Post on 09-Jul-2018
1
The First Cut is the Deepest:
Stock Market Re-entry Decision and Hot Stove Effect
Ozlem Arikan
1 Arie E. Gozluklu
2
Gi H. Kim3 Hiroaki Sakaguchi
4
This Version: 19th
of May, 2016
Abstract
This paper analyses how the first-time experience in the stock market affects the re-entry decision
of inexperienced market participants. Using a unique dataset of Finnish retail investors, we show
that the complete withdrawal from the stock market after an initial loss - the hot stove effect (HSE) -
is prevalent in the Finnish market. The HSE is especially pronounced after heavy losses in the first-
time investment. The HSE is stronger in the bear market after the burst of the dot-com bubble. While
the gender of the investor does not play a role, age significantly affects the re-entry decision: the older
the investor the more likely that she will withdraw from the market entirely after an initial loss. We
find that the HSE is not significantly weaker for Nokia shares, which is a well-recognized name for
Finnish market participants. Our findings have important implications for financial literacy.
KEYWORDS: hot stove effect, market participation, Finnish market, trader characteristics,
financial literacy, re-entry decision.
1,2,3,4
University of Warwick; Correspondence to Gi Kim at: gi.kim@wbs.ac.uk. We would like to thank the
participants of WBS Behavioural Science group meeting for valuable comments and suggestions.
2
The First Cut is the Deepest:
Stock Market Re-entry Decision and Hot Stove Effect
Abstract
This paper analyses how the first-time experience in the stock market affects the re-entry decision
of inexperienced market participants. Using a unique dataset of Finnish retail investors, we show
that the complete withdrawal from the stock market after an initial loss - the hot stove effect (HSE) -
is prevalent in the Finnish market. The HSE is especially pronounced after heavy losses in the first-
time investment. The HSE is stronger in the bear market after the burst of the dot-com bubble. While
the gender of the investor does not play a role, age significantly affects the re-entry decision: the older
the investor the more likely that she will withdraw from the market entirely after an initial loss. We
find that the HSE is not significantly weaker for Nokia shares, which is a well-recognized name for
Finnish market participants. Our findings have important implications for financial literacy.
KEYWORDS: hot stove effect, market participation, Finnish market, trader characteristics,
financial literacy, re-entry decision.
3
1. Introduction
Individuals are increasingly more responsible to make their own financial decision to smooth
their consumption over life-time. In the US, the retirement system moves toward defined contribution
plans where individuals are asked to make financial choices. Similarly, the UK government promotes
private pension schemes “to save money for later in life”. These changes in the pension system as a
result of the ageing population in the developed economies, shifts the financial decision making about
future consumption to naïve individual investors (van Rooij et al., 2011).
Most recent academic papers (Bali et al., 2009; Barberis, 2000; Campbell and Viceira, 2005;
Levi, 2015; Siegel, 2014) and advice from the finance industry agree that stocks are an important
investment vehicle for longer investment horizons.1 While there is no consensus on the optimal
weight one should invest in risky assets, even for long-term investors (Bodie, 1995; Pastor and
Stambaugh, 2012), there is a consensus that stocks -especially a well-diversified stock portfolio such
as the market index- as an asset class should be included in a wealth portfolio to exploit
diversification benefits.2 For example, Viceira (2001)’s model suggests that employed individuals
should invest more in stocks, in particular if their labor income is not correlated with stock returns.
However, there is an extant literature on limited participation puzzle (Allen and Gale, 1994; Brav et
al., 2002; Guvenen, 2009; Vissing-Jorgensen, 2002). Even those who participate in the stock market,
often enter the market only with a limited number of stocks (Calvet et al., 2007; Goetzmann and
Kumar, 2008).
Earlier studies so far mainly focus on first-time market participation (Guiso and Japelli, 2005;
Guiso et al., 2008; van Rooij et al., 2011) and investor attrition (Seru et al., 2010), which is related to
the exit decision. In particular, they show that such decisions are highly dependent on financial
literacy, especially when it comes to investment in stocks (van Rooij et al., 2011). Most of the
1 Target-date (life-cyle) funds offered by Fidelity or Vanguard suggest 90% stock allocation for long-term investors (Viceira, 2008, Pastor and
Stambaugh, 2012). 2 Stocks, in principle, should also provide hedge against inflation, since stocks are claims to productive capacity as oppose to bonds with nominal claims (Bodie, 1976; Schotman and Schweitzer, 2000; Kim and In, 2005).
4
individual investors are also subject to behavioral biases (see, for example, the survey by Barberis and
Thaler, 2003) such as disposition effect -selling stocks too early after gains and holding losing
securities for too long-, insufficient (naïve) diversification strategies, and overconfidence resulting in
excessive trading especially among male market participants (Barber and Odean, 2001).
This paper has a different focus. To the best of our knowledge, it is the first empirical paper
that studies the stock market re-entry versus withdrawal decision of inexperienced naïve investors.
Exploiting the detailed Finnish dataset used in earlier studies (e.g., Seru et al., 2010; Grinblatt and
Keloharju, 2000, 2001a, 2001b), we first confirm that such individual investors are subject to a hot
stove effect (HSE), that is, individuals shy away from the stock market after an initial bad experience
as shown by the previous literature (e.g., Seru et al., 2010). In particular, we identify the HSE by
focusing on the differences in re-entry decisions among inexperienced investors after initial gains and
losses. If the HSE effect is dominant, we shall observe a higher withdrawal rate from the market
among inexperienced investors after an initial loss. However, if the prospect theory (Kahnemann and
Tversky, 1979) is a better description of the behavior of individual investors, then we shall observe a
more risk-seeking attitude, hence a higher re-entry ratio, after the initial loss. We then confirm that
the magnitude of the first investment, investor characteristics such as gender and age, bull versus bear
markets, or exiting with a well-known stock such as Nokia affect the re-entry decision. In the
robustness section, we also test whether our results are sensitive to the number of stocks in the initial
portfolio, to the location of the investor, Helsinki versus outside Helsinki, and to the exclusion of the
most focal firm (Nokia) in our sample.
We confirm for our data set that the HSE is prevalent in the Finnish market after controlling
for market conditions, time-fixed effects and the duration of the first-time experience in the stock
market, and that the HSE is more pronounced after heavy losses. We also show that the HSE is
stronger in the bear market, in the gloomy period after the burst of the dot-com bubble. While the
gender of the investor is a determining factor for market re-entry, that is, females are less likely to
5
come back in line with the findings of Barber and Odean (2001), it does not play a role in the context
of the HSE. We did not find a significant difference in male versus female re-entry decisions after
exiting with a loss. Age, on the other hand, significantly affects the re-entry decision after an initial
loss: the older the investor the more likely that (s)he will withdraw from the market entirely after an
initial loss. The HSE is not significantly weaker for the first-time holders of Nokia shares. Overall,
our results imply that the first-time investor does not exhibit risk-seeking behavior after an initial loss,
and shy away from the market, shutting down an important channel to transfer consumption across
time.
