Behavioral Portfolio Analysis of Individual Investors 1

Electronic copy available at: http://ssrn.com/abstract=1629786

1

Behavioral Portfolio Analysis of Individual Investors 1

Arvid O. I. Hoffmann*

Maastricht University and Netspar

Hersh Shefrin

Santa Clara University

Joost M. E. Pennings

Maastricht University, Wageningen University, and University of Illinois at Urbana-Champaign

Abstract: Existing studies on individual investors’ decision-making often rely on observable socio-demographic

variables to proxy for underlying psychological processes that drive investment choices. Doing so implicitly ignores

the latent heterogeneity amongst investors in terms of their preferences and beliefs that form the underlying drivers

of their behavior. To gain a better understanding of the relations among individual investors’ decision-making, the

processes leading to these decisions, and investment performance, this paper analyzes how systematic differences in

investors’ investment objectives and strategies impact the portfolios they select and the returns they earn. Based on

recent findings from behavioral finance we develop hypotheses which are tested using a combination of transaction

and survey data involving a large sample of online brokerage clients. In line with our expectations, we find that

investors driven by objectives related to speculation have higher aspirations and turnover, take more risk, judge

themselves to be more advanced, and underperform relative to investors driven by the need to build a financial

buffer or save for retirement. Somewhat to our surprise, we find that investors who rely on fundamental analysis

have higher aspirations and turnover, take more risks, are more overconfident, and outperform investors who rely on

technical analysis. Our findings provide support for the behavioral approach to portfolio theory and shed new light

on the traditional approach to portfolio theory.

JEL Classification: G11, G24

Keywords: Behavioral Portfolio Theory, Investment Decisions, Investor Performance, Behavioral Finance

* Corresponding author: Arvid O. I. Hoffmann, Maastricht University, School of Business and Economics,

Department of Finance, P.O. Box 616, 6200 MD, The Netherlands. Tel.: +31 43 38 84 602. E-mail:

[email protected].

1 The authors thank Jeroen Derwall and Meir Statman for thoughtful comments and suggestions on previous

versions of this paper. Any remaining errors are our own.

Electronic copy available at: http://ssrn.com/abstract=1629786

2

I. Introduction

The combination of increased self-responsibility for retirement and an aging population has led a

growing number of people to become accountable for their own financial futures. Considering

the significant impact of current investment choices on future lifestyles (Browning and Crossley,

2001), it is important to understand how individual investors differ when it comes to the

triangular relationship among the decisions they make, the processes leading to these decisions,

and the resulting investment performance.

To date, our understanding of these relationships remains limited (Wilcox, 2003), as existing

research either studies only part of this triangle (Nagy and Obenberger, 1994) or uses observable

socio-demographic variables such as gender, age, or transaction channel to proxy for the

underlying psychological processes that drive investors’ decision-making (Graham, Harvey, and

Huang, 2009).2 In so doing, these studies implicitly assume that investors in the same age

bracket, having the same gender, or using the same transaction channel are homogenous in their

underlying psychological processes and the impact these have on their decision-making.

Recent literature on latent heterogeneity suggests that identifying the influence of

unobservable variables such as investors’ preferences and beliefs is key to achieving a better

understanding of financial market participants’ choices and behavior (Heckman, 2001; Pennings

and Garcia, 2009). Unobservable, individual-level differences may help to explain the underlying

mechanisms of a wide variety of behavioral anomalies (Dhar and Zhu, 2006; Graham et al.,

2 A well-known finding is women’s outperformance of men, on a risk-adjusted basis, due to the accumulation of

transaction costs by overconfident male investors who trade heavily (see e.g., Barber and Odean, 2000). Other

important results are that older investors have better diversified portfolios and trade less aggressively than their

younger counterparts (Dorn and Huberman, 2005; Goetzmann and Kumar, 2008), whereas investors switching from

phone-based to online trading are found to trade more actively, more speculatively, and less profitably than before

the switch (Odean and Barber, 2002).

3

2009; Lee, Park, Lee, and Wyer, 2008), but to date they have not been widely used to explain

individual investors’ decision-making or performance.

Our investigation into the role of individual differences focuses on the following questions:

How do investors differ from each other in respect to the type of information upon which they

rely to develop their strategies? How do investors differ from each other in respect to their

general investing objectives and risk attitudes? To what extent do differences among investors

impact the composition of their portfolios, trading activity, and investment performance?

To address these questions, we develop a dynamic behavioral theoretical framework and an

empirical study. The theoretical framework is a behavioral extension of the traditional Euler

equation approach and combines preferences, beliefs, and other variables that are typically

unobservable such as investors’ ambition level and risk attitude to explain how investors make

portfolio choices.3 As such, the framework reflects some of the essential features of behavioral

portfolio theory (BPT) (Shefrin and Statman, 2000) and findings from studies on overconfidence

(Barber and Odean, 2001; Kahneman and Lovallo, 1993). BPT emphasizes the role of behavioral

preferences in portfolio selection and proposes that individual investors’ portfolio choices and

consequently return performance reflect characteristics such as aspirations, hope, fear, and

narrow framing. In this respect, BPT helps to explain why some investors simultaneously buy

bonds and lottery tickets by investigating multiple objectives (e.g., protection from poverty at

retirement and potential for a shot at riches) as well as aspirations (Statman, 2002). Studies on

overconfidence emphasize the role of beliefs and help to explain why some investors are overly

3 In the remainder of this paper, we refer to “observable” variables when discussing variables that can be constructed

from secondary data, such as transaction records, and “unobservable” variables when discussing variables that as a

general matter cannot be observed using secondary data, but require primary data collection, such as our investor

survey. Thus, although technically the latter variables are not “unobservable” for our sample we continue to use this

terminology throughout this paper for reasons of consistency.

4

optimistic (Barber and Odean, 2001) and develop excessively bold forecasts (Kahneman and

Lovallo, 1993).

Our empirical study combines individual investors’ survey responses with their trading

records to create a unique dataset combining soft and hard data over an extended time period.

The survey allows us to directly measure investor characteristics that typically remain

unobservable, such as their objectives and strategies. Instead of using proxies based on, for

example, demographics, we directly measure these aspects of investors’ underlying preferences

and beliefs (Graham et al., 2009). Together with their trading records, this allows us to relate

investors’ decision-making processes with their observed choices instead of inferring the first

from the latter (cf. Manski, 2004). We empirically identify different segments of individual

investors, label and profile these segments, and compare their return performance.

In doing so, we contribute to the literature in several ways. We (1) characterize some of the

key ways in which individual investors differ from each other in terms of both preferences and

beliefs, (2) develop a stylized dynamic behavioral portfolio selection model to explain how

differences in preferences and beliefs lead to differences in investors’ portfolio decisions, (3)

develop a series of hypotheses based on predictions stemming from the model, and (4) present a

series of empirical findings, some of which serve to test our hypotheses, and some of which

strike us as surprising and at odds with conventional wisdom. Our most striking result is that

overtrading does not necessarily result in underperformance. Rather, underperformance depends

on the circumstances. Investors with strong beliefs that stem from using fundamental analysis

trade more frequently but still outperform investors using other strategies.

II. Traditional Portfolio Analysis: Stylized Model

5

A traditional model of dynamic portfolio choice (Merton, 1971; Viceira, 2001) involves an

expected utility maximizing investor choosing, at each time t, consumption (ct) and securities (xt

= xt,1 …, xt,J) given initial wealth (W), a stochastic stream of labor income (Lt), and stochastic

prices (qt). The standard Euler condition for this problem involves the purchase of a marginal

unit of security j at time t, and requires that the marginal benefit of this purchase be equal to the

marginal cost. The marginal benefit is the expected marginal utility of consumption at time t+1

generated by the marginal investment in security j. The marginal cost is the foregone marginal

utility of consumption at time t, as the increased expenditure on security j comes at the expense

of less consumption at time t.

In the traditional model, the expected utility function has the form E( t t (u(ct)), where is

a subjective time preference discount factor, and the expectation is taken over a subjective

probability belief (P) which an investor associates with the underlying stochastic process. The

Euler condition has the following form:

qt,j u/ ct = E(qt+1,j u/ ct+1) (1)

In words, purchasing an additional unit of security j at time t reduces consumption by qj,t units,

with each unit reduction of consumption resulting in the decline in utility at t of u/ ct. At t+1,

the additional unit of security j will result in the ability to purchase qj,t+1 units of consumption,

whose consumption increases discounted utility by u/ ct+1. At the optimum, the foregone

utility at t is exactly matched by the increased expected utility at t+1.

6

The traditional model postulates that investors’ subjective probability beliefs (P) are

objectively correct, and implicitly assumes that markets are efficient.4 In this setting, the purpose

of the portfolio is to manage the risk profile of the investor’s consumption stream, based on

initial wealth (W) and stochastic labor income (L). This means that the portfolio serves to hedge

uncertain labor income so as to smooth consumption over time. Unless labor income is highly

volatile, most trading activity would only involve marginal adjustments to a diversified portfolio

with the purpose of rebalancing or dealing with liquidity needs to finance consumption.

III. Behavioral Portfolio Analysis: Stylized Model

The behavioral approach to portfolio choice emphasizes additional motives for trading besides

rebalancing and consumption-related liquidity. These motives are connected to a series of

phenomena documented in the behavioral literature including:

Probability weighting and reference point effects involving gains and losses, psychophysics,

emotions, and aspirations (Kahneman and Tversky, 1979; Lopes, 1987)

Mental accounting (Thaler, 1985; Thaler, 2000)

Ambiguity aversion (Fox and Tversky, 1995; Heath and Tversky, 1991)

Status quo bias (Mitchell, Mottola, Utkus, and Yamaguchi, 2006).

The disposition effect (Shefrin and Statman, 1985)

The attention hypothesis (Barber and Odean, 2008)

Lack of diversification (Benartzi and Thaler, 2001; Goetzmann and Kumar, 2008)

4 Operationally, this means that prices q can be expressed in terms of a stochastic discount factor m according to q =

E(mx), where the expectation is formed with respect to (P).

7

Realization and evaluation utility (Barberis and Xiong, 2008)

Insufficient saving due to a lack of self-control (Shefrin and Thaler 1988, Benartzi and

Thaler 2007)

A. Behavioral Euler Equation

Consider a behavioral analogue to the traditional framework, which can capture the particular

phenomena just described. We begin with a full optimization extension to (1), which we

subsequently interpret in terms of a quasi-optimization analogue. Write the analogue of expected

utility as an objective function U. Let U have as its arguments consumption stream c = [ct],

portfolio choices x = [xt], changes in portfolio positions y = [xt – xt-1], prices q = [qt], and

probability beliefs (P). The arguments c, x, and y are random variables, with c and x being the

objects of choice. The inclusion of x, y, and q as arguments allows for an investor’s preferences

to reflect not only consumption, but also the performance of his or her portfolio and the impact

of gains and losses from trading.

