market Participation Among Elderly U ...

19 | P a g e

access to beneficiaries of trust funds. In respect to the variable, how well one understands the

stock market, we find that it is insignificant throughout the robustness specifications though there

is variability in the coefficient where half of the robustness specifications result in a positive

coefficient and the other half, a negative coefficient, for this self-assessment variable.

Internet-specific variables, use of Internet for financial research and use of an online

brokerage, are proven to be ambiguous when the definition of stockholding is deconstructed.

When looking at the Internet as a tool for financial research, we find that the variable is positive

and significant in the 401K, Individual Company, and IRA/Keogh specifications and statistically

insignificant in the remaining specifications, trust funds and mutual funds. This may be

explained by the fact that trust funds are controlled and thus advised by the contributor/owner of

the trust fund, and not the beneficiary. For mutual funds, financial research may be seen to be

insignificant since there is a portfolio manager and less research is required when investing in

mutual funds. The online brokerage variable is rendered insignificant in two of the

specifications; 401K and trust fund, but exhibits a positive and significant relationship in the

other specifications. I suggest that the online brokerage variable in the 401K specification is

insignificant since the 401K is an employer-sponsored plan; thus, companies may contribute and

match employee contributions with company stock and options, reducing the use of brokerages

in general. For trust funds, the explanation follows from the previous considerations of trust

fund’s insignificance in the various variables.

VII. Conclusion

From my empirical results, I conclude that finance-oriented Internet usage increases the

propensity of elderly U.S. households to hold stock in their investment portfolio. With additional

robustness specifications and controls, I confirm my hypothesis that access to the reduced

Page 20: Determinants of Stock Market Participation Among Elderly U ...

20 | P a g e

participation costs is significant and indeed a matter to consider. In addition, I attempt to

introduce the concept of accessibility of reduced participation costs through finance-oriented

Internet usage with the goal of re-approaching reality in economic models.

Since Internet has become widely available, the question that needs to be answered must

be more specific. This has motivated me to focus on the accessibility of reduced participation

costs, and question whether the reduced participation costs are experienced by all households.

From my empirical results, I find that it is important to consider the specific effect rather than the

mass effect. In doing so, my models predict that the households in question tend to have higher

stock market participation rates when they use the Internet for specifically financial reasons.

From the HRS 2009 Internet Survey, use of the Internet for financial research increases the

probability of owning stock by 21.8% while the use of an online brokerage boosts that

probability by 38.7%.

Page 21: Determinants of Stock Market Participation Among Elderly U ...
Page 22: Determinants of Stock Market Participation Among Elderly U ...

20 | P a g e

Time PeriodMean SD Mean SD

1889-1978 0.80 5.67 6.98 16.54

1889-1898 5.80 3.23 7.58 10.021899-1908 2.62 2.59 7.71 17.211909-1918 -1.63 9.02 -0.14 12.811919-1928 4.30 6.61 18.94 16.181929-1938 2.39 6.50 2.56 27.901939-1948 -5.82 4.05 3.07 14.671949-1958 -0.81 1.89 17.49 13.081959-1968 1.07 0.64 5.58 10.591969-1978 -0.72 2.06 0.03 13.11

Observed Historical Return Differential

% real return on a riskless security*

% real return on S&P 500

Source: Mehra and Prescott (1985)

*Securities used for data were 90-day T-bills from 1931-1978, T-Certificates from 1920-1930, and 60-day to 90-day Prime Commercial paper for prior to 1920

Year

% of families with stock holdings

% of families with directly held stocks

% of families with savings bonds

% of families with directly held bonds

% of families with transaction accounts

% of families with certificates of deposit

1989 31.80 16.80 23.90 5.70 85.50 19.901992 36.90 17.00 22.30 4.30 86.90 16.701995 40.50 15.20 22.80 3.10 87.40 14.301998 48.90 19.20 19.30 3.00 90.60 15.302001 53.00 21.30 16.70 3.00 91.40 15.702004 50.30 20.70 17.60 1.80 91.30 12.702007 53.20 17.90 14.90 1.60 92.10 16.102010 49.90 15.10 12.00 1.60 92.50 12.20

Notes:

Financial Market Participation Rates for All U.S. Households

Source: 2010 Survey of Consumer Finances

A Transaction Account is a liquid account that is primarily comprised of checking, savings, and money market deposit accounts, and money market mutual funds; Certificates of Deposits are interest-bearing deposits with a set term; Savings Bonds refer to a bond that is subject to a fixed interest rate for a fixed period of time; Bonds refers to mortgage-backed bonds and corporate or foreign bonds , tax-exempt and other government bills.