The rest of the paper is organized as follows. In Section 2, we discuss the related literature
and introduce the predictions regarding the HSE. In Section 3, we provide the empirical analysis
describing the data, our identification strategy, non-parametric univariate analysis and multivariate
logit results. In Section 4, we show the robustness results. We conclude in Section 5.
2. Related Literature and Predictions
It is common in financial economics to assume that the individuals have stable preferences
(aversion) towards risk which describe the individual decision making when facing risky choices (for
example, Holt and Laury, 2002). The psychology literature, on the other hand, provides an alternative
description based on experiential learning which formulizes Mark Twain’s observation3, namely the
hot stove effect (HSE). Denrell and March (2001) argue that as part of the adaptation process
individuals try to reproduce successful outcomes, and hence alternatives that have experienced
relatively good outcomes in the past are more likely to be sampled than are alternatives that have
experienced relatively poor outcomes (Holland 1975). This type of experience-based learning results
3 “We should be careful to get out of an experience only the wisdom that is in it—and stop there: lest we be like the cat that sits down on a hot
stove lid. She will never sit down on a hot stove lid again—and that is well; but also she will never sit down on a cold one.” Mark Twain (1897).
6
in missing out potentially good outcomes and creates bias against risky choices. Fujikawa (2009)
provides experimental evidence on the existence of the HSE.
In the finance context, Strahilevitz et al. (2011) document that investors also make decisions
based on past experience, and find that they are reluctant to repurchase stocks previously sold for a
loss. Huang (2013) extends Strahilevitz et al. (2011)’s findings and show that investors are also
reluctant to buy similar stocks from the same industry after a loss. Seru et al. (2010) find that past
negative experience in the stock market also affects the exit decision from the market. In an IPO
setting, Kaustia and Knüpfer (2008) confirm that investors repeat behavior that has produced good
outcomes in the past and avoids behavior that has produced poor outcomes. They find that an increase
in the returns that an investor earns on past IPO investments has a positive impact on this investor’s
propensity to participate in future IPOs. On the other hand, Malmendier and Nagel (2009) show that
not only the personal experience in the stock market but also individual experience of macroeconomic
shocks related to the stock market affect stock market participation.
However, none of the earlier papers directly test how an initial experience in the stock market
affect the re-entry decision to the market. If the naïve investors behave according to experience-based
learning, we expect that their re-entry decision will be conditional on their first-time experience.
Therefore in line with the previous literature cited above, we predict the following:
Prediction 1: Investors with a limited first-time experience in the stock market are more likely
to withdraw from the market completely after an initial loss. In other words, the hot stove effect
(HSE) reduces the likelihood of stock market re-entry.
One can argue that not only an initial loss but also the magnitude of the realized loss may
impact investors’ behavior. For example, Strahilevitz et al. (2011) find that if a stock is originally sold
for a loss, the likelihood of repurchase of the same stock drops nearly linearly with the magnitude of
that loss. Similarly, Seru et al. (2010) suggest that magnitude of losses is important in investors’ exit
7
decision.4 They find that an investor whose performance is one standard deviation worse than the
mean is about 15% less likely to continue trading in the market. In a savings context, Choi et al.
(2009) show that individual investors over-extrapolate from their personal return experience. They
find that within a given time period, investors who experience particularly rewarding outcomes from
saving in their 401(k)—a high average and/or a low variance rate of return—increase their 401(k)
savings rate more than investors who have less rewarding experiences with saving. Therefore, our
second prediction is as follows:
Prediction 2: The bigger the loss in the first-time stock market experience the less likely is
that the investor comes back to the market. In other words, the hotter the stove, the stronger is the
effect on market re-entry decision.
In both survey and experimental settings, De Bondt (1993) finds that non-professional
investors are overly optimistic in bull markets and overly pessimistic in bear markets. Using market
based data, Bange (2000) confirms that investors are positive feedback traders. She finds that small
investors increase their equity holdings following market run-ups, and decrease their holdings after
downturns. Amromin and Sharpe (2009) find that expected risk and returns are strongly influenced
by expected economic conditions. When investors believe that macroeconomic conditions are more
expansionary, they tend to expect both higher returns and lower volatility. Similarly, Malmendier and
Nagel (2009) examine whether people who live through different macroeconomic histories differ in
their level of risk taking. They find that individuals, who have experienced low stock market returns
throughout their lives, report lower willingness to take financial risk, are less likely to participate in
the stock market, invest a lower fraction of their liquid assets in stocks if they participate, and are
more pessimistic about future stock returns. However, they also find that the effects of economic
4 Seru et al. (2010) focus on learning via trading, and consider investors with some experience in the market, that is, those with at least seven
round-trip trades.
8
downturns vanish over time; individuals are influenced more strongly by recent returns than distant
returns.
The literature above suggest that individual investors are more optimistic in the boom periods
than in the bear periods, and past macroeconomic conditions affect individuals’ decisions, but this
effect vanishes over time. Therefore, it is worthwhile to ask whether first-time investors who exit the
market with a loss in a bear period are less likely to come back.
Prediction 3: First-time investors leaving the market with a loss in a bear period are more
likely to withdraw completely from the market. In other words, the HSE is stronger in a bear period.
An important question is how the investor characteristics such as gender and age affect
experience-based decision making in the context of HSE. Previous research which examines gender
differences in risk aversion generally finds that females are more risk averse than males (Halek and
Eisenhauer 2001). Eckel and Grossman (2008) review experimental evidence examining risk
preferences across genders, and finds that while most results reveal that females are more risk averse
than males, results are less clear in contextual settings. Schubert et al. (1999) argue that although
there is some evidence that women are more risk averse than men it is questionable whether such
stereotypic risk attitudes can be confirmed in real life. They posit that behavior in abstract gambling
experiments may not correspond to risk behavior in contextual decisions. They find that when the
context is an insurance choice or a financial investment choice, there are no differences between the
risk attitudes across genders. In the abstract gambling treatment, men are found to be more risk averse
than women in the loss frame and the opposite is found in the gain frame. Using market based data,
Atkinson et al. (2003) find that male- and female-managed funds do not differ significantly in terms
of performance, risk, and other fund characteristics. Their results suggest that differences in
investment behavior often attributed to gender may be related to investment knowledge and wealth
constraints.
9
As previous research examining gender differences in risk preferences does not give clear-cut
results, to explore gender difference in our setting we test the following null hypothesis:
Hypothesis 1: Gender does not affect investors’ re-entry decisions after the first-time bad
experience in the stock market.