In the neoclassical framework, investors with predictable streams of labor income make

small but frequent adjustments to their portfolios, by weighing the costs of foregone marginal

current consumption associated with a marginal security purchase against the expected marginal

future consumption so generated. Notice that the criterion driving portfolio choice is

consumption and savings.

In the corresponding behavioral framework, the consumption-savings feature is augmented

by additional considerations. When a behavioral investor contemplates a marginal increase in his

or her holdings xt,j of security j at time t, (s)he adds three additional components to the

neoclassical calculus. Those components take the form of U/ xt,j, U/ yt,j, and U/ yt+1,j. The

behavioral Euler condition is:

8

qt,j U/ ct = t+1 (qt+1,j U/ ct+1) + U/ xt + U/ yt - t+1 U/ yt+1 (2)

The term t+1 (qt+1,j U/ ct+1) is the analogue to E(qt+1,j u/ ct+1), with the summation t+1 over

the support of outcomes at t+1. The term U/ xt captures the effects of marginal evaluation

utility, meaning the psychological feelings the investor experiences from the value of his or her

portfolio at different points in time. Here an investor’s sense of wellbeing at a given moment,

apart from his or her consumption, is enhanced when his or her portfolio grows, and is

diminished when it falls. The terms U/ yt and U/ yt+1 capture realization utility (Barberis and

Xiong, 2008), meaning the impact of trading a position. In this respect, an investor who sells at a

gain might experience positive realization utility whereas an investor who sells at a loss might

experience negative realization utility (cf. Thaler and Johnson, 1990).5 The minus sign associated

with the term t+1 U/ yt+1 in (2) reflects the fact that increasing xt reduces yt+1 = xt+1 – xt.

B. Preferences: Evaluation Utility and Realization Utility

In the stylized thought process that underlies condition (2), the investor’s decision regarding a

security at time t balances the benefits from trading and holding a security against the associated

costs. As in the neoclassical condition (1), increases in future consumption appear as benefits and

decreases in current consumption appear as costs. As for the psychological benefits that appear

on the right-hand-side of (2), consider the determinants of evaluation utility and realization

utility.

5 Parenthetically, transaction costs can be captured by augmenting the model to feature both bid and ask prices. In

the current framework, there is a single transaction price and so bid-ask spreads are zero.

9

Evaluation utility reflects the emotional experience associated with holding a position in a

security. Among the determinants of evaluation utility are three variables described in Shefrin

and Statman (2000) and Shefrin (2008), namely SP, , and P(A).

The variable SP, known as a security-potential function, is similar to expected utility. It is

associated with gains and losses to the value of a position, and relates to feelings associated with

thrill-seeking, presence or absence of anxiety, and value-expressive benefits derived from, for

example, holding stocks of socially responsible firms (Statman, 2004).

In contrast to an expected utility function, SP features rank-dependent weights in place of

probabilities. Rank-dependent weights reflect particular emotions, such as fear and hope.6

Investors who are overly fearful act as if they overweight unfavorable events relative to more

favorable events. Notably, although the probability weight attached to an event does not vary

with portfolio decisions, the decision weight assigned to an outcome can vary with the position

an investor takes in a security. In particular, when holding a long position in security j, a fearful

investor will tend to overweight the probability that the return is negative. However, should the

same investor instead hold a short position in security j, s(he) would overweight the probability

that the return is positive.

The variable refers to an aspiration level. For example, the investor’s aspiration might be

that the portfolio s(he) selected at t-1 be worth at least t at time t. Correspondingly, P(At) is the

probability the investor assigns to meeting that goal. Investors who set both high aspirations and

high probabilities of achieving those aspirations are said to be ambitious. A key feature of the

portfolio selection framework developed by Shefrin and Statman (2000) is that ambitious

investors take on high risk.

6 Psychological-based decision theories tend to use an inverse-S shaped weighting function for distribution functions

(Kahneman and Tversky, 1979; Tversky and Kahneman, 1992). This corresponds to a U-shaped weighting function

for density functions in which probabilities of extreme events are exaggerated. .

10

The realization utility terms on the right-hand-side of (2) embody the impacts of pride or

regret directly associated with the act of trading. Examples are the feeling of pride associated

with selling at a gain or the feeling of regret associated with selling at a loss.

C. Beliefs: Biases, Framing and Probability Weighting

The behavioral approach emphasizes the importance of both preferences and beliefs. The

discussion in the previous paragraphs has emphasized preferences. In shifting the emphasis from

preferences to beliefs, we identify three key issues. First, investors typically have erroneous

beliefs stemming from behavioral biases. Examples of biases are excessive optimism and

overconfidence. Excessive optimism can lead investors to overestimate expected returns (De

Bondt and Thaler, 1985), whereas overconfidence can lead them to underestimate risk (Barber

and Odean, 2000; Odean, 1998). In conjunction, this can lead to forecasts which are too bold (cf.

Kahneman and Lovallo, 1993). Moreover, most individual investors have only the vaguest

notion of how security returns are jointly distributed (Benartzi and Thaler, 2001).

Second, because of framing effects, behavioral investors ignore information relating to

return covariance. This is the key reason underlying the violation of stochastic dominance in

prospect theoretic choice experiments (Kahneman and Tversky, 1979). Therefore, the beliefs

used in connection with (2) across securities might not be compatible with a single set of beliefs

P.7 Instead, we follow the approach of prospect theory with narrow framing and assume that

investors’ beliefs consist of marginal distributions for each security, which are applied to (2) on a

7 In the behavioral model, investors seek to achieve equality (2) for each security, thereby balancing the marginal

benefits and marginal costs of increasing the amount held of each security j. However, computing the values of

U/ ct, U/ ct+1, U/ xt,j, U/ yt,j, and U/ yt+1,j is a tall order requiring full knowledge of the joint distribution of

all security returns. For this reason, we postulate that investors use a heuristic approach to estimate marginal benefits

and costs, and as a result select suboptimal portfolio strategies.

11

security by security basis.8 In particular, investors are assumed to ignore covariances when

choosing their portfolios.

Third, preferences and beliefs interact through probability weighting. In this regard, decision

weights are applied to subjective probabilities. However, both preferences and beliefs also

combine to impact at least two other psychological phenomena, aversion to ambiguity and status

quo bias, a topic to which we now turn.

D. Ambiguity Aversion and Status Quo Bias

Ambiguity aversion reflects discomfort stemming from a lack of knowledge of the underlying

probabilities (Ellsberg, 1961). For example, knowing that an urn contains 100 balls, of which 50

are red and 50 are black is different from knowing that an urn contains 100 balls whose color is

either black or red, but with no knowledge of the fraction of each. Status quo bias involves the

tendency to preserve the status quo instead of to make a change from the status quo.

Both aversion to ambiguity and status quo bias play key roles in the portfolio issues we

analyze. Aversion to ambiguity can lead investors to hold relatively few securities, leading for

example to x=0 for most securities. Status quo bias involves an underlying reluctance to trade

(Samuelson and Zeckhauser, 1988), leading for example to y=0 holding much of the time.9

8 Prospect theory takes as its starting point expected utility theory and replaces the utility function with a value

function, probabilities with probability weights, and a single complex optimization with a collection of simpler

optimizations. Notably, the value function and probability weighting function are consistent across the simpler

optimizations. 9 An example that is often used to explain the relation between status quo bias and regret avoidance is the following.

Suppose you own stock worth $1,000 in Company A and can exchange it for $1,000 of stock in Company B. Given

your investment assessment, you choose to hold your current shares. Your neighbor holds $1,000 in Company B

and, for reasons similar to yours, decides to switch his shares for $1,000 of Company A. During the next six months,

the value of each person’s stock in Company A falls to $700. Which one feels the greater regret? According to the

psychology literature, the answer is that your neighbor will experience more regret because he took an action that

produced an unfavorable consequence, and can easily imagine having done otherwise. In contrast, you took no

action, and so it is more difficult for you to imagine acting differently.

12

Notably, solutions featuring zero holdings (x=0) or zero trading (y=0) are instances of corner

solutions. Shefrin (2008) points out that behavioral preference maps involving rank-dependent

weighting feature many kinks, and these in turn give rise to corner solutions where marginal

conditions like (2) fails to hold with equality. Aversion to ambiguity is manifest in a strong fear

response, the fear that a highly unfavorable event will occur if the investor takes a decision with

limited knowledge of the underlying probabilities. Such fear can lead an investor to view any

non-zero position in a particular security as unattractive, be that position long or short. Formally,

this would entail U/ xt < 0 for xt > 0 and U/ xt > 0 for xt < 0 with a point of non-

differentiability (kink) at xt = 0. Similar remarks apply to status quo bias (in respect to U/ yt).

Status quo bias does not mean that investors refrain from trading altogether, only that other

forces must be strong enough to counter the bias. Aversion to ambiguity and status quo bias

induce investors to hold a few securities rather than many, and to trade intermittently rather than

continuously. Certainly, if at some point in time, the needs for hedging, rebalancing, and

liquidity are sufficiently strong, investors will overcome status quo bias and trade. Likewise,

investors can overcome status quo bias if they have enough confidence in their stock picking

skills to feel little potential for regret (Kahneman, Knetsch, and Thaler, 1991), derive sufficiently

high evaluation utility from their portfolios, or have bold enough forecasts (cf. Kahneman and

Lovallo, 1993).

Bold forecasts stem from conviction, a combination of familiarity, strong opinions, and

confidence, just the opposite of ambiguity. Consider some of the findings in the psychology

literature about the influence of information on decision makers’ degree of conviction. An often

cited study of horse race handicappers by Slovic and Corrigan (1973) analyzes how confidence

and accuracy change as functions of the amount of racing sheet information. Accuracy increases

13

with the amount of information, until a point of information overload is reached, after which it

slightly declines (Oskamp, 1965). However, confidence increases steadily with the amount of

information (Locander and Hermann, 1979). Hahn et al. (1992) confirmed that decision quality is

an outcome of both time pressure and information load.