*Direct holdings are those held outside of a managed asset such as mutual funds, trusts, managed investment accounts, annuities, and tax-deferred retirement accounts.

Table 1

Table 2

Page 23: Determinants of Stock Market Participation Among Elderly U ...

21 | P a g e

Observations Mean Std. Dev. Min. Max.

Age (in years) 4381 65.72 8.98 32 96

Wealth (in hundreds of thousands of dollars) 2611 3.67 7.03 0.00001 87

Education (in years) 4370 14.18 2.24 0 17

Use of Internet for Financial Research (1 = yes, 0 = no) 4308 0.57 0.50 0 1

Use of Online Brokerage (1 = yes, 0 = no) 4380 0.08 0.27 0 1

Follows Stock Market Closely (1 = yes, 0 = no) 4358 0.61 0.49 0 1

Understands Stock Market Well (1 = yes, 0 = no) 4345 0.83 0.37 0 1

Works for Pay (1 = yes, 0 = no) 4425 0.46 0.50 0 1

Summary Statistics

Contact letters were sent to 5,742 HRS respondents, inviting them to participate in the 2009 Internet Survey. The 2009 Internet sample was drawn from respondents who reported Internet access in the HRS 2008 Core survey, plus those who did not respond to the 2008 Core survey but had been selected for the 2003, 2006, or 2007 Internet surveys.

As was the case in prior HRS Internet surveys, roughly 20% of the eligible pool was reserved for a control group. A total of 4,433 respondents completed the 2009 Internet Survey, for a simple response rate of 77.2%.

2009 Health and Retirement Internet Survey Description

The 2009 Internet Survey is the fourth in a series of surveys conducted on the Internet. Completed interviews were obtained from 4,433 HRS respondents.

Table 3

Table 4

Page 24: Determinants of Stock Market Participation Among Elderly U ...

22 | P a g e

Regular(0)

Baseline(1)

401K(2)

Trust Fund(3)

Mutual Funds

(4)

Individual Companies

(5)IRA/Keogh

(6)

Intercept 0.5744 0.5838 1.0689 -0.4348 -0.0216 -0.417 -0.0685(0.0848) (0.0811) (0.1055) (0.0715) (0.1052) (0.103) (0.1109)

Years of Education (in years) 0.012*** 0.0119*** 0.0135*** 0.0028 0.0156*** 0.0111** 0.0217***(0.0038) (0.0038) (0.0049) (0.0033) (0.0049) (0.0047) (0.0051)

Age (in years) -0.0017** -0.0018** -0.0149*** 0.0071*** 0.0025** 0.006*** 0.0005(0.0009) (0.0009) (0.0011) (0.0008) (0.0011) (0.0011) (0.0012)

Financial Wealth (in hundred of thousands of dollars) 0.0062*** 0.00642*** 0.00512*** 0.0176*** 0.01*** 0.01*** 0.01***(0.0011) (0.0012) (0.0016) (0.001) (0.002) (0.002) (0.002)

Use of Internet for Financial Research (1 = yes, 0 = no) 0.0672*** 0.0582** -0.0024 -0.0071 0.0833*** 0.0694***(0.0178) (0.0233) (0.0157) (0.0231) (0.0226) (0.0242)

Use of Online Brokerage (1 = yes, 0 = no) 0.0624** 0.0115 0.0173 0.1188*** 0.2126*** 0.0928***(0.0252) (0.0332) (0.0219) (0.0323) (0.031) (0.0331)

Follows Stock Market Closely (1 = yes, 0 = no) 0.1388*** 0.1092*** 0.0554** -0.0163 0.145*** 0.1018*** 0.1687***(0.0184) (0.0194) (0.0251) (0.0168) (0.0251) (0.0244) (0.0266)

Understands Stock Market Well (1 = yes, 0 = no) 0.0328 0.0228 0.0427 0.0215 -0.0049 0.0269 -0.0225(0.0233) (0.0234) (0.0306) (0.0204) (0.0306) (0.03) (0.0322)

Works for Pay [Proxy: Income] (1 = yes, 0 = no) -0.0095 -0.0076 -0.0165 -0.0121 -0.0204 0.0088 0.0304(0.0153) (0.015) (0.02) (0.0134) (0.0198) (0.0193) (0.0206)