Previous literature which examines the relationship between age and risk aversion is mixed
(Ameriks and Zeldes, 2004). Halek and Eisenhauer (2001) find that on average, risk aversion
decreases by age until age 65 and increases thereafter. Bellante and Green (2004) examine relative
risk aversion in portfolio allocation decisions among the elderly population, and find that risk
aversion distinctly decreases as among elderly but find only a modest increase in risk aversion
between the ages of 70 and 90. Bakshi and Chen (1994) on the other hand find that as age increases
the risk premium also increases implying that on average older people are more risk averse than
younger people. However, they also find that as people age their demand for financial assets
increases: older people invest more in financial markets whereas in younger ages housing is a more
urgent need. The authors’ findings jointly imply that although older people invest in financial assets
more than younger ones, they invest less in risky assets such as stocks. Some studies confirm that as
people age they invest less in stocks (Agnew et al., 2003; Holden and Van Derhei, 2001). Others
suggest a different pattern: the share of financial wealth in equities increases over the working life and
then declines or stays flat after the retirement (Poterba and Samwick, 1997; Yoo, 1994). In contrast,
some studies find no evidence supporting a gradual reduction in portfolios comprised of stocks with
age (Ameriks and Zeldes, 2004). In sum, previous literature has mixed evidence on how age impacts
risk aversion and investment behavior. Therefore, our hypothesis, in null form, is as follows:
Hypothesis 2: Age does not affect investors’ re-entry decisions after the first-time bad
experience in the stock market.
10
Frieder and Subrahmanyam (2005) find that individual investors prefer to invest in stocks
with easily recognized products. In line with this, Lou (2014) finds that advertising spending is
associated with individual investor buying and a contemporaneous rise in abnormal stock returns,
which is then reversed in subsequent years suggesting that investors' initial response to changes in
advertising is indeed biased and excessive. In addition, they find that the return pattern is significantly
stronger among firms producing consumer goods (e.g., Apple Computer) than those producing non-
consumer goods (e.g., US Steel), consistent with the intuition that advertising for consumer goods
(e.g., iPod) is more likely to attract consumer/investor attention. Similarly, Billett et al. (2014) find
that individual investors prefer more prestigious brands to less prestigious brands. However, to the
best of our knowledge, the repurchase decision of well-known brands after a bad initial experience is
still an empirical question. Therefore, we explore this question with the following null hypothesis:
Hypothesis 3: Investors are equally likely to re-enter if the first-time bad experience in the
stock market is with a well-known stock such as Nokia.
3. Empirical analysis
3.1. Data and Sample Construction
The transaction data used in this paper is well-established in the literature and identical to that
used in Seru et al. (2010). The data come from the Nordic Central Securities Depository (NCSD) and
cover all trading in all Finnish stocks over a nine-year period. The subset of the same dataset is used
in Grinblatt and Keloharju (2000, 2001a, 2001b).5 The original data contain 62,946,476 transactions
of nearly 1.3 million investors including individual investors and institutional investors spanning a
period from January 1995 to December 2003. We restrict the entries to transactions placed by
individual investors only, which reduce the number of transactions down to 4,965,147 by about 0.3
5 The detailed description on the data can be provided in these references.
11
million retail investors. From this sample, we identify each investor’s first-time stock market
experience and her decision to re-enter the market subsequent to the first experience.
a. First-time stock market experience. The first-time stock market experience is measured
based on investor’s first ever stock investment since she opened the brokerage account in
our sample. In order to do this, we first remove the accounts that existed before the
beginning of our sample period because we do not know if they had been already involved
in any stock trade before. For every investor who opened the account after the beginning
of our sample period, we identify the time when she enters and exits the market. The entry
date is defined as the first date on which she purchases shares of any stock. We exclude
from the sample the investors who purchase multiple stocks on the entering date.6 The exit
date is then identified as the date on which a given investor sells all of her shares of the
stock, that is, she closes the position in full since her first-time entry. We do not consider
for our analysis any investor who never exits market. To make sure that this investor has
kept positions only in the stock they initially purchased, we require them to have not
involved in trades of any other stock between enter and exit date7. Our analysis focuses on
the investors entering the market with a single stock (as opposed to multiple different
stocks) as they represent better inexperienced naïve investors. For the same reason, we
exclude from our analysis the accounts of those investors who ever took a short position,
which is unlikely in case of naïve investors.
6 We consider in the robustness section the case of investors with multiple stocks. 7 Note that investors are allowed to add positions on the same single stock during 7 calendar days following the initial entry date or a period between the initial entry data and the exit date, whichever shorter
12
b. Stock market re-entry. For each investor we define the re-entry date as the first date on
which she purchases any stock within one calendar year (that is, 365 days) since her exit. 8
It ensures that any other transaction does not occur between exit and re-entry date.
Once entry and exit dates are identified, we measure the average (raw) returns of stocks
purchased and sold for the periods between the two dates.9 In doing so, we use actual transaction
prices (adjusted for stock splits) if available, and otherwise we complement the dataset with the
closing prices from Thomson Reuters Datastream. Any intraday trades are netted to obtain a quantity-
weighted price for the daily transaction. In the case of multiple trades occurring for a given stock
between two dates, we assume that the amount of shares purchased first were sold first to compute the
quantity-weighted average returns of multiple trades (Grinblatt and Keloharju, 2000). We only focus
on investors who trade stocks that are available in Datastream, and who exit the market before the end
of 2002 (that is, one year prior to the end date of our transaction database). We do this to avoid the
potential bias where investors may re-enter in 2004 but are categorized as ‘no re-entry’ due to the
truncation of our sample at the end of 2003.
After clearing and merging, our final sample contains 11,543 individuals and 176 different,
publicly-listed Finnish stocks. Panel A of Table 1 presents some investor characteristics in our
sample, and Panel B shows the breakdown of the individual’s first investment by listing up the top 10
most-frequently traded stocks. The frequency of the exit and the re-entry by year is given in Panel C.
The summary statistics of variables used in our analysis are presented in Panel D.
As shown in Panel A, the male accounts are predominant taking 71% of total accounts in the
sample. This is in line with Barber and Odean (2001) documenting with the U.S. brokerage data that
men trade 45% more than women. The median age of investor is 38 and the bulk (84%) of investors is
8 We consider the one-year fixed window to reduce any noise that looking over longer horizon would introduce (e.g., see, Seru et al., 2010). We
also perform our tests without imposing the one-year window, though unreported but available upon request, and our results are largely
unchanged. 9 As a robustness check, we also employ risk-adjusted returns and our results are largely unchanged.
13
aged between 20 and 65. While the elderly people over 65 years old take 7% of the accounts,
interestingly, we have significant number of accounts by underaged (below 20 years old) investors
constituting 9% of all individual accounts. Since in Finland a person must be 18 years old to set up a
brokerage account, it must be a guardian who opened the account, and made any transfer of shares
into or out of the account. Berkman et al. (2014) show the underage accounts make up for 2.5% of
total trading days attributable to the group of the underage investors.
As for the first-time investment performances, about 21% (2,451 accounts) experience a loss
as opposed to 78% (9,006 accounts) with a gain. Much higher frequency of a realized gain compared
to a loss is consistent with well-documented disposition effect, that is, the propensity of investors to
sell assets on which they have experienced gains, and to hold on to assets with which they have
experienced losses (see, Shefrin and Statman, 1985 and Odean, 1998 among others). More
specifically, the distribution of first investment returns is plotted in Figure 1(a), where we have the
bell-shaped return distribution with the mean (0.2956) and standard deviation (0.9554). The average
duration of first-time investments is 209 days (the minimum of 1 day and the maximum of 2,562
days) (Figure 1(b)). It is interesting to see about 10% of investors in our sample stay in the Finnish
stock market for a relatively short period (that is, less than 5 days).10
In Panel B we list the ten most popular stocks chosen by investors for their first-time
investment, and 36% of investors enter the stock market by buying Nokia shares. This is not
surprising since Nokia is by far the largest firm in the Finnish market accounting for 36% of the total
stock market capitalization on average during the sample period (ranging daily from 16% to a high of
64% at one point in 2000). It is also interesting to note that investors tend to herd on particular stocks
such that seven of ten investors in our sample choose one of top ten stocks from the large pool of 237
different stocks. It is shown in Panel C that 53% of investors (6,229 out of 11,543) do not re-enter
10 For the robustness, we conduct our analysis for the sample of investors whose holding periods are longer than five days. Our results remain unchanged.