The findings in Heath and Tversky (1991) on the determinants of ambiguity aversion

provide further insight into the drivers of conviction. They show that ambiguity aversion is

reduced by a sense of familiarity and expertise. Fox and Tversky (1995) establish that the degree

of ambiguity aversion in a particular choice increases when decision makers contrast the choice

with a situation in which he or she has more knowledge, or someone else has more knowledge.

E. Setting the Stage for Hypotheses Development

To set the stage for the development of our hypotheses, we recapitulate some of the interpretive

features in the stylized behavioral Euler approach embodied within condition (2). In respect to

preferences, consider (2) to be a dynamic extension of the mental accounting version of BPT in

Shefrin and Statman (2000). Here securities are evaluated relative to goals defined by aspiration

levels and success probabilities, with each mental account and associated aspiration level

corresponding to a different goal (cf. Das et al., 2010). Examples of the types of goals we

consider in the remainder of the paper relate to capital growth, retirement saving, hobby, and

speculation (cf. Lewellen, Lease, and Schlarbaum, 1980). In this paper, we refer to these types

of goals as “objectives.”

As in BPT, condition (2) pertains to two points in time. However, unlike BPT, (2) contains

terms pertaining to realization utility, which impacts trading behavior. In addition, the relative

strength of the different terms in (2) is assumed to reflect the general nature of different types of

14

goals. For example, we think it reasonable to assume that realization utility is stronger for

investors whose primary objective is thrill seeking than investors whose primary objective in

investing involves savings for retirement.

In respect to beliefs, we interpret the model as if investors are quasi-rational, relying on

subjective marginal distributions rather than on joint return distributions. These distributions are

all implicitly conditional. To this end, our paper focuses on the source of the conditioning.

Examples include the media (financial news), past prior prices (technical analysis), and financial

variables (fundamental analysis) (cf. Lease, Lewellen, and Schlarbaum, 1974). In this paper, we

refer to these types of information sources as “strategies.”

As a system, (2) is more akin to a consumer choice model than a mean-variance portfolio

model. In this respect, securities are selected and held for their attributes, and their contribution

to satisfying needs (cf. Wilcox, 2003). Just as each consumer purchases only a small subset of

available products, so do behavioral investors hold only a small subset of available securities, at

least directly. The determinants of which securities are held at any time reflect the interaction

among ambiguity aversion, status quo bias, and boldness of beliefs, as in the “bold forecasts,

timid choices” framework of Kahneman and Lovallo (1993). In addition, the quasi-rational

feature of (2) might involve investors holding different types of securities, not because they value

diversification, but because they have a taste for variety. Although variety might mimic

diversification, investors ignore covariance information in (2), and so do not value diversification

per se.

In the next section we develop a series of hypotheses, based on the behavioral Euler

condition (2) and some related assumptions. Our first major assumption is that (2) features the

“bold forecasts, timid choices” property, in which forecasts need to be sufficiently bold to

15

overcome status quo bias. Our second major assumption is that investors implement (2) in a

manner similar to making consumer choices, meaning that they do not value diversification per

se, although they might have a taste for variety in their security holdings. Our additional

assumptions pertain to the manner in which evaluation utility, realization utility, and aspiration

variables vary across investor objectives, and confidence and accuracy vary across investor

strategy. We develop these additional assumptions in the next section, where we use condition

(2) to describe how variation in boldness across investor strategies, and variation in aspirations

across investor objectives predicts variation in trading patterns and associated returns.

IV. Hypotheses

Overconfidence pertains to beliefs, and status quo bias pertains to preferences. The behavioral

approach emphasizes the psychological features associated with both preferences and beliefs. In

this paper, we focus on the role of investment objectives as a reflection of investor preferences,

and the role of investment strategy as a reflection of investor beliefs. In this section, we develop

hypotheses about the impact of both strategies and objectives.

Our hypotheses relate to individual differences across the spectrum of investors. In this

regard, overconfidence leads some investors to trade too much, while status quo bias leads other

investors to trade too little (Goetzmann and Kumar, 2008; Rantapuska, 2006). Overconfidence

leads investors’ forecasts to be excessively bold, while status quo bias leads to timid choices and

inaction (Kahneman and Lovallo, 1993). As discussed below, in our framework, both features

can operate simultaneously with the result that investors trade only intermittently, when their

beliefs are sufficiently bold to outweigh status quo bias.

16

Status quo bias is strong. Mitchell et al. (2006) provide evidence that 80% of participants in

401(k) accounts initiate no trades in a two-year period, and an additional 11% make only one

trade. Therefore, few investors in their sample rebalance. Similarly, Ameriks and Zeldes (2004)

find that 50% of the investors in their sample do not rebalance over a nine year period. In a

related vein, Choi et al. (2008) find that 80% of investors in 401(k) plans maintain the plan’s

default savings contribution and investment option.

Investors who trade rarely if ever lie at one end of the spectrum. At the other end of the

spectrum lie investors who trade on a daily basis. Barber et al. (2009) report that 17% of traders

in Taiwan are day traders. For most day traders, overconfidence is strong. In the main, our

hypotheses deal with investors lying in the middle of the spectrum, where status quo bias and

overconfidence operate in tension.

Active trading stems from conviction. An overconfident investor with sufficiently high

conviction in his or her stock picking skills will tend to overcome status quo bias and engage in

frequent trading (cf. Kahneman and Lovallo, 1993). In respect to beliefs, our hypotheses for

active traders focus on the nature of the information upon which investors rely, and the degree to

which that information generates conviction. We suggest that variation in investors’ trading

activity will be influenced by the nature of the information upon which their trading strategies

depend. If investors possess information that generates high conviction, the resulting

overconfidence leads to bolder forecasts. Bolder forecasts are able to overcome investors’ status

quo bias that would otherwise cause timid choices and inaction. This relationship is a key feature

of the hypotheses we develop below, especially in respect to trading strategies.

A. Investment Strategies

17

First, compare investors who rely on fundamental analysis as a strategy with those who rely on

technical analysis. Investors using fundamental analysis examine all underlying conditions

relevant for future stock price developments. Besides financial statements, these include

economic, demographic, and geopolitical factors. In contrast, investors relying on technical

analysis only study the stock price movements themselves, believing that historical data provides

indicators for future stock price developments.

To us, this suggests that fundamental analysis typically involves more information than

technical analysis (cf. Shleifer and Summers, 1990). Investors relying on fundamental analysis

are therefore more likely to become more familiar with the firms they follow than investors

relying on technical analysis. After all, fundamental analysis serves to focus primary attention on

details pertaining to the firms themselves, whereas technical analysis focuses attention on price

patterns generated by firms’ stocks. This focus on firm fundamentals instead of the kind of

pattern recognition tasks inherent in technical analysis leads us to conclude that familiarity bias

will tend to be stronger by investors relying on fundamental analysis than investors relying on

technical analysis.

In the language of Kahneman and Lovallo (1993), investors who rely on fundamental

analysis are more inclined to adapt an “inside view” (Kahneman and Lovallo, 1993) and become

overconfident than those who rely on technical analysis, as confidence is an increasing function

of the amount of information (Locander and Hermann, 1979). We hypothesize that as a result

their forecasts become bolder and they more easily overcome status quo bias, leading to less

timid choices. Thus, based on condition (2) we expect investors who rely on fundamental

analysis to trade more frequently than those who rely on technical analysis, ceteris paribus. In

short:

18

H1: Relative to investors relying on technical analysis, investors relying on fundamental analysis

will form bolder beliefs, and their greater overconfidence will induce them to trade more

frequently.

Apart from a very small segment of highly skilled investors who hold concentrated portfolios

(Barber, Lee, Liu, and Odean, 2009; Goetzmann and Kumar, 2008), overtrading typically leads

to underperformance due to the accumulation of transaction costs (Barber and Odean, 2000). As

there is no a priori reason to expect that investors using fundamental analysis are more skilled

than investors using other strategies, we expect:

H2: Relative to investors relying on technical analysis, investors relying on fundamental analysis

will earn lower risk and style adjusted returns.

Second, compare investors who rely on fundamental analysis with those relying on their

intuition. In the behavioral framework, investors do not place high intrinsic value on

diversification. In the spirit of prospect theory’s isolation effect (mental accounting, narrow

framing) (Kahneman and Tversky, 1979), investors act as if they implement condition (2) on a

security-by-security basis, rather than as part of an integrated optimization.10

As a result, status

10

Prospect theory is a boundedly rational theory of choice involving maximization of a weighted value function.

The maximization does not typically correspond to a full optimization, as complex decision tasks are often

simplified into smaller subtasks with important information about the subtasks being omitted. In this respect, the

value function used to make decisions corresponds to a “proxy” of the decision maker’s utility function. Decision

makers rely on proxies because they lack the ability required to compute utility. The use of proxies featuring

omissions can result in suboptimal choice, of which a notable example is the selection of stochastically dominated

risks. A key feature of prospect theory is that the value function and weighting function are common across decision

tasks. In the present analysis, think of equation (2) featuring a proxy for the expected utility terms, in which the

omitted information involves the contribution to utility from securities other than j. If we follow the prospect theory

assumption, then the proxy will be the same across securities.

19

quo bias will typically lead to underdiversification. Ceteris paribus, (2) implies that investors

holding more securities will tend to be those with stronger convictions in their stock picking

skills and in possession of better and more information which leads them to make bolder

forecasts (Kahneman and Lovallo, 1993). Only in these cases, will investors be able to overcome

status quo bias and be willing to invest in multiple stocks and thus make less timid choices.

As discussed previously, it is likely that the latter features correlate with reliance on

fundamental analysis. As such, investors who rely on fundamental analysis will tend to hold a

larger number of different stocks in their portfolios than other investors.11

Conversely, investors

who only rely on intuition, and therefore less information, will tend to have less conviction

regarding their stock picking skills for most securities and their status quo bias leads them to

make timid choices. As a result, these investors may be biased towards a small(er) number of

stocks with which they are familiar (Huberman, 2001). Goetzmann and Kumar (2008) point out

that as investors increase the number of stocks in their portfolios, they tend to choose stocks

which co-move, thereby depriving themselves of the benefits of diversification. Moreover, to

avoid feelings of regret (Kahneman et al., 1991) investors relying on intuition will exhibit a

strong status quo bias and hold fewer securities in their portfolios. In short:

H3: Investors relying on fundamental analysis will hold a larger number of different stocks in

their portfolio than investors relying on intuition.

B. Investment Objectives

11

The common proxy assumption implies that at the origin, the decision to trade is determined by whether or not the

net benefit is sufficient to overcome the obstacle imposed by the kink, with the latter being common across

securities.