Email [Proxy: Regular Internet Usage] (1 = yes, 0 = no) 0.024(0.0322)

Observations 2521 2484 2240 2119 2341 2196 2140

R2 0.0576 0.0717 0.0914 0.1915 0.0869 0.1282 0.097

F-Statistic 21.02 23.90 28.07 62.47 27.75 40.21 28.61

P-Value of F-Statistic 0.00 0.00 0.00 0.00 0.00 0.00 0.00Standard errors are in parentheses; *** p<0.01, ** p<0.05, * p<0.1. This table reports OLS estimates concerning the stock market participation decision of elderly households with internet access in the United States. Model (0) is a pre-baseline specification that indicates regular usage of the Internet. Model (1) is the baseline specification and is defined in Chapter V, while models (2) through (7) are robustness specifications described in section VI. Models (1) through (7) include goal-oriented Internet usage, which is specified as models that include the variables, Use of Internet for Financial Research and Use of Online Brokerage.

401K refers to an employer-provided retirement account; Trust fund refers to an account funded by somone other than the beneficiary; Mutual funds are a managed asset that contain a basket of investment products; Individual companies refers to an account that holds the shares one owns in individual public corporations (direct holdings); IRA/Keogh is a tax-deferred personal retirement account.

Regression Table

Table 5

Page 25: Determinants of Stock Market Participation Among Elderly U ...
Page 26: Determinants of Stock Market Participation Among Elderly U ...

23 | P a g e

VariableIntercept .232

(0.32)Years of Education 0.038***

(0.015)Age -0.008**

(0.003)Financial Wealth 0.075***

(0.011)Use of Internet for Financial Research 0.218***

(0.067)Use of Online Brokerage 0.387***

(0.129)How Closely One Follows Stock Market 0.331***

(0.071)How Well One Understands Stock Market 0.074

(0.084)Works for Pay (Income) -0.04

(0.061)Observations 2484Pseudo R2 0.0915

LR Chi2 222.07P-Value of LR Chi2 0.00Standard errors are in parentheses; *** p<0.01, ** p<0.05, * p<0.1. This table reports marginal effect estimates of the independent variables in model (1)

Marginal Effects of Baseline-Specification (Model 1)Table 6

Page 27: Determinants of Stock Market Participation Among Elderly U ...

24 | P a g e

Yes % No % Total %Yes 848 32.53% 1759 67.47% 2607 100%

% 47.03% 72.84% 61.81%

No 955 59.28% 656 40.72% 1611 100%

% 52.97% 27.16% 38.19%

Total 1803 42.70% 2415 57.30% 4218 100%

% 100% 100% 100%

Yes % No % Total %Yes 316 11.92% 2335 88.08% 2651 100%

% 90.03% 59.34% 61.85%

No 35 2.14% 1600 97.86% 1635 100%

% 9.97% 40.66% 38.15%

Total 351 8.19% 3935 91.81% 4286 100%

% 100% 100% 100%

Uses Internet for Financial Research

Uses an Online BrokerageHousehold Owns Stock

Household Owns Stock

Internet Specific Variables and Stockholding

Sample Characteristics

Table 7

Table 8

Household Owns Stock Freq. % Freq. % Freq. % Freq. % Freq. % Freq. % Freq. % Freq. %Yes 2669

61.71%1010

26.87%2661

60.03%1589

39.80%1253

32.09%2323

37.20%387

10.54%887

20.80%No 1656

38.29%2749

73.13%1772

39.97%2403

60.20%2652

67.91%1376

62.80%3285

89.46%3377

79.20%Total 4325

100.00%

3759100.00%

4433100.00%

3992100.00%

3905100.00%

3699100.00%

3672100.00%

4264100.00%

IRA/Keogh

*Direct holdings are those held outside of a managed asset. Managed Asset account types include mutual funds, trusts, managed investment accounts, annuities, and tax-deferred retirement accounts.

Account Definitions: Overall is composed of all accounts (direct and indirect), Individual refers to accounts that hold shares directly in individual corporations and thus is considered a directly held account, indirectly held account is composed of all other accounts, which are Mutual Fund, 401K, IRA/Keogh, Trust, and Other accounts; Other accounts are all stockholding accounts with the exception of personal brokerage accounts, mutual fund accounts, 401K accounts, IRA/Keogh accounts, and Trust accounts.