14
(within a year) and, a large number (2,213 accounts) of investors exit after 30th April in 2000, labeled
as 2000b, which coincides with the burst of dot-com bubble.
[Insert Table 1 here]
[Insert Figure 1 here]
3.2. Empirical Methodology and Result
In this chapter, we describe our empirical analysis and present the main results. For each of
the predictions derived in Section 2, we present the results of a non-parametric univariate analysis
exhibited in a series of figures, followed by a multivariate logit regression analysis including several
controls.
Prediction 1: The existence of the HSE
The hot stove effect (HSE) predicts that investors who have experienced a negative return on
a particular stock will not invest in any other stocks, let alone the same stock they sold for a loss. In
order to investigate this, we relate each investor’s stock market re-entry decision to her past stock
market experience. More specifically, each investor is first categorized according to her first-time
investment performance as either Loss, representing those people who earned a negative return, or
Gain, those who experienced a positive return. Then, we examine for each group the proportion of
investors withdrawing from the stock market by not being engaged in stock trades any more (within
one year) after their first-time investment gain (or loss). We call this proportion ‘the withdrawal rate’.
If the HSE exists in the stock market we should observe a significantly higher withdrawal rate for
Loss investors compared to Gain investors. The difference in withdrawal rates between two groups
can be considered as capturing how strongly individuals are subject to the HSE: the bigger the
difference, the stronger the HSE.
15
As seen in the left panel of Figure 2(a), investors in a Loss group show much higher
likelihood of not returning to the market than their counterparts in a Gain group. About 72.8% of
accounts in Loss are not engaged in any trade for any stock at all within one year since they sold a
particular stock for a loss. Strikingly, we observe a much lower ratio of 48.9% for the Gain group. It
is previously documented with the U.S. retail investor data (e.g., Strahilevitz et al, 2011) that
investors tend to avoid repurchasing the same stock that they sold for a loss. Our finding is, however,
more comprehensive in that investor’s tendency to not reinvest after experiencing a loss is not
restricted to the case of a same stock, and is also carried over to the case of different stocks they never
invested in before. We argue that this finding is consistent with investors being subject to the HSE as
opposed to becoming more risk-seeking after a loss (a preference consistent with the Prospect theory).
In a broader sense, our result is also in a similar vein with Malmendier and Nagel (2011) that
individuals express a lower willingness to take financial risk, and are less likely to participate in the
stock market, especially when they have experienced low real stock-market returns in their lives.
[Insert Figure 2 here]
While our univariate results hint at the existence of the HSE, one would argue that there exist
other characteristics, both individual- and market-level, that could drive our results. In order to
control for various factors that may affect investor’s stock market re-entry decision, we carry out
multivariate analyses by employing the logit regression framework. Our baseline regression is as
follows:
Withdrawali,t+1 = β0 + β1Lossi,t + β2MktReti,t + β3MktVoli,t + β4Yeari,t + β5Durationi,t + 𝜖𝑖,𝑡 (1)
16
The dependent variable is Withdrawal equal to one if an individual neither purchase nor sell any stock
within one year since unwinding her first investment in full, and otherwise zero. Loss is a dummy
variable, of our main interest, equal to one if the first investment position was unwound with a loss,
and otherwise zero. MktRet and MktVol is return and volatility (a standard deviation of daily return)
respectively of OMX Helsinki Index, which are measured for one month period after the investor’s
exit. Year is a year dummy variable with year 1995 a reference level, representing the year in which
an investor exit the market. Due to the dramatic changes in stock market conditions during the year of
2000, two time dummies, 2000a and 2000b are used respectively for the year before and after the
burst of dot-com bubbles (that is, 30th of April). Duration is the length of time in days between entry
and exit date. The latter is an important control to account for a potential disposition effect. To make
our baseline model as parsimonious as possible, we choose to include in the baseline regression in
Equation (1) only the basic set of control variables: stock market conditions when investors exit the
market summarized by its return and volatility, and the holding period of first investments. We also
include year dummies to ensure to remove time trends or any aggregate effects, such as time-varying
aggregate risk aversion.
[Insert Table 2 here]
Table 2 shows main results of our baseline regression. Consistent with our Prediction 1, it is
shown that the coefficient on Loss is positive and statistically highly significant with p-value virtually
zero. Being other things equal, the odds of investors withdrawing (vs. non-withdrawing) from stock
markets increase by more than two times (= 𝑒0.9963 = 2.70) when their first experience is negative.
The coefficients on other variables also show the expected signs. It is interesting to find the positive
coefficient on Duration, which suggests that the longer investors are engaged in their first-time stock
17
investing, the more likely they withdraw from the market. The effect of duration seems robust as it
carries a highly significant coefficient throughout the analyses regardless of the model specification.
It is also interesting to see that stock market conditions at the time of exit are shown to be
relevant to investor’s re-entry decision. For example, the negative, and statistically highly significant,
coefficient on MktVol suggests the likelihood of withdrawal is low if the exit was made earlier during
a turbulent time. In a smilar vein, the coefficients on time dummies for a volatile period of 1999-2000
(Year1999, Year2000a, and Year2000b) are negative with a relatively large magnitude and all highly
significant. This can be interpreted as investors attributing their investment perfomance to luck rather
than to skill in case of a loss, also known as self-serving bias (Heider, 1944, Campbell and Sedikides,
1999; Myers, 2015). Therefore, a negative experience would be less likely to deter them from
reinvesting in stock markets.
Overall, our main results from the baseline regression support our conjecture that once an
investor had a negative experience with a particular single stock, she shows a tendency to cease
trading any of stocks alltogher, let alone the stock they traded earlier.
Prediction 2: The HSE and the Magnitude of Loss
As with the analogy in Mark Twain’s cat, the hotter the stove the cat has sat on, the stronger is
the cat’s unwillingness to sit on any of stoves again, hot or cold. In other words, the worse is the
experience from stock market participation in the first trial the less likely it is that the investor comes
back to the market. To investigate the relation between the HSE and the painfulness of past
investment experience, we first plot in Figure 3 the withdrawal rates by the magnitue of loss or gain.