20

Investment objectives are imbedded in investors’ preferences. Aspiration levels constitute an

important component of objectives. A key implication of behavioral portfolio theory is that

investors whose goals involve high aspirations act as if they have a high tolerance for risk,

implying that investors who set high aspiration levels in combination with an associated high

probability of achieving those levels, will tend to choose risky portfolios (Shefrin and Statman,

2000). Risky portfolios are portfolios that are more exposed to market risk and overweight small

firms (Barber and Odean, 2001). Hence we hypothesize:

H4: Investors with higher aspiration levels have higher risk profiles than investors with lower

aspiration levels.

H5: Investors with higher risk profiles will hold riskier portfolios (i.e. with higher exposure to

the market and small-firm factors) than investors with lower risk profiles.

As previously discussed, because of familiarity bias, investors who rely on fundamental analysis

are likely to have high conviction in their stock picking skills. In addition to leading them to

make bold forecasts, we suggest that familiarity also leads them to be more ambitious than

investor whose beliefs feature more ambiguity. This is because ambiguity involves uncertainty

about P(A), the probability of achieving the aspiration level. In turn, ambiguity aversion induces

pessimism about P(A), which results in less risk taking. This leads to the following hypothesis:

H6: Investors relying on fundamental analysis will have the highest aspirations and risk profiles.

21

The behavioral framework links investments objectives to trading behavior. In this regard,

investors saving for retirement or building a financial buffer and investors who invest to

speculate or exercise a hobby lie at opposite ends of a continuum. For investors who mainly

invest as a hobby or to speculate, the second and third term in (2) loom large, as these relate to

the benefits from (anticipated) evaluation utility (Barberis and Xiong, 2008) and thrill seeking

(Grinblatt and Keloharju, 2006). To experience these positive emotions such investors will trade

more frequently than other investors. Hence we hypothesize:

H7: Investors who invest primarily as a hobby or to speculate will trade more frequently than

investors whose primary investment objective is to build a financial buffer or save for retirement.

Additionally, investors who mainly invest as a hobby or to speculate might have very high

conviction, make bold forecasts, tolerate risk, and set ambitious targets. Indeed, recent literature

shows that investors who trade to entertain themselves (Dorn and Sengmueller, 2009) or to

speculate – essentially seeing stocks as a lottery ticket providing a shot at riches (Statman, 2002)

– have higher aspirations and take more risk relative to investors who do not associate investing

with gambling (Kumar, 2009). In contrast, investors whose primary investment objective is to

build a financial buffer or save for retirement are likely to have lower aspirations and choose

more conservative portfolios. In short:

H8: Investors whose primary investment objective is to build a financial buffer or save for

retirement have lower aspirations and take less risk than investors who invest primarily as a

hobby or to speculate.

22

V. Data and Methods

The analyses in this paper draw on transaction records of all clients and questionnaire data

obtained for a sample of clients of the largest online broker in The Netherlands. Due to trading

restrictions, we exclude accounts owned by minors (age <18 years). We also exclude accounts

with a beginning-of-the-month value of less than €250 and accounts owned by professional

traders to ensure we deal with active accounts owned by individual investors. Imposing these

restrictions leaves 65,325 individual accounts with over 9 million trades from January 2000 until

March 2006.

A. Brokerage Records

Opening positions as well as complete transaction records are available for all prospective

participants of the survey, regardless whether they choose to participate or not, allowing us to

control for sample selection bias. The typical record consists of an identification number, an

account number, transaction time and date, buy/sell indicator, type of asset traded, gross

transaction value, and transaction commissions.

B. Survey Sampling and Selection

In 2006, we designed and performed an online survey amongst all clients of the online broker. In

total, 6,565 clients completed the questionnaire. To prevent biased responses, the purpose of the

survey was framed in a neutral way and no reference to the objective of the study at hand was

made. In the call to participate, respondents were requested to “provide their opinion of the

online broker”. Brokerage clients who participated in the survey could win a personal computer

23

that was raffled amongst respondents who fully completed the questionnaire. Amongst other

questions, the questionnaire probed for investors’ preferences as reflected in their investment

objective, beliefs as reflected in their investment strategies, aspiration level as reflected by their

ambition level, risk-taking behavior as reflected in the risk profile of their current investment

portfolio, and their sophistication as reflected in their self-categorization into novice, advanced,

or very advanced investor classes. Figure 1 provides an overview of the questions we used.

After matching transaction records with questionnaire data, a sample of 5,500 clients and

corresponding accounts remain for which both hard (transaction) and soft (survey) data are

available and which have an account history of at least 36 months.

[Figure 1 about here]

C. Descriptive Statistics

In Table 1 (Panel A) we report descriptive statistics for the respondents to the investor survey.

We also report these descriptives for the non-respondents to test for selection bias (Panel B).

Of the sample of 5,500 investors for which both accounting and survey data is available,

58% is male and the mean age is about 50 years. The mean (median) number of total trades over

the sample period is 76.45 (30.00). Average (median) monthly turnover is about 42% (11%). The

average (median) portfolio value is €45,915 (€15,234). Combining the average portfolio value

with a total portfolio value of €50,000-€60,000 for the average Dutch investor (Bauer,

Cosemans, and Eichholtz, 2009) indicates that our average client invests more than three-fourth

of his or her total investment portfolio at this particular online broker, showing that we do not

investigate investors’ “play accounts” (Goetzmann and Kumar, 2008) but deal with

24

representative and serious investor accounts. In fact, 40.8% of our survey respondents only hold

an investment account at this particular broker. Of the respondents who do also hold an

investment account at another broker, 51.6% indicate that this comprises less than half of their

total portfolio. As a robustness check, we compare the results of investors who only invest

through this particular broker with those who also have another brokerage account but find no

significant differences regarding our hypotheses. Following Seru, Shumway, and Stoffman

(2008) we measure experience by the number of months an investor has been trading since

account opening. The results in Table 1 show that the mean (median) experience is about 40.21

(39.00) months. As compared to recent findings by Odean and Barber (2000) and Goetzmann

and Kumar (2008) our investors’ portfolios are better diversified, although still far from well-

diversified. The mean (median) number of stocks held by our investors is 6.57 (4.00) while the

mean (median) Herfindahl-Hirschmann Index (HHI) is 27.78% (21.14%). Comparing the HHI

with the normalized HHI (HHI*) indicates that investors’ portfolio weights are not uniformly

distributed. Rather, investors distribute their overall portfolio value unevenly over different

assets. Mean (median) monthly returns over the period 2000-2006 are -0.30% (0.30%). On

average, the respondents to the investor survey are relatively risk-seeking, with a mean (median)

score of 5.31 (6.00) (1=very defensive, 7=very speculative).

A comparison between the respondents to the investor survey (Panel A) with non-

respondents (Panel B) shows that relative to non-respondents, the respondents feature more

females, are older, transact more frequently, have higher portfolio values, are more experienced,

better diversified (hold a larger number of different stocks and have a lower HHI), obtain a better

monthly return performance, and take more risk (all p<0.00). Although the differences between

survey respondents and non-respondents are relatively small, they suggest that the sample of

25

clients which completed the investor survey tend to be relatively sophisticated investors with a

sizeable portfolio which adds to the relevance and importance of the study at hand.

[Table 1 about here]

D. Measuring Investor Performance

Investor performance is defined as the monthly change in market value of all securities in an

investor’s account (Bauer et al., 2009). End-of-the-month account value is net of transaction

costs the investor incurred during the month. As performance is measured on a monthly basis,

assumptions have to be made considering the timing of deposits and withdrawals of cash and

securities. To be conservative, we assume that deposits are made at the start of each month and

withdrawals take place at the end of each month. Analyses with the assumption that deposits and

withdrawals are made halfway during the month yield similar results. Hence, we calculate net

performance as

)(

)(

1

1

itit

itititnet

itDV

NDWVVR , (3)

where Vit is the account value at the end of month t, NDWit is the net of deposits and withdrawals

during month t, and Dit are the deposits made during month t.

Gross performance is obtained by adding back transaction costs incurred during month t,

TCit, to end-of-the-month account value,

)(

)(

1

1

itit

ititititgross

itDV

TCNDWVVR . (4)

Only direct transaction costs (commissions) are considered. We do not add back any indirect

transactions costs (market impact and bid-ask spreads). The trades of most individual investors

are relatively small, making market impact costs unlikely. Moreover, Keim and Madhavan

26

(1998) show that bid-ask spreads may be imprecise estimates of the true spread, as trades are

often executed within the quoted spread.

E. Attributing Investor Performance

To obtain investors’ abnormal performance, we attribute the returns on investor portfolios to

different risk and style factors using the Carhart (1997) four-factor model. This model adjusts

investor returns for exposure to market (RMRF), size (SMB), book-to-market (HML), and

momentum (UMD) factors. Following Bauer et al. (2009), we construct these factors for the

Dutch market, as our sample of investors mainly invests in Dutch securities.12

The market return

in the RMRF factor is represented by the return on the MSCI Netherlands equity index. All

factor-mimicking portfolios are constructed according to the procedure by Kenneth French.13

The following time series model is estimated to obtain risk and style adjusted returns:

K

k

itktikiit FR1

. (5)

In this model Rit represents the excess return on investor i’s portfolio, βik is the loading of

portfolio i on factor k, and Fkt is the month t excess return on the k’th factor-mimicking portfolio.

The intercept αi measures abnormal performance relative to the risk and style factors. The factor

loadings indicate whether a portfolio is tilted towards market risk or a particular investment

style.

F. Segmenting Investors

12

In terms of volume (value) 95% (85%) of all trades are transactions in Dutch securities. This suggests the

presence of a home bias among Dutch investors, which has previously been documented by French and Poterba

(1991) for the US, UK, and Japan and by Karlsson and Norden (2007) for Sweden. Hence, we find that Dutch

versions of the factor-mimicking portfolios lead to a better model fit than do international factors. 13

See http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html.

27

We group the 5,500 investors for which we obtained both hard transaction and soft survey data

into groups based on their preferences or beliefs. Investment objectives pertain to preferences,

whereas strategies pertain to beliefs.

While the investors in our sample typically have only one investment objective (e.g., saving

for retirement, building a financial buffer, speculate, exercise a hobby), they combine different

strategies to attain this objective (e.g., combining financial news with intuition or professional

advice).14

Therefore, we use univariate sorting to distinguish different segments of investors

based on investment objective (cf. Kumar, Page, and Spalt, 2009), but have to use cluster

analysis to discern segments based on investment strategy (Hair, Anderson, Tatham, and Black,

1998).