Stockholding by Account Type

401KOverall Indirectly Held OtherIndividual

(Directly Held) Mutual Fund Trust

Page 28: Determinants of Stock Market Participation Among Elderly U ...

25 | P a g e

Years of Education

No Yes Total No Yes Total0 1 3 4 0 2 2 41 0 1 1 1 0 1 12 0 0 0 2 0 0 03 2 0 2 3 2 0 24 0 2 2 4 2 0 25 0 0 0 5 0 0 06 3 1 4 6 4 0 47 4 1 5 7 5 0 58 7 6 13 8 11 2 139 25 12 37 9 38 2 40

10 49 19 68 10 66 2 6811 52 23 75 11 72 1 7312 620 553 1173 12 1141 46 118713 183 207 390 13 376 20 39614 246 337 583 14 551 40 59115 84 143 227 15 212 19 23116 264 497 761 16 686 95 78117 290 611 901 17 798 121 919

Total 1830 2416 4246 Total 3966 351 4317

Internet Specific Variables on Education

Years of Education

Uses Internet for Financial Research

Use of Online Brokerage

Household Owns Stock

30 40 50 60 70 80 90 Total %Yes 4

0.15%41

1.56%706

26.78%1053

39.95%628

23.82%194

7.36%10

0.38%2636

100.00%

36.36% 46.59% 64.47% 62.46% 57.99% 65.32% 66.67% 61.66%

No 70.43%

472.87%

38923.73%

63338.62%

45527.76%

1036.28%

50.31%

1639100.00%

63.64% 53.41% 35.53% 37.54% 42.01% 34.68% 33.33% 38.34%

Total 110.26%

882.06%

109525.61%

168639.44%

108325.33%

2976.95%

150.35%

4275100.00%

100% 100% 100% 100% 100% 100% 100% 100%

Age on StockholdingAge

Table 9

Table 10

Page 29: Determinants of Stock Market Participation Among Elderly U ...
Page 30: Determinants of Stock Market Participation Among Elderly U ...

26 | P a g e

Author Haliassos and Bertaut (1995) Bertaut (1998) Choi, Laibson, and

Metrick. (2000)Barber and Odean (2002)

Vissing-Jorgenson (2002)

Hong, Kubik, and Stein (2004)

Lusardi, Rooij, and Alessie (2007) Bogan (2008) Cole and Shastry

(2009)Glaser and Klos (2012)

Sample Period 1983 1983, 1989 1997-1999 Jan 1991 to Dec 1996 1968-1993 1992 2005 1992, 2002 1980, 1990, 2000 2001

Database Survey of Consumer Finances

Survey of Consumer Finances Hewitt Associates, LLC

Unidentified Large Discount Brokerage Firm

Panel Study of Income Dynamics, Consumer Expenditure Survey

Health and Retirement Survey DNB Household Survey Health and Retirement

Survey U.S. Census German SAVE data

Sample Coverage

U.S. Households U.S. Households Employees of two large companies U.S. Online Investors U.S. Households Older U.S. Households Dutch population

[Ages 22-90] Older U.S. Households U.S. Households German Households

Data observations

4103 1368Alpha (10,000+ participants), Omega (50,000 participants)

1607 1081 7,465 1,115 3774 4 to 14 million 1827

Definition of Stock

Stocks in Public Corporation, Mutual Funds, excluding stocks in company in which a household member was employed

Stocks in 401k

Stocks in Public Corporation, Mutual fund, Investment Trust, IRA

Stocks excluding IRA, Keogh account, 401K, Contribution Pension Plans

Financial Market Participation (income from interest, dividends, net rental income, royalty income, or income from estates and trusts)

Stocks in Public Corporation, Real Estate Mutual Funds (and similar assets)

Measurement of Stockholding

dummy (household owns stock)

dummy (household owns stocks) value (Trades)

dummy (household begins online trading in month t)

dummy (household owns stock)

dummy (household owns stock)

dummy (household owns stock)

dummy (household owns stock in 2002)

dummy (household has investment income)

dummy (household owns stocks)

value (Turnover)

Age Age

5 dummy variables, age<35, age=35-44, age=45-54, age=65-74, age>75, omitted variable is age=55-65

Age Age Age Age

5 dummy variables, age=30-40, age=41-50,age=51-60, age>60, omitted variable is age<30

Age19 dummy variables, 3-year age groups from 18-75 year olds

5 dummy variables (2nd, 3rd, 4th, 5th highest age quintile), omitted variable is lowest age quintile