In doing so, investors are sorted into deciles based on the realized returns (scaled by their volatilites)
they earned for their first-time stock market participation. We adjust the returns with volatilites to
accounts for a preference for volatile stocks. We also control for the duration of the initial investment
18
to capture the disposition effect.11 Due to the relatively small number of accounts who sold a stock for
a loss in our sample, only two of ten deciles (that Decile 1 and Decile 2) are assigned to those in
Loss, Decile 3 is a mixture of accouts in Loss and Gain, and other six (Decile 4 to Decile 10) goes to
accounts in Gain. For instance, Decile 1 represents the group of investors whose Vol-AdjustedReturn
was lower than -5.5% of returns, Decile 2 between -5.5% and -.5%, , and Decile 3 between -.5% and
1.4%. Likewise, Decile 4 between 1.4% and 2.8% while Decile 10 higher than 25.6%.
Interestingly, Figure 3 reveals the U-shape patterns in withdrawal rates across ten investor
groups. The rate is highest for Decile 1 (that is, the group with lowest returns) and decreases with the
returns until Decile 6 and then increase from Decile 7 to Decile 10. Considering the deciles separately
for Loss and Gain, we first see that withdrawal rates are overall higher for Loss than for Gain, which
is in line with our finding in previous section. Second, the rate decreases monotonically with the
magnitude of returns for Loss deciles. This is consistent with our conjecture that the more painful
investors’ first experience the more likely they shun the stock market later. Even though the
experience is limited to a single stock, the pain is severe enough to keep them away from the universe
of stocks. Third, the increasing, though not monotonically, pattern in withdrawal rates with the
magnitude for Gain deciles appears at odds with the HSE, according to which the rate should be
lower for high return deciles. In particular, it is very puzzling to see the highest withdrawal rates
among investors (that is, Decile 9 and Decile 10) having reaped extremely high returns (45% at
minimum) from their earlier investment. Earning a hefty return as high as 40-50% at minimum (for a
80 day holding period on average in the sample) is a very rare event. Presumably, once having
experienced the positive rare event in the market, investors, especially if naïve and unsophisticated,
may become averse to the possibility of another rare event where they end up with a huge negative
return from next investments.
11
We thank an anonymous referee for suggesting these controls.
19
In order to look into this relation, we re-run our baseline regression in Equation (1) with the
addition of new variables, Loss×Magnitude(Loss). Magnitude(Loss) is a continuous variable
measuring the absolute value of negative realized returns on the first investment. Our main finding
from the univariate analysis carries through to the multivariate regression analysis. As presented in
Table 3, the coefficient on Loss×Magnitude is positive and statistically significant (with the p-value
being 0.0232), meaning that the larger the losses on the first investment the less likely that investors
re-enter the market. As an illustration, being other thigs equal, the odds of an investor withdrawing
from the market increase by 1.72 times (= 𝑒0.5468 = 1.72) when the loss is heavier by the magnitude
of 1 (that is, realized return decreases by 100 percentage point).
[Insert Table 3 here]
[Insert Figure 3 here]
Prediction 3: The HSE and Bull vs. Bear Markets
As we notice earlier in Table 2, the time of exit seems to influence an investor’s re-entry
decision: people withdraw less if they exit in a boom period (that is, 1999 and 2000a as in Figure
4(a)). Now we want to give a closer look at how investor’s re-entry decision would be affected by the
time when she exits the market, especially with an investment loss. Our conjecture is that HSE is
stronger (weaker) for those who exit the market in bear (bull) periods. We examine how withdrawal
rates differ by year of exit as presented in Figure 4(b). The most important observation from the graph
is that withdrawal rates do not show much differences between Loss and Gain groups until the year of
2000a. But the gap becomes significantly wider for the later periods (that is, 2000b, 2001, and 2002).
In other words, investors are more likely to withdraw from the market if it is in the early 2000’s when
they realized losses than it is in the late 1990’s.
20
It is important to note that there is a dramatic change in stock market conditions in Finland
(see Figure 4(a)) before and after 30th of April in 2000, which coincides with the burst of dot-com
bubble in the US. To see a clearer picture, we divide investors into two groups with one having exited
in a bull period (that is, before 30th of April in 2000), and the other in a bear period (that is, after 30th
of April in 2000). Then we compare the withdrawal rates for two groups as in Figure 4(c). It is clearly
seen that the HSE is pronounced among exit-in-bear investors: as high as 80% of them with a loss (vs.
45% with a gain) never come back to stock markets again. This comes with a big contrast to the case
of exit-in-bull investors whose withdrawal rate after a loss is merely 50%.
Why would the HSE be less pronounced for those investors who leave stock markets in the
bull period? Our explanation is that witnessing the stock market run-ups during the period of 1997-
1999 would lead to an optimistic view on the stock market, even though the investors themselves had
a negative experience. This optimism would diminish the pain from the earlier (negative) experience,
and make investors less likely to withdraw from the market. On the other hand, investors form a
pessimistic view on the market in the bear period, which reinforces their tendency to shun the market
after the negative experience.
We modify the baseline logit regression specification by introducing the interaction term
between Loss and bear (vs. bull) market dummy variable, Exit-in-Bear, being equal to one if an
investor fully unwinds the first investment before 30th of April in 2000, and otherwise zero. Table 4
presents a clear result where Loss×Exit-in-Bear has a positive coefficient of 0.5282, which is also
statistically highly significant. Its economic magnitude is also important: the withdrawal rate after
experiencing a loss is 1.7 times (= 𝑒0.5282 = 1.70) higher for exit-in-bear investors than their exit-in-
bull counterparts. Overall, our finding is consistent with Prediction 3 stating the HSE are stronger
(weaker) for those who exit the market in bear (bull) periods.
[Insert Table 4 here]
21
[Insert Figure 4 here]
Prediction 4a: The HSE and Investor Gender
Gender is known to be one of the important factors in affecting stock market entry decisions
(e.g., Barber and Odean, 2001, Croson and Gneezy, 2009), so it is conceivable that the gender also
affects the re-entry decision. The preliminary result in Figure 2(b), however, shows that gender
appears to have no significant influence on the HSE. We observe a higher withdrawal rate after Loss
by about 25 percentage points than after Gain regardless of investor’s gender. Focusing only on a
Loss group, we have a higher withdrawal rate for women than for men, which is in line with the
finding of Barber and Odean (2001) that men trade more excessively than women.
To test the effect of gender, we include in our baseline regression the additional variable,
Female, that is, the dummy variable equals to one if women and zero otherwise. As shown in Panel A
of Table 5, Female is seen to have a positive and statistically significant coefficient, indicating
women are less likely to return to the stock market than men regardless of the performance of their
initial investment. This result is in line with the literature of women trading less excessively. The
coefficient on the interaction term, Loss×Female, however, is not statistically significant at all, which
confirms that investor gender has no significant influence on the HSE.
Prediction 4b: The HSE and Investor Age
Investor’s attitude for risk-taking may change over their life cycle (e.g., Bakshi and Chen,
1994) and therefore one might expect that the HSE varies with age. Figure 5(a) presents the
withdrawal rates for Loss and Gain by different age groups, and the differences in rates between Loss
and Gain are plotted for each group in Figure 5(b). Figure 5(a) shows that withdrawal rates for Loss
investors increase with investor age whereas the rates for Gain are largely unchanged for different age
groups. Figure 5(b) provides a clearer picture for which the withdrawal rate of investors over the age
22
of 65 after Loss is about 30 percentage points higher than the rate after Gain. This difference,
however, becomes smaller for the young below 20 years old and is only 20 percentage points. This
result would imply that the older are more prone to the HSE. It is very interesting to see the youngest
group (that is, the age of 20-) is least subject to the HSE. This may be because the underaged accounts
in Finland are maintained by guardians of the underage, who are supposedly informed and financially
more sophisticated (Berkman et al., 2014).