The univariate sorting results indicate five segments of investors based on their dominant

investment objective. These segments are labeled Capital Growth, Hobby, Saving for

Retirement, Speculation, and Building Financial Buffer.

To distinguish segments of investors based on investment strategy, we use a non-hierarchical

cluster analysis following Hair, Anderson, Tatham and Black (1998).15

The cluster analysis

groups together investors with similar scores on certain (combinations of) strategies. In

particular, differences between segments in terms of scoring are maximized and within segments

minimized (Punj and Stewart, 1983). This procedure leads to six segments, which are labeled

14

In its original specification, behavioral portfolio theory (Shefrin and Statman, 2000) is a static framework that

describes an investors’ portfolio as consisting of multiple layers, each layer corresponding to a particular investment

goal or objective. In this paper, we incorporate elements of behavioral portfolio theory into a dynamic framework to

develop hypotheses about investors’ trading behavior on an ongoing basis, focusing on investors’ most important

investment objective. 15

Nonhierarchical clustering procedures are less susceptible to outliers in comparison to hierarchical clustering

procedures. In addition, unlike hierarchical clustering, K-means clustering as used in this study is able to analyze

large data sets as this procedure does not require prior computation of a proximity matrix of the distance/similarity

of every case with every other case (Punj and Stewart, 1983).

28

Financial News, Financial News and Intuition, Intuition, Technical Analysis, Fundamental

Analysis, and Financial News, Intuition, and Professional Advice.

Table 2 reports descriptive statistics for these segments in regard to a number of observable

variables, while Table 3 does the same for the unobservable variables. Observable variables are

variables which can be constructed from the secondary data (transaction records) as obtained

from the online brokerage firm. Unobservable variables cannot be constructed using secondary

data, but require primary data as obtained by our investor survey.

[Tables 2 and 3 about here]

VI. Profiling Investor Segments

This section profiles the different segments of investors as obtained previously using a

combination of observable and unobservable variables.

A. Segments based on Investment Objectives

Table 2 shows that male investors are especially well represented in the segments Hobby (0.62)

and Speculation (0.64). The latter segments also contain the youngest (47.31 and 48.61 years,

respectively) investors, whereas those in the segment Speculation also trade most heavily during

the sample period (99.25 times). Monthly turnover is highest in the segment Speculation

(78.87%) and lowest in the segment Saving for Retirement (26.44%). Investors in the segment

Capital Growth have the largest portfolio value (€62,646) while Hobby investors have the

smallest accounts in terms of value (€24,139). Investors Saving for Retirement are most

29

experienced (43.49 months) and best diversified both in terms of number of stocks and the HHI

(holding 7.57 different stocks and having a HHI of 24.39%) while investors in the segment

Speculation are least experienced (34.47 months) and less diversified (5.63 different stocks, HHI

is 30.38%).16

The profiles of the segments Speculation and Hobby thus obtained, containing

younger male investors who overtrade and underdiversify, are in line with recent findings on

speculative trading as gambling (Kumar, 2009) or entertainment (Dorn and Sengmueller, 2009).

Table 3 demonstrates that the segment Speculation has the highest score on ambition level

(3.52), the most speculative risk profile (5.80), reports to have the lowest percentage of novice

investors (24.13%), and the highest percentage of advanced (59.74%) and very advanced

(15.48%) investors, respectively. Together with the high turnover and dominance of males in this

segment, these findings confirm and enrich earlier work that finds that especially male investors

are subject to overconfidence and trade excessively (Barber and Odean, 2001). Additionally,

these findings confirm the prediction by Statman (2002) that investors who perceive investing as

playing the lottery may have particularly high aspiration levels and be subject to overconfidence.

Not surprisingly, we find that investors in the segment Saving for Retirement have lower

ambition levels (3.26), a less speculative risk profile (4.98) and are more modest about their level

of sophistication (only 7.64% of this group of investors judge themselves to be very advanced).

B. Segments based on Investment Strategy

Table 2 shows that the fraction of males is highest (0.64) in the segment Fundamental Analysis

and lowest in the segments Financial News and Financial News, Intuition, and Professional

Advice (both 0.55). The number of trades during the sample period is highest for investors in the

segment Fundamental Analysis (106.09 times) and lowest in the segment Intuition (59.80 times).

16

The tables also report HHI*, which measures the difference in HHI relative to equal weighting.

30

The previous combination of gender and turnover is consistent with earlier work by Barber and

Odean (2001) who find that relative to women, men are overconfident and trade heavily. The

combination of using fundamental analysis and excessive trading is in line with our expectations

that especially investors who feel they have more complete information are likely to make bold

forecasts and overcome their status quo bias, leading to less timid choices in terms of transaction

frequency. The average age is highest (51.05) in the segment Financial News, Intuition, and

Professional Advice and lowest (48.40) in the segment Intuition. Monthly turnover is highest

(46.63%) in the segment Financial News and Intuition and lowest (36.20%) in the segment

Technical Analysis. The segment Fundamental Analysis has the highest portfolio value (€72,509)

while the segment Intuition has the lowest portfolio value (€31,379). Investors in the segment

Financial News are most experienced (41.93 months) while those in the segment Technical

Analysis are least experienced (37.34 months). We also find interesting differences between

segments with regard to portfolio diversification. The segment Fundamental Analysis is best

diversified (8.05 different stocks, HHI is 25.68%), while the segment Intuition has the worst

diversification (5.68 different stocks, HHI is 30.56%). These investors may have less conviction

in their capabilities as they have less complete information, resulting in forecasts that are more

conservative and not sufficiently bold to overcome their status quo bias, leading to timid choices

(cf. Kahneman and Lovallo, 1993).

Table 3 demonstrates that investors in the segment Fundamental Analysis have the highest

ambition level (3.43), while investors in other segments, such as Intuition (3.09) and Financial

News (3.10) have more modest ambitions. In line with the previous results, investors in the

segment Fundamental Analysis have the most speculative risk profile (5.52), whereas investors

in the segment Financial News have the least speculative risk profile (5.09). Finally, whereas the

31

segments Fundamental Analysis (16.53%) and Technical Analysis (10.13%) have the highest

percentage of investors who regard themselves to be very advanced, these numbers are

considerably lower in the other segments, reaching a minimum in the segment Financial News

(3.95%). The lower score of the latter category of investors indicates that they may be less likely

to be overconfident about their own abilities. Instead of trying to make an independent estimate

of a company’s attractiveness using, for example, fundamental or technical analysis, they rely on

widely available financial news to make their investments.

VII. Performance per Investor Segment

In this section we compare the raw returns and alphas of the different segments of investors as

previously identified. We expect important differences between segments in terms of

performance due to the previously identified differences with respect to observable (e.g.,

turnover, age, transaction frequency, and portfolio diversification) as well as unobservable

variables (ambition level, risk profile, sophistication) and the predictions of the behavioral

portfolio framework. Table 4 reports the investment performance per investor segment.

[Table 4 about here]

A. Segments based on Investment Objectives

Panel A of Table 4 shows that the segment Speculation has the worst raw return (gross), while

the segment Capital Growth does best. The average investor in the segment Speculation loses

0.38% per month in gross terms, whereas the average investor in the segment Capital Growth

gains 0.68% per month.

32

The right hand side of Panel A shows that the performance difference between the different

segments of investors widens when transaction costs are taken into account. The return of the

segment Speculation incurs the most transaction costs, which is intuitive considering this

segment’s high turnover. The raw net return of this segment is now -2.22% per month, whereas

the performance of the segment Capital Growth is still positive with 0.22% per month.

After also adjusting for both risk and style tilts, the segment Capital Growth still achieves

the best performance with a net alpha of -0.40%, whereas the segment Speculation remains the

worst performer with a net alpha of -1.28%. The latter result is in line with the observable and

unobservable characteristics of the investors in this segment. Investors whose objective is to

speculate have high ambition levels, high risk profiles, high turnover, and judge themselves to be

very advanced. These characteristics are typical for overconfident investors who overtrade and

consequently underperform (Barber and Odean 2001). In addition, the factor loadings show that

these investors are heavily investing in small cap stocks, which may be a risky strategy in

combination with the lower levels of diversification we find for this segment.

B. Segments based on Investment Strategy

Panel B of Table 4 shows that the segment Technical Analysis has the worst raw return (gross),

while the segment Financial News and Intuition does best, closely followed by Fundamental

Analysis. The average investor in the segment Technical Analysis gains only 0.07% per month in

gross terms, whereas the average investor in the segment Financial Analysis and Intuition gains

0.86% and Fundamental Analysis 0.76% per month, respectively.

The right hand side of Panel A shows that when transaction costs are taken into account the

segment Technical Analysis remains the worst performer and the segment Financial News and

33

Intuition stays the best. The raw net return of the segment Technical Analysis becomes negative

at -0.92% per month, while the performance of the segment Financial News and Intuition stays

mildly positive at 0.13%.

This pattern also remains the same after adjusting for risk and style tilts, although the

difference between segments now narrows. The segment Financial News and Intuition

achieves the best performance with a net alpha of -0.46%, closely followed by the segment

Fundamental Analysis, which obtains a net alpha of -0.47%. The segments Technical Analysis

and Financial News, Intuition, and Professional Advice are the worst performers, having a net

alpha of -0.73% and -0.71% per month, respectively. The superior performance of the segments

Financial News and Intuition and Fundamental Analysis is interesting and suggests some stock-

picking skills.17

After all, these investors’ stock choices must be good enough to overcome the

detrimental effect of the relatively high level of transactions of these segments. The inferior

performance of the segment Financial News, Intuition, and Professional Advice is remarkable

and suggests that the advice of investment professionals may not be very helpful for the

performance of individual investors, but is associated with a relatively high number of

transactions and turnover. Finally, the inferior performance of the segment Technical Analysis

illustrates the limited usefulness of past stock market information for future return performance.

VIII. Testing of Hypotheses

This section reports the results of testing the hypotheses of the behavioral portfolio framework as

presented in section IV. To determine whether investment objectives and strategies result in

significant differences between investors regarding their investment behavior and return

17

It should be noted, however, that the alphas are negative across all groups.

34

performance we employ a series of t-tests and ANOVA’s (Hair et al., 1998). Detailed results are

provided in Tables 2-4.