Labor Income Labor Income Labor Income 1999 Salary 8 dummy variables, top range is >$125,000

Conditional Mean of Non-Financial Income Labor Income Household Income Household Income Household Income

4 dummy (2nd, 3rd, 4th, 5th income quintiles), omitted variable is lowest income quintile

Financial Wealth

Financial Net Worth Financial Net Worth Total plan balance at end of 1999

Financial Net Worth, Stock/Net Worth

Financial Wealth of household i in period t, nominal terms and real terms

4 dummy variables (2nd, 3rd, 4th, 5th wealth quintile), omitted variable is lowest wealth quintile

4 dummy variables (2nd, 3rd, 4th wealth quartile) omitted variable is lowest wealth quartile

Household Net Worth

4 dummy variables (2nd, 3rd, 4th, 5th highest wealth quintiles), omitted variable is lowest wealth income quintile

Sociability (Peer Effects)

dummy (sociability index, defined by questions concerning “Know neighbors”, “Visit neighbors”, “Attend Church” )

dummy (Education of peers: intermediate vocational/ second/ pre-university, Education of peers: higher vocational, university)

dummy (sociability)

Education 3 dummy variables, defined as less than high school, high school, some college, omitted variable is college degree

2 dummy variables (less than high school, college, omitted variable is at least high school)

3 dummy variables, 16+ years, 13-15 years, 9-12 years, omitted variable is less than 9 years

Years of education 4 dummy variables (intermediate vocational, secondary pre-university, higher vocational, university, omitted variable is less than intermediate

Years of education Years of education 4 dummy variables (general, intermediate, secondary degree, upper secondary omitted variable is less than general school )

Job Status

3 dummy variables (retired in 1983, retired between 1983-1989, remain retired in 1989)

dummy (retired) dummy (retired)

Internet Variable

Dummy (low tech, i.e. No computer usage)

dummy (1992 computer usage)

dummy (regular internet use in 2008)

Methodology (Regression Form)

Logit Bivariate Probit Binary Logit Logit MLE OLS OLS AND GMM Univariate Probit OLS OLS and GMM

Literature SurveyTable 11

Page 31: Determinants of Stock Market Participation Among Elderly U ...
Page 32: Determinants of Stock Market Participation Among Elderly U ...

27 | P a g e

Appendix:

Correlation Matrix:

Page 33: Determinants of Stock Market Participation Among Elderly U ...

28 | P a g e

Codebook:

ONLINE_BROKERAGE Level: RESPONDENT Type: Numeric How often do you do each of the following activities on the Internet? Buy or sell stocks, mutual funds, or bonds online ..................................................................... 3692 1. Never 332 2. Rarely 238 3. Sometimes 118 4. Often 53 9. QUESTION SKIPPED Recoded, (1=0, 2=0, 3=1, 4=1, 9=.) =============================== FIN_RESEARCH Level: RESPONDENT Type: Numeric How often do you do each of the following activities on the Internet? Get financial information online, such as stock quotes or mortgage interest rates ........................................................................ 1855 1. Never 898 2. Rarely 869 3. Sometimes 686 4. Often 125 9. QUESTION SKIPPED Recoded, (1=0, 2=0, 3=1, 4=1, 9=.) =============================== EDU Level: RESPONDENT Type: Numeric Years of Education ..................................................................... 4370 0-17. ACTUAL VALUE 63 Blank. Missing ===============================

AGE Level: RESPONDENT Type: Numeric Respondent’s Age ........................................................................ 4381 32-96. ACTUAL VALUE 52 Blank. Missing =============================== FOLLOWSTOCK Level: RESPONDENT Type: Numeric

How closely do you follow the stock market? ........................................................................ 449 1. Very Closely 2224 2. Somewhat 1685 3. Not at all 75 9. QUESTION SKIPPED Recoded, (1=1, 2=1, 3=0, 9=.) =============================== UNDERSTANDSTOCK Level: RESPONDENT Type: Numeric How would you rate your understanding of the stock market? ........................................................................ 77 1. Extremely Good 336 2. Very Good 1447 3. Somewhat Good 1109 4. Somewhat Poor 727 5. Very Poor 649 6. Extremely Poor 88 9. QUESTION SKIPPED Recoded, (1=1, 2=1, 3=1, 4=0, 5=0, 6=0, 9=.) ===============================

Page 34: Determinants of Stock Market Participation Among Elderly U ...

0 | P a g e