Next we add the age variables to the baseline regression in Equation (1) to conduct
multivariate analysis. In Panel B of Table 5, Age is a categorical variable with four levels (0-19, 20-
39, 40-64, and 65+). The bin consisting of accounts at age 0-19 is the reference level. It is clearly seen
that the magnitude of coefficient on the interation term, Loss×Age, increases monotonically with the
age, indicating the oldest (youngest) is most (least) likely to withdraw from the market after
experiencing a loss. The tendency of the oldest, in particular, not returning to the market is also
statistically significant. As an illustration, investors with age of 65+ are 1.7 times (= 𝑒0.5390 =
1.71) more likely to withdraw from stock markets than underaged investors after they exited the
market with an investment loss. This result is not likely to be driven by the death of investors because
we examine their re-entry within one year since their exit. Overall these results suggest that older
people tend to be more affected by the HSE.
[Insert Table 5 here]
[Insert Figure 5 here]
Prediction 5: The HSE and Brand Stocks
Billett et al. (2014) find that individual investors prefer more prestigious brands to less
prestigious brands. If we observe the same pattern in the context of the stock market, investors may
be less scarred by a negative experience if it is with brand stocks, and less likely to withdraw from the
23
market. In this vein, we predict that the HSE would be weaker if investor’s past experience involves
brand stocks. In examining this relation, we use the Nokia stock as representing the brand stock in
Finland. This is conceivable because of the unique feature of Finnish stock market, unlike others, that
Nokia was undoubtedly the most famous brand especially during our sample period. Nokia was by far
the largest firm in the Finnish market accounting for 36% of the total stock market capitalization, and
36% of individual investors initiated their stock trading by buying Nokia. As shown in Table 6 with a
new dummy variable, Nokia, equal to one if the first investment was on Nokia and otherwise zero,
the variable of our interest, Loss× Nokia, is shown to have a coefficient of expected negative sign.
Though the significance is low, it is still interesting to see that investor’s re-entry decision is
somehow linked to the brand awareness of the stock. Note that the variable Nokia itself, however,
carries a highly significant negative coefficient.
[Insert Table 6 here]
3.3. Robustness
3.3.1. Investor location
According to Viceira (2001)’s model, especially those households whose labor income is
negatively correlated with stock market returns should invest in the market. Given that the majority of
the companies in the Finnish stock market are located in Helsinki, one can argue that the investor
location could a good proxy for the correlation between labor income and stock market returns. So,
the labor income of individuals living in Helsinki is presumably more correlated with stock market
returns. Therefore, according to the model, the Finnish stock market should be more appealing for the
people living outside Helsinki. In this section, we want to check whether our results would be
different depending on where the investor is located. Based on this conjecture, we introduce a new
24
dummy variable, Helsinki, being equal to one for investor living in Helsinki and zero otherwise. Table
7 reveals two interesting results: (i) the HSE is still prevalent after controllong for investor location,
and (ii) investors living in Helsinki are less subject to the HSE (based on the coefficient on Loss×
Helsinki being negative and marginally significant). To mitigate the concern on the selection bias, we
also employ the matching methodology (see Appendix 2) and compare two groups of investors in
Figure 6(a). Our main result carries through, and we see that investors living outside Helsinki are
more subject to the HSE, this is precisely the group of investors who should consider the stock market
for their retirement consumption.
[Insert Table 7 here]
[Insert Figure 6 here]
3.3.2. Nokia effects
Given that Nokia is a predominant entry in the sample, one may argue that our results could
be firm-specific and may not hold in general. Therefore in this section we check the robustness of our
results for the sample excluding the Nokia shares. In Table 8, we exclude all accounts who invested in
Nokia for their first investments, and re-run the baseline regression in Equation (1). The main results
remain unchanged and thus it is unlikely that our findings are driven by a particular stock.
[Insert Table 8 here]
3.3.3. Multiple-stock entries
25
In the main analysis, we focus on investors who once enter and exit the market by trading
only a single stock. Even though we intend to understand the re-entry decision of naïve investors
(e.g., household investors), one criticism may be that our results cannot be generalised to a broader
class of investors, e.g., financially more sophisticated market participants. Hence in this section we
perform a robustness test for our results and include investors entering and exiting the stock market
by trading a portfolio of stocks (as opposed to a single stock).12 In doing so, we define a new dummy
variable, Multiple, being equal to one for a multiple-stock entry investor and zero for a single stock
one. Our main results are unchanged after enlarging our sample. Interestingly, we have a positive
coefficient on the interaction variable, Loss× Multiple, suggesting that investors who trade a portfolio
of stocks are more subject to the HSE than those who trade a single stock. This may be due to the fact
that investment failure comes more painful when investors would think that trading a portfolio is a
less risky and correct way of investment. To mitigate the selection bias, we construct a matched
sample (see Appendix 2), and compare investors for the two cases (that is, trading a single stock vs. a
portfolio of multiple stocks) in Figure 6(b). Matching sample results confirm that the multiple stock
investors are more subject to the HSE.
[Insert Table 9 here]
4. Conclusion
In this paper, we document that complete withdrawal from the stock market is more likely after an
initial first-time loss. Using a detailed dataset which contains all individual transactions of the market
participants from the Finnish stock market, we show that the HSE is strong even after controlling for
market conditions, time-fixed effects and the duration of the first-time investment. The magnitude of
the first-time loss exacerbates the HSE. Moreover, the broad market performance (bear vs. bull)
12
For the detail on the sample construction for the multiple stock investors, see Appendix 1.
26
during the stock market exit has a strong impact on the re-entry decision. While the gender does not
play a role in the context of the HSE, investor age is positively correlated with the HSE, that is, the
older the investor the more likely that she will withdraw from the market entirely after an initial loss.
Our results do not depend on the number of stocks in the initial portfolio, investor location or the first
stock being Nokia.
There is scant evidence on the re-entry decision of inexperienced investors. Our paper is an initial
attempt to fill the gap. The existence of the HSE among the first-time investors, and thus the fact that
inexperienced investors largely do not exhibit gambling behavior has important implications on
incentives for stock market participation, and contributes to the discussion on financial literacy; if the
re-entry decision heavily depends on the performance of the limited first-time experience in the stock
market, then investors should be informed about risks involved in investing stocks in the short-run,
and encouraged to consider the stock market for long-run investment horizon. Neither trading
excessively nor withdrawing completely allow the stock markets to serve as a vehicle to transfer
consumption to the period when is most valued, that is, during the retirement period.
27
References
Allen, F., and D. Gale. 1994. Limited Market Participation and Volatility of Asset Prices. The
American Economic Review. 84(4): 933-955.