As predicted by H1, investors relying on fundamental analysis are more overconfident than

those relying on technical analysis as reflected by the higher proportion of fundamental traders

who report to be either “advanced” (t(1584) = 5.64, p < 0.00) or “very advanced” (t(1584) =

3.78, p < 0.00) and the substantially larger proportion of technical traders who report to be

“novice investors” (t(1584) = 9.05, p < 0.00). Additionally, as predicted, trading frequency is

higher (t(1584) = 3.54, p < 0.00) for the more overconfident fundamental traders than for the less

overconfident technical traders.

Surprisingly, we have to reject H2, as despite their frequent trading, the risk and style

adjusted return performance of fundamental traders is actually higher than those of technical

traders (t(1584) = 2.06, p = 0.04). These results show that overtrading does not necessarily result

in underperformance (cf. Barber and Odean, 2000). Rather, underperformance depends on the

circumstances. In this case, we distinguish between traders relying on fundamental versus

technical analysis. We find that although fundamental investors trade more, they may not be

“overconfident” in the traditional sense, as their high level of confidence is actually warranted by

a detailed insight in the underlying economic fundamentals and their frequent trading leads to

higher returns even after accounting for transaction costs. These investors may learn by trading,

leading to superior returns (cf. Glaser and Weber, 2007; Nicolosi, Peng, and Zhu, 2009).

We accept H3: Investors relying on fundamental analysis are better diversified than

investors relying solely on their intuition as represented by the larger number of different stocks

that are held by the former group (t(1486) = 6.07, p < 0.00) and their lower HHI score (t(1420) =

3.83, p < 0.00). This result is in line with the discussion above, as relative to other investors,

35

investors relying on fundamental analysis use more information, which we hypothesize generates

more conviction with respect to their stock-picking skills. Investors relying on fundamental

analysis instead of intuition are more likely to be more sophisticated and therefore less impacted

by behavioral biases (Dhar and Zhu, 2006), such as regret (Kahneman et al., 1991) and status

quo bias (Samuelson and Zeckhauser, 1988), which can be related to under-diversification.

Our tests confirm H4: Investors with higher aspirations (above-median ambition level) take

more risk (t(5709) = 5.71, p < 0.00) as reflected by their risk profile (M = 5.37) than investors

with lower aspirations (below-median ambition level) (M = 5.04). As such, we confirm a key

feature of portfolio selection under the behavioral framework developed by Shefrin and Statman

(2000): ambitious investors are more comfortable taking on high risk.

We also confirm H5: The portfolios of investors with above-median risk profiles have higher

exposure (t(2152) = 7.16, p < 0.00) to the market factor (M = 1.41) than investors with below-

median risk profiles (M = 1.21). Also, investors with higher risk profiles invest more (t(2152) =

3.64, p < 0.00) in small caps (M = .71) than investors with below-median risk profiles (M = .59).

Thus, investors with a higher tolerance for risk also select more risky portfolios as indicated by

the respective factor loadings of their portfolios.

In line with H6, investors relying on fundamental analysis have the highest aspirations as

reflected by their ambition levels (F(5, 5452) = 17.35, p < 0.00) and take the most risk as

reflected in their risk profile (F(5, 5258) = 7.70, p < 0.00). The latter result may be surprising

from a “noise trader” perspective (Black, 1986) in which traders relying on technical analysis are

the ones taking most risk, but is in line with a behavioral portfolio framework in which investors

with the most information have the highest convictions in their stock picking skills, make bolder

36

forecasts, and set the most ambitious goals. To achieve these high aspirations, they ultimately

take more risk than other investors (cf. Fisher and Statman, 1997).18

Our tests confirm H7: Investors whose primary investment objective is to speculate or

exercise their hobby trade more frequently (F(4, 5495) = 9.32, p < 0.00) than investors who

invest primarily to build a financial buffer or save for retirement.

Finally, the evidence confirms H8: Investors whose primary investment objective is to build

a financial buffer or save for retirement have lower aspirations (F(4, 5453) = 17.38, p < 0.00)

and take less risk (F(4, 5259) = 36.99, p < 0.00) than investors who invest primarily as a hobby

or to speculate.

IX. Conclusions

Recent work (Barber et al., 2009) shows individual investors’ tendency to underperform relative

to the market. To date, variables which are relatively easy-to-observe such as age, gender, and

transaction channel have been used to explain this underperformance and are used as proxies for

typically unobservable psychological biases such as overconfidence, loss aversion, and

familiarity. To the best of our knowledge, the existing literature has not directly measured these

biases using consumer behavior methods such as investor surveys (Graham et al., 2009). Neither

has the existing literature positioned its findings of underperformance in a behavioral portfolio

framework by employing underlying variables which are less easy-to-observe such as investment

objective and strategy.

18

Behavioral portfolio theory acknowledges that different layers/parts of investors’ portfolios are connected to

different intrapersonal risk profiles (Shefrin and Statman, 2000).

37

In this paper we use a unique dataset involving 5,500 individual investors which contains

both “hard” accounting and “soft” survey data. We use these data to identify segments of

investors based on their dominant investment objectives and the investment strategies they use.

These investor segments are subsequently profiled using a combination of observable and

unobservable characteristics. Finally, the cross-sectional return performance of different

segments is analyzed and our behavioral hypotheses tested.

Our explanation for differences in return performance between different segments of

investors is novel in that instead of using proxies, we use a survey to measure directly investors’

underlying behavioral tendencies and psychological biases. We obtain data on a variety of

variables which typically remain unobservable and combine this with a selection of observable

variables. As such, we profile investor segments and test the hypotheses of our behavioral

portfolio framework. In doing so, we contribute to the emerging, but limited body of literature

investigating latent heterogeneity in finance (Heckman, 2001).

Our results might be useful for policy makers, as they show that “the usual suspects” of

individuals who trade excessively might differ from the actual culprits. We find that investors

using fundamental analysis actually trade more than investors relying on technical analysis,

which contrasts with the common belief but fits a behavioral portfolio framework. To the extent

that fundamental investors “think” they know the underlying fundamentals that drive stock prices

but actually do not, there is a clear target group for educational incentives that has not received

the attention it deserves until now. These investors may be provided with questionnaires and

self-administered investment quizzes to evaluate their true knowledge about market

fundamentals and tailor-made education offered by government agencies or financial authorities.

38

Figure 1: variables constructed from survey responses

Variables Answer categories

Investment objective

What is your most important investment objective with

the investment portfolio at this brokerage firm?

1 - Capital growth: achieve a higher expected

return than on a savings account

2 - Hobby: interest in stock market

3 – Saving for retirement: being able to stop

working on an earlier age

4 – Speculation: try to profit from short-term

developments on the stock market

5 – Building financial buffer: building a

financial buffer for future expenses

Investment strategy

Which strategies do you use as a basis for your

investment decisions (multiple answers possible)?

1 – Financial news: I base my investment

decisions on financial news

2 – Intuition: I base my investment decisions on

my personal intuition

3 – Technical analysis: I base my investment

decisions on technical analysis

4 – Fundamental analysis: I base my investment

decisions on fundamental analysis

5 – Professional advice: I base my investment

decisions on the professional advice from an

investment advisor

6 – Tips from others: I base my investment

decisions on tips from others such as friends or

family.

7 – Other

Ambition level

How ambitious do you consider yourself to be? 1 – I am not ambitious

2 – I am a bit ambitious

3 – I am moderately ambitious

4 – I am quite ambitious

5 – I am very ambitious

Risk profile

Investors answer a set of questions, measuring their

sensitivity for losses, time horizon, and subjective

probabilities of chance events. This leads to a

categorization between 1 and 7.

1 – Saving (no investment in (risky) equity)

2 – Very defensive

3 – Defensive

4 – Careful

5 – Offensive

6 – Speculative

7 – Very speculative

Investor Sophistication

What kind of investor do you consider yourself to be? 1 – A novice investor

2 – An advanced investor

3 – A very advanced investor

39

Table 1: descriptive statistics

A: 5,500 Respondents of the Investor Survey Mean Std. Dev 5th Pctl 25th Pctl Median 75th Pctl 95th Pctl

Gender (male =1) 0.58

Age in 2006 (years) 49.70 12.73 28.00 40.00 50.00 59.00 70.00

Trades (#) 76.45 132.00 1.00 9.00 30.00 83.00 311.00

Turnover (%) 42.40 121.00 0.00 3.89 10.99 31.48 173.05

Portfolio value (€) 45,915 142,576 1,057 5,321 15,234 42,406 166,840

Experience (months) 40.21 20.91 9.00 22.00 39.00 60.00 72.00

Number of stocks held 6.57 7.39 1.00 2.00 4.00 8.00 20.00

HHI (%) 27.78 23.28 1.10 9.80 21.14 39.73 78.42

HHI* (%) 17.20 21.55 0.16 4.06 9.06 20.74 70.69

Monthly Net Returns -0.003 0.059 -0.071 -0.010 0.003 0.010 0.041

Risk Profile (1-7) 5.31 1.61 2.00 4.00 6.00 7.00 7.00

B: 59,825 Non-Respondents of the Investor Survey Mean Std. Dev 5th Pctl 25th Pctl Median 75th Pctl 95th Pctl

Gender (male =1) 0.61***

Age in 2006 (years) 45.92*** 12.28 27.00 37.00 45.00 55.00 67.00

Trades (#) 44.41*** 104.00 0.00 2.00 10.00 38.00 210.00

Turnover (%) 33.10*** 189.00 0.00 0.50 4.50 17.26 128.51

Portfolio value (€) 28,253*** 163,483 542 2,289 7,158 21,703 106,459

Experience (months) 34.21*** 23.02 2.00 13.00 31.00 55.00 72.00

Number of stocks held 6.24*** 7.11 1.00 2.00 4.00 8.00 19.00

HHI (%) 35.99*** 20.71 1.00 21.64 36.81 47.29 76.05

HHI* (%) 25.85*** 26.00 0.01 7.63 17.41 32.31 91.09

Monthly Net Returns -0.02*** 0.095 -0.016 -0.023 -0.002 0.010 0.043

Risk Profile (1-7) 4.83*** 1.86 2.00 3.00 5.00 7.00 7.00

This table presents descriptive statistics for a sample of 65,325 investor accounts at a Dutch online broker. We split

the sample into 5,500 investors who participated in our investor survey and 59,825 who did not. The sample period

is from January 2000 to March 2006. The variables are defined as follows: Gender refers to the fraction of accounts

hold by a male investor only. Age is the age in years of the main account holder. Trades is the total number of stock

trades per account during the sample period. Turnover is the average of the value of all stock purchases and sales in

a given month divided by the beginning-of-the-month account value. Portfolio value is the average market value of

all assets in the investor’s portfolio. Experience is the number of months an investor has been trading. Number of

stocks held refers to the number of different stocks an investor has in portfolio at the end of the sample period. HHI

refers to the Herfindahl-Hirschmann Index value for an investors’ portfolio at the end of the sample period (the HHI

is defined as the sum of the squared portfolio weights of all assets. For the purpose of the HHI calculations, mutual

funds are assumed to consist of 100 equally-weighted, non-overlapping, positions). HHI* refers to the normalized

index: (H – (1/N)) / (1 – (1/N)). Comparing HHI with HHI* makes clear how different the value from the index is

from uniform weights. Monthly net returns is the average raw return per month corrected for transaction costs. Risk

profile refers to the self-reported riskiness of investors’ portfolios (1=very defensive, 7=very speculative). The table

shows for each variable the mean, median, and standard deviation, as well as 5th

, 25th

, 75th

, and 95th

percentile

values. If there is a statistically significant difference between attribute means reported for the two samples (survey

respondents and non-respondents), it is noted by asterisks in the mean columns of the non-respondent sample. The

mean comparison tests allow for different variances within the two groups. ***/**/* indicate that the means are

significantly different at the 1%/5%/10% level.