Atkinson, S. M., S. B. Baird and M. B. Frye. 2003. Do Female Mutual Fund Managers Manage
Differently?. Journal of Financial Research. 26 (1): 1-18.
Bakshi, G. S. and Z. Chen. 1994. Baby Boom, Population Aging, and Capital Markets. Journal of
Business. 67(2): 163-202.
Bali, T. G., Demirtas, K. O., Levy, H., and A. Wolf. 2009. Bonds Versus Stocks: Investors’ Age And
Risk Taking. Journal of Monetary Economics. 56(6): 817-830
Bange, M. M. 2000. Do The Portfolios Of Small Investors Reflect Positive Feedback Trading?
Journal of Financial and Quantitative Analysis. 35(2): 239-255.
Barber, B. M., and T. Odean. 2001. Boys Will Be Boys: Gender, Overconfidence, and Common
Stock Investment. Quarterly Journal of Economics. 116(1): 261-292.
Barberis, N., and R. Thaler. 2003. A Survey Of Behavioral Finance. Handbook of the Economics of
Finance. 1: 1053-1128.
Barberis, N. 2000). Investing For The Long Run When Returns Are Predictable. Journal of finance.
55(1): 225-264.
Bellante, D. and C. A. Green. 2004. Relative Risk Aversion Among The Elderly. Review of Financial
Economics. 13(3): 269-281.
Berkman, H., P. D. Koch, and P. J. Westerholm. 2014. Informed Trading Through The Accounts Of
Children. Journal of Finance. 69(1): 363-404.
Billett, M. T., Z. Jiang and L. L. Rego. 2014. Glamour Brands And Glamour Stocks. Journal of
Economic Behavior & Organization. 107(B): 744-759.
Bodie, Z. 1995. On The Risk Of Stocks In The Long Run. Financial Analysts Journal. 51(3), 18-22.
28
Bodie, Z. 1976. Common Stocks As A Hedge Against Inflation. The Journal of Finance. 31(2), 459-
470.
Brav, A., Constantinides, G. M., and C. C. Geczy. 2002. Asset Pricing With Heterogeneous
Consumers And Limited Participation: Empirical Evidence. NBER Working Paper.
Calvet, L. E., J. Y. Campbell, and P. Sodini. 2007. Down or out: Asssessing the Welfare Costs of
Household Investment Mistakes, Journal of Political Economy. 115(5), 707-747.
Campbell, J. Y., and L. M. Viceira. 2005. Strategic asset allocation for pension plans, Oxford
Handbook Of Pensions And Retirement Income, Oxford.
Campbell, W.K., and C. Sedikides. 1999. Self-threat Magnifies The Self-Serving Bias: A Meta-
Analytic Integration. Review of General Psychology. 3(1): 23-43.
Choi, J. J., D. Laibson, B. C. Madrian, and A. Metrick. 2009. Reinforcement Learning and Saving
Behavior. Journal of Finance. 64(6): 2515-2534.
Croson, R. and R. Gneezy. 2009. Gender Differences in Preferences. Journal of Economic Literature.
47(2): 448-474.
De Bondt, W. P. 1993. Betting On Trends: Intuitive Forecasts Of Financial Risk And Return.
International Journal of forecasting. 9(3): 355-371.
Denrell, J. and J.G. March. 2001. Adaptation As Information Restriction: The Hot Stove Effect.
Organization Science. 12(5): 523-538.
Eckel, C. C. and P. J. Grossman. 2008. Men, Women And Risk Aversion: Experimental Evidence.
Handbook of experimental economics results. 1(113):1061-1073.
Frieder, L. and A. Subrahmanyam. 2005. Brand Perceptions And The Market For Common Stock.
Journal of financial and Quantitative Analysis. 40(01): 57-85.
Fujikawa, T. 2009. On The Relative Importance Of The Hot Stove Effect And The Tendency To Rely
On Small Samples. Judgement and Decision Making. 4(5), 429–435.
Goetzmann, W. N., and A. Kumar. 2008. Equity Portfolio Diversification. Review of Finance. 12(3),
29
433-463.
Grinblatt, M., and M. Keloharju. 2000. The Investment Behavior and Performance of Various
Investor Types: A Study of Finland’s Unique Data Set. Journal of Financial Economics.
55(1):43–67.
Grinblatt, M., and M. Keloharju. 2001a. How Distance, Language, and Culture Influence
Stockholdings and Trades. Journal of Finance. 56(3):1053–73.
Grinblatt, M., and M. Keloharju. 2001b. What Makes Investors Trade? Journal of Finance.
56(2):589–616.
Guiso, L., Sapienza, P., and L. Zingales. 2008. Trusting The Stock Market. Journal of Finance. 63(6):
2557-2600.
Guiso, L., and T. Jappelli. 2005. Awareness And Stock Market Participation. Review of Finance. 9(4):
537-567.
Guvenen, F. 2009. A Parsimonious Macroeconomic Model For Asset Pricing. Econometrica. 77(6):
1711-1750.
Halek, M. and J. G. Eisenhauer. 2001. Demography Of Risk Aversion. Journal of Risk and
Insurance. 68(1):1-24.
Heider, F. 1944. Social Perception And Phenomenal Causality. Psychological Review. 51(6):358-374.
Holden, S. and J. VanDerhei. 2001. 401 (k) Plan Asset Allocation, Account Balances, And Loan
Activity In 1999. SSRN Working Paper.
Holt, C. A., and S.K. Laury. 2002. Risk Aversion And Incentive Effects. American Economic Review.
92(5):1644-1655.
Holland, J. H. 1975. Adaptation In Natural And Artificial System: An Introduction With Application
To Biology, Control And Artificial Intelligence, University of Michigan Press.
Huang, X. 2015. Mark Twain’s Cat: Industry Investment Experience, Categorical Thinking and Stock
Selection. Working Paper.
30
Kahneman D. and A. Tversky. 1979. Prospect Theory: An Analysis Of Decision Under Risk.
Econometrica. 47(2):263–292.
Kausta, M. and S. Knupfer. 2008. Do Investors Overweight Personal Experience? Evidence from
IPO Subscriptions. Journal of Finance. 63(6):2679-2702.
Kausta, M. and S. Knupfer. 2012. Peer Performance And Stock Market Entry. Journal of Financial
Economics. 104(2):321-338.
Kim, S., and F. In. 2005. The Relationship Between Stock Returns And Inflation: New Evidence
From Wavelet Analysis. Journal of Empirical Finance. 12(3):435-444.
Levy, H. 2015. Aging Population, Retirement, and Risk Taking. Management Science. Forthcoming.
Lou, D. 2014. Attracting Investor Attention Through Advertising. Review of Financial Studies.
Forthcoming.
Malmendier, U. and S. Nagel. 2011. Depression Babies: Do Macroeconomic Experiences Affect
Risk-Taking? Quarterly Journal of Economics. 126(1):373-416.
Myers, D.G. 2015. Exploring Social Psychology, New York: McGraw Hill Education.
Odean, T. 1998. Are Investors Reluctant to Realize Their Losses? Journal of Finance. 53(5):1775-
1798.
Pástor, Ľ., and R. F. Stambaugh. 2012. Are Stocks Really Less Volatile In The Long Run? Journal of
Finance. 67(2): 431-478.
Poterba, J. M. and A. A. Samwick. 1997. Household Portfolio Allocation Over The Life Cycle.
NBER Working Paper.
Schotman, P. C., and M. Schweitzer. 2000. Horizon Sensitivity Of The Inflation Hedge Of
Stocks. Journal of Empirical Finance. 7(3):301-315.
Schubert, R., M. Brown, M. Gysler and H. W. Brachinger. 1999. Financial Decision-Making: Are
Women Really More Risk-Averse? American Economic Review. 89(2):381-385.
31
Seru, A., Shumway, T., and N. Stoffman. 2010. Learning By Trading. Review of Financial Studies. 23
(2):705-739.
Shefrin, H., and M. Statman. 1985. The Disposition To Sell Winners Too Early And Ride Losers Too
Long: Theory and Evidence. Journal of Finance. 40(3):777-790.
Siegel, J. 2014. Stocks For The Long Run, McGraw-Hill Companies.
Strahilevitz, M. A., Odean, T., and B.M. Barber. 2011. Once Burned, Twice Shy: How Naive
Learning, Counterfactuals, and Regret Affect the Repurchase of Stocks Previously Sold.
Journal of Marketing Research. 48:102-120.
Twain, M. 1897. Following The Equator: A Journey Around the World. American Publishing Co.,
Hartford, C.T.
Schubert, R., M. Brown, M. Gysler and H. W. Brachinger. 1999. Financial Decision-Making: Are
Women Really More Risk-Averse? American Economic Review. 89(2):381-385.
Van Rooij, M., Lusardi, A., and R. Alessie. 2011. Financial Literacy And Stock Market
Participation. Journal of Financial Economics. 101(2):449-472.
Viceira, L. M. 2001. Optimal Portfolio Choice For Long-Horizon Investors With Non-tradable Labor
Income. Journal of Finance. 56(2):433-470.
Viceira, L. M. 2008. Life-Cycle Funds, In Annamaria Lusardi, ed., Overcoming the Saving Slump:
How to Increase the Effectiveness of Financial Education and Saving Programs.
Vissing-Jorgensen, A. 2002. Limited Asset Market Participation And The Elasticity Of Intertemporal
Substitution. NBER Working Paper.
32
Appendix:
1. The entry and exit date for multiple stock investors
In this section, we describe how to measure the entry and exit date for investors who initiate their
trading with more than a single stock. In doing so, we define the initial investment window for a given
investor as a period between the first buying trade and the first selling trade (excluding the first
selling date) or 7 calendar days since the first buying trade, whichever shorter (see Figure A.1). The
exit date is defined as the date on which a given investor had no position on any stocks in a portfolio
for the first time since the first entry. The reentry date is defined as the date on which a given investor
bought one or more stock for the first time since the exit. Figure A.1 illustrates the method:
Figure A.1. Illustration of entry and exit dates for multiple stock investors
33
2. Construction of matched investor samples: (i) Helsinki vs. Non-Helsinki and (ii)
Single Stock vs. Multiple Stocks
We matched 2,828 accounts living in Helsinki (Helsinki subset) with 8,715 accounts living
outside Helsinki (non-Helsinki subset), and matched 1,105 accounts who bought multiple stocks
during the initial investment window with 11,543 accounts who bought a single stock during the
window. In order to construct those matched samples, we used the one-to-one matching method
similar to Huang and Stoll (1996). For an account i in the Helsinki (Multiple-stocks) subset and an
account j in the non-Helsinki (Single-stock) subset, the distance between i and j is calculated as:
Distance(i,j) = ∑ [2(𝑥𝑖
𝑘− 𝑥𝑗𝑘)
𝑥𝑖𝑘− 𝑥𝑗
𝑘 ]2
𝑘 , where 𝑥𝑘 is standardized variable to be matched (2)
Variables were standardised by subtracting the minimum value and being divided by the range of
the variable in the subset such that the standardised variables were in the range between 0 and 1. All
matched pairs were kept regardless the matched distance between them. The procedure is as follows:
1. Randomly sample an account i from Helsinki (Multiple-stocks) subset.
2. Identify gender and exit-year (and Helsinki for matching the single-stock subset with
the multiple-stocks subset) of account i.
3. Extract accounts in the non-Helsinki (Single-stock) subset who have the same gender
and exit-year as account i (candidate accounts).
4. Calculate the distances based on Age and Days between the window-end and the exit
between account i and each candidate account according to Formula 1.
5. Find an account j in the candidate accounts whose distance to account i is the smallest.
6. Exclude account i from the Helsinki (Multiple-stocks) subset and account j from the
non-Helsinki (Single-stock) subset.
34
7. Repeat step 1 to 6 until all accounts in the Helsinki (Multiple-stock) subset are
matched to accounts in the non-Helsinki (Single-stock) subset.
In order to minimize the influence of the order of sampling, we iterated the procedure by 100 time
and chose the best matched samples on the basis of aggregated distances of each matched pairs. Note
that the range of aggregated distances of 100 iterations was [57.48, 61.12] for Helsinki-Non Helsinki
matching, and [27.42:28.15] for Single-Multiple matching.
35
Figure 1. First-time Stock Market Investment
(a) Realized raw return
This graph presents the distribution of realized returns on first-time investment. The y-axis represents the
number of investors and the x-axis is the rate of return. The width of each bar is .05. The red dotted lines
represent the boundaries between return deciles.
(b) Duration
This graph plots the distribution of durations of first-time stock investment with the y-axis representing the
number of investors, and the x-axis being the number of days between enter and exit date. The width of each
bar is 5 days.
36
Figure 2. Investment Profit and Investor Withdrawal Rate
(a) Investment profit and investor withdrawal rate: whole sample
This graph presents the withdrawal rate of investors since their first-time investment. The y-axis represents
the withdrawal rate, and the x-axis is the profit on the first-time stock investment. The bars represent the
withdrawal rate and error bars are corresponding 95% confidence intervals computed by the bootstrap
(1,000 resamples).
(b) Investment profit and investor withdrawal rate: subsample by gender
This graph presents the withdrawal rate of male investors vs. female investors since their first-time
investment. The y-axis represents the withdrawal rate, and the x-axis is the profit on the first-time stock
investment. The bars represent the withdrawal rate and error bars are corresponding 95% confidence
intervals computed by the bootstrap (1,000 resamples).
37
Figure 3. Investor Withdrawal Rate and Magnitude of Investment Profit
This graph presents the withdrawal rate of investors for different magnitudes of investment profits. The y-
axis represents the withdrawal rate predicted by a logit model, and the x-axis is the deciles of return
portfolio. The portfolio number in x-axis is in order of the volatility adjusted returns after controlling for the
duration of initial holding period: Decile 1: Less than -5.5, Decile 2: -5.5~-.5, Decile 3: -.5~1.4, Decile 4:
1.4~2.8, Decile 5: 2.8~4.5, Decile 6: 4.5~6.5, Docile 7: 6.5~9.2, Decile 8: 9.2~14.1, Decile 9: 14.1~.25.6,
Decile 