40

Table 2: descriptives per investor segment – observable variables

Segments based on investment objective Gender (male=1) Age in 2006 (years) Trades (#) Turnover (%) Portfolio value (€) Experience (months) # Stocks held HHI (%) HHI* (%)

Capital Growth (N=2422 ) 0.56 50.99 79.62 35.61 62,646 41.88 7.27 25.32 15.66

Hobby (N=1395 ) 0.62 47.31 65.20 43.43 24,139 39.18 5.43 32.25 19.28

Saving for Retirement (N=353 ) 0.53 49.85 75.33 26.44 49,359 43.49 7.57 24.39 15.46

Speculation (N=688 ) 0.64 48.61 99.25 78.87 33,579 34.47 5.63 30.38 17.69

Building Financial Buffer (N=642 ) 0.55 50.85 65.13 35.46 45,915 40.53 6.39 28.82 19.21

P-value of F-test 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

(7.68)*** (21.83)*** (9.32)*** (20.03)*** (18.19)*** (20.15)*** (12.21)*** (15.52)*** (5.63)***

Segments based on investment strategy Gender (male=1) Age in 2006 (years) Trades (#) Turnover (%) Portfolio value (€) Experience (months) # Stocks held HHI (%) HHI* (%)

Financial News (N=963 ) 0.55 50.64 67.04 44.71 38,992 41.93 5.69 28.27 17.96

Financial News and Intuition (N=1235 ) 0.58 50.44 82.49 46.63 57,227 39.87 7.39 26.99 16.34

Intuition (N=1442 ) 0.59 48.40 59.80 39.89 31,379 41.06 5.68 30.56 18.69

Technical Analysis (N=878 ) 0.58 49.35 79.11 36.20 39,470 37.34 6.11 26.82 16.39

Fundamental Analysis (N=708 ) 0.64 49.45 106.09 43.10 72,509 41.12 8.05 25.68 15.74

Financial News, Intuition, and Professional Advice (N=274 ) 0.55 51.05 84.81 46.46 47,705 38.16 7.81 25.21 16.18

P-value of F-test 0.01 0.00 0.00 0.40 0.00 0.00 0.00 0.00 0.07

(2.99)** (5.82)*** (13.76)*** (1.03) (10.39)*** (5.98)*** (12.68)*** (4.37)*** (2.02)*

This table presents descriptive statistics for a sample of 5,500 investor accounts at a Dutch online broker regarding a number of observable variables. We split the sample into

5 (6) segments using univariate sorting (cluster analysis) on investment objective (strategy). N refers to the number of investor accounts within each segment. The sample

period is from January 2000 to March 2006. The variables are defined as follows: Gender refers to the fraction of accounts hold by a male investor only. Age is the age in

years of the main account holder. Trades is the total number of stock trades per account during the sample period. Turnover is the average of the value of all stock purchases

and sales in a given month divided by the beginning-of-the-month account value. Portfolio value is the average market value of all assets in the investor’s portfolio.

Experience is the number of months an investor has been trading. Number of stocks held refers to the number of different stocks an investor has in portfolio at the end of the

sample period. HHI refers to the Herfindahl-Hirschmann Index value for an investors’ portfolio at the end of the sample period (the HHI is defined as the sum of the squared

portfolio weights of all assets. For the purpose of the HHI calculations, mutual funds are assumed to consist of 100 equally-weighted, non-overlapping, positions). HHI*

refers to the normalized index: (H – (1/N)) / (1 – (1/N)). Comparing HHI with HHI* makes clear how different the value from the index is from uniform weights. The table

shows for each variable the mean. We report the p-value of F-tests to detect significance differences between segments, reporting the F-ratio between brackets. ***/**/*

indicate that the means are significantly different between the segments at the 1%/5%/10% level.

41

Table 3: descriptives per investor segment – variables which are typically unobservable

Segments based on investment objective Ambition (1-5) Risk Profile (1-7) Novice Investor (%) Advanced Investor (%) Very Advanced Investor (%)

Capital Growth (N=2422 ) 3.21 5.15 37.57 54.50 7.31

Hobby (N=1395 ) 3.16 5.54 44.44 49.17 5.87

Saving for Retirement (N=353 ) 3.26 4.98 37.39 54.67 7.64

Speculation (N=688 ) 3.52 5.80 24.13 59.74 15.84

Building Financial Buffer (N=642 ) 3.15 5.05 39.88 54.98 4.67

P-value of F-test 0.00 0.00 0.00 0.00 0.00

(17.38)*** (36.99)*** (20.80)*** (5.70)*** (20.08)***

Segments based on investment strategy Ambition (1-5) Risk Profile (1-7) Novice Investor (%) Advanced Investor (%) Very Advanced Investor (%)

Financial News (N=963 ) 3.10 5.09 46.52 49.12 3.95

Financial News and Intuition (N=1235 ) 3.30 5.24 35.79 56.60 7.21

Intuition (N=1442 ) 3.09 5.43 48.06 46.12 5.48

Technical Analysis (N=878 ) 3.31 5.29 34.85 53.99 10.13

Fundamental Analysis (N=708 ) 3.43 5.52 15.25 67.80 16.53

Financial News, Intuition, and Professional Advice (N=274 ) 3.34 5.27 31.75 62.77 4.74

P-value of F-test 0.00 0.00 0.00 0.00 0.00

(17.35)*** (7.70)*** (54.11)*** (22.70)*** (23.97)***

This table presents descriptive statistics for a sample of 5,500 investor accounts at a Dutch online broker regarding a number of unobservable variables. We split the sample

into 5 (6) segments using univariate sorting (cluster analysis) on investment objective (strategy). N refers to the number of investor accounts within each segment. The sample

period is from January 2000 to March 2006. The variables are defined as follows: Ambition refers to the self-reported ambition level of an investor (1=not ambitious, 5=very

ambitious). Risk profile refers to the self-reported riskiness of investors’ portfolios (1=very defensive, 7=very speculative). Novice/advanced/very advanced investor refers to

the self-reported “sophistication” of investors and reports the percentage of investors per segment in each of the three categories. The table shows for each variable the mean.

We report the p-value of F-tests to detect significance differences between segments, reporting the F-ratio between brackets. ***/**/* indicate that the means are significantly

different between the segments at the 1%/5%/10% level.

42

Table 4: investment performance per investor segment

Gross Returns Net Returns

A: Segments based on investment objective

1 2 3 4 5 P-value of F-test 1 2 3 4 5 P-value of F-test

Raw Return 0.0068 0.0034 0.0065 -0.0038 0.0057 0.00 (5.88)*** 0.0022 -0.0064 0.0003 -0.0222 0.0003 0.00 (25.02)***

Alpha (Carhart) -0.0040 -0.0066 -0.0061 -0.0128 -0.0054 0.00 (13.51)***

Factor Loadings:

RMRF 1.22 1.40 1.19 1.63 1.34 0.00 (19.50)***

SMB 0.57 0.75 0.56 0.86 0.66 0.00 (9.73)***

HML 0.22 0.24 0.21 0.27 0.21 0.48 (0.88)

UMD -0.03 -0.05 0.01 0.00 -0.05 0.28 (1.28)

Adj. R 2 (%) 64.45 58.65 64.42 56.53 63.12 0.00 (16.46)***

B: Segments based on investment strategy

1 2 3 4 5 6 P-value of F-test 1 2 3 4 5 6 P-value of F-test

Raw Return 0.0041 0.0086 0.0025 0.0007 0.0076 0.0012 0.00 (3.65)*** -0.0027 0.0013 -0.0054 -0.0092 0.0003 -0.0065 0.00 (6.49)***

Alpha (Carhart) -0.0057 -0.0046 -0.0058 -0.0073 -0.0047 -0.0071 0.00 (4.29)***

Factor Loadings:

RMRF 1.31 1.32 1.33 1.24 1.30 1.32 0.60 (0.73)

SMB 0.67 0.67 0.67 0.57 0.55 0.70 0.08 (2.00)*

HML 0.26 0.24 0.21 0.20 0.22 0.17 0.19 (1.48)

UMD 0.00 -0.06 -0.06 0.02 -0.01 0.03 0.00 (4.00)***

Adj. R 2 (%) 62.70 63.58 61.92 58.52 63.69 63.61 0.00 (3.95)***

This table presents investment performance per investment segment. We report raw gross returns, raw net returns, and alphas and factor loadings based on net returns. For

panel A the numbers 1-5 refer to the following investor segments: 1=Capital growth, 2=Hobby, 3=Saving for retirement, 4=Speculation, 5=Building financial buffer. For

panel B the numbers 1-6 refer to the following investor segments: 1=Financial news, 2=Financial news and intuition, 3=Intuition, 4=Technical analysis, 5=Fundamental

analysis, 6=Financial news, intuition, and professional advice. We report the p-value of F-tests to detect significance differences between segments, reporting the F-ratio

between brackets. ***/**/* indicate that the means are significantly different between the segments at the 1%/5%/10% level.

43

Reference List

Ameriks, J. and S. Zeldes (2004). How Do Household Portfolios Vary With Age? Working

Paper, Columbia University.

Barber, B. M., Y-T Lee, Y-J Liu, and T. Odean (2009). Just How Much Do Individual Investors

Lose by Trading? Review of Financial Studies, 22(2), 609-32.

Barber, B. M. and T. Odean (2000). Trading is Hazardous to Your Wealth: the Common Stock

Investment Performance of Individual Investors. The Journal of Finance, 55(2), 773-806.

Barber, B. M. and T. Odean (2001). Boys Will be Boys: Gender, Overconfidence, and Common

Stock Investment. The Quarterly Journal of Economics, 1 261-92.

Barber, B. M. and T. Odean (2008). All That Glitters: The Effect of Attention and News on the

Buying Behavior of Individual and Institutional Investors. Review of Financial Studies, 21

785-818.

Barberis, N. and W. Xiong (2008). Realization Utility. NBER Working Paper No.14440.

Bauer, R., M. Cosemans, and P. M. A. Eichholtz (2009). Option Trading and Individual Investor

Performance. Journal of Banking and Finance, 33(4), 731-46.

Benartzi, S. and R. H. Thaler (2001). Naive Diversification Strategies in Defined Contribution

Savings Plans. American Economic Review, 91(1), 79-98.

Black, F. (1986). Noise. The Journal of Finance, 41(3), 529-43.

Browning, M. and T. F. Crossley (2001). The Life-Cycle Model of Consumption and Saving.

Journal of Economic Perspectives, 15(3), 3-22.

Carhart, M (1997). On Persistence in Mutual Fund Performance. The Journal of Finance, 52 57-

82.

Choi, J., J. Beshears, D. Laibson, and B. Madrian (2008). The Importance of Default Options for

Retirement Savings Outcomes: Evidence from the United States, in Lessons from Pension

Reform in the Americas, S. J. Kay and T. Sinha, eds. Oxford University Press.

Das, S., H. Markowitz, J. Scheid, and M. Statman (2010). Portfolio Optimization with Mental

Accounts. Journal of Financial and Quantitative Analysis, 45(2), 311-334.

De Bondt, W. F. M. and R. H. Thaler (1985). Does the stock market overreact? The Journal of

Finance, 40(3), 793-805.

Dhar, R. and N. Zhu (2006). Up Close and Personal: Investor Sophistication and the Disposition

Effect. Management Science, 52(5), 726-40.

Dorn, D. and G. Huberman (2005). Talk and Action: What Individual Investors Say and What

They Do. Review of Finance, 9 437-81.

Dorn, D. and P. Sengmueller (2009). Trading as Entertainment? Management Science, 55(4),

591-603.

Ellsberg, D. (1961). Risk, Ambiguity, and the Savage Axioms. Quarterly Journal of Economics,

74(4), 643-69.

Fisher, K. L. and M. Statman (1997). Investment Advice from Mutual Fund Companies. Journal

of Portfolio Management, 24(1), 9-25.

Fox, C. R. and A. Tversky (1995). Ambiguity Aversion and Comparative Ignorance. The

Quarterly Journal of Economics, 110(3), 585-603.

French, K. R. and J. Poterba (1991). Investor diversification and international equity markets.

American Economic Review, 81 222-6.

Glaser, M. and M. Weber (2007). Why Inexperienced Investors do not Learn: They do not Know

their past Portfolio Performance. Finance Research Letters, 4 203-16.

44

Goetzmann, W. N. and A. Kumar (2008). Equity Portfolio Diversification. Review of Finance,

12 433-63.

Graham, J. R., C. R. Harvey, and H. Huang (2009). Investor Competence, Trading Frequency,

and Home Bias. Management Science, 55(7), 1094-106.

Grinblatt, M. and M. Keloharju (2006). Sensation Seeking, Overconfidence, and Trading

Activity. NBER Working Paper No.12223.

Hahn, M., R. Lawson, and Y. G. Lee (1992). The Effects of Time Pressure and Information Load

on Decision Quality. Psychology and Marketing, 9(5), 365-78.

Hair, J. F., R. E. Anderson, R. L. Tatham, and W. C. Black (1998). Multivariate Data Analysis.

Upper Saddle River, New Jersey: Prentice Hall.

Heath, C. and A. Tversky (1991). Preference and Belief: Ambiguity and Competence in Choice

under Uncertainty. Journal of Risk and Uncertainty, 4 5-28.

Heckman, J. J. (2001). Micro Data, Heterogeneity, and the Evaluation of Public Policy: Nobel

Lecture. Journal of Political Economy, 109(4), 673-748.

Huberman, G. (2001). Familiarity breeds investment. Review of Financial Studies, 14(3), 659-80.

Kahneman, D., J. L. Knetsch, and R. H. Thaler (1991). Anomalies: The Endowment Effect, Loss

Aversion, and Status Quo Bias. Journal of Economic Perspectives, 5(1), 193-206.

Kahneman, D. and D. Lovallo (1993). Timid Choices and Bold Forecasts: A Cognitive

Perspective on Risk Taking. Management Science, 39(1), 17-31.

Kahneman, D. and A. Tversky (1979). Prospect theory: an analysis of decision under risk.

Econometrica, 47 263-91.

Karlsson, A. and L. Norden (2007). Home sweet home: Home bias and international

diversification among individual investors. Journal of Banking and Finance, 31 317-33.

Keim, D. and A. Madhavan (1998). The Cost of Instututional Equity Trades. Financial Analysts

Journal, 54 50-69.

Kumar, A. (2009). Who Gambles in the Stock Market? The Journal of Finance, Forthcoming.

Kumar, A., J. Page, and O. Spalt (2009). Religious Beliefs, Gambling Attitudes, and Financial

Market Outcomes. Working Paper.

Lease, R. C., W. G. Lewellen, and G. G. Schlarbaum (1974). The Individual Investor: Attributes

and Attitudes. The Journal of Finance, 29(2), 413-33.

Lee, H.-J., J. Park, J.-Y. Lee, and R. S. Wyer (2008). Disposition Effects and Underlying

Mechanisms in E-Trading of Stocks. Journal of Marketing Research, 45(3), 362-78.

Lewellen, W. G., R. C. Lease, and G. G. Schlarbaum (1980). Portfolio Design and Portfolio

Performance: The Individual Investor. Journal of Economics and Business, 32(3), 185-97.

Locander, W. B. and P. W. Hermann (1979). The Effect of Self-Confidence and Anxiety on

Information Seeking in Consumer Risk Reduction. Journal of Marketing Research, 16(May),

268-74.

Lopes, L. (1987). Between Hope and Fear: The Psychology of Risk. Advances in Experimental

Social Psychology, 20 255-95.

Manski, C. F. (2004). Measuring Expectations. Econometrica, 72(5), 1329-76.

Markowitz, H. M. (1952). Portfolio Selection. The Journal of Finance, 7 77-91.

Merton, R. C. (1971). Optimum Consumption and Portfolio Rules in a Continuous-Time Model.

Journal of Economic Theory, 3 373-413.

Mitchell, O., G. Mottola, S. Utkus, and T. Yamaguchi (2006). The Inattentive Participant:

Portfolio Trading Behavior in 401(k) Plans. Working paper, Wharton.

45

Nagy, R. A. and R. W. Obenberger (1994). Factors Influencing Individual Investor Behavior.

Financial Analysts Journal, 50(4), 63-8.

Nicolosi, G., L. Peng, and N. Zhu (2009). Do Individual Investors Learn from their Trading

Experience? Journal of Financial Markets, 12 317-36.

Odean, T. (1998). Volume, Volatility, Price, and Profit When All Traders Are above Average.

The Journal of Finance, 53(6), 1887-934.

Odean, T. and B. M. Barber (2002). Online Investors: Do the Slow Die First? Review of

Financial Studies, 15(2), 455-87.

Oskamp, S. (1965). Overconfidence in Case-Study Judgments. Journal of Consulting

Psychology, 29(3), 261-5.

Pennings, J. M. E. and P. Garcia (2009). Risk & Hedging Behavior: The Role and Determinants

of Latent Heterogeneity. The Journal of Financial Research, Forthcoming.

Punj, G. and D. W. Stewart (1983). Cluster Analysis in Marketing Research: Review and

Suggestions for Application. Journal of Marketing Research, 20(May), 134-48.

Rantapuska, E. (2006). Essays on Investment Decisions of Individual and Institutional Investors.

Helsinki School of Economics: Doctoral Dissertation.

Samuelson, W. and R. Zeckhauser (1988). Status Quo Bias in Decision Making. Journal of Risk

and Uncertainty, 1(1), 7-59.

Seru, A., T. Shumway, and N. Stoffman (2008). Learning by Trading. Working Paper,

University of Chicago and University of Michigan.

Shefrin, H. (2008). A Behavioral Approach to Asset Pricing. Boston: Elsevier.

Shefrin, H. and M. Statman (1985). The Disposition to Sell Winners too Early and Ride Losers

Too Long: Theory and Evidence. The Journal of Finance, 40 777-90.

Shefrin, H. and M. Statman (2000). Behavioral Portfolio Theory. The Journal of Financial and

Quantitative Analysis, 35(2), 127-51.

Shleifer, A. and J. O. Summers (1990). The Noise Trader Approach to Finance. Journal of

Economic Perspectives, 4(2), 19-33.

Slovic, P. and B. Corrigan (1973). Behavioral Problems of Adhering to a Decision Policy. Talk

presented at The Institute for Quantitative Research in Finance, May 1, Napa, CA.

Statman, M. (2002). Lottery Players / Stock Traders. Financial Analysts Journal, 58(1), 14-21.

Statman, M. (2004), What Do Investors Want? Journal of Portfolio Management, 153-161.

Thaler, R. H. (1985). Mental Accounting and Consumer Choice. Marketing Science, 4(3), 199-

214.

Thaler, R. H. (2000). Mental Accounting Matters, in Choices, Values, and Frames, D.

Kahneman and A. Tversky, eds. New York: Cambridge University Press, 241-68.

Thaler, R. H. and E. J. Johnson (1990). Gambling With the House Money and Trying to Break

Even: The Effects of Prior Outcomes on Risky Choice. Management Science, 36(6), 643-60.

Tversky, A. and D. Kahneman (1992). Advances in Prospect Theory: Cumulative Representation

of Uncertainty. Journal of Risk and Uncertainty, 5(4), 297-323.

Viceira, L. M. (2001). Optimal Portfolio Choice for Long-Horizon Investors with Nontradable

Labor Income. The Journal of Finance, 56(2), 433-70.

Wilcox, R. T. (2003). Bargain Hunting or Star Gazing? Investors' Preferences for Stock Mutual

Funds. Journal of Business, 76(4), 645-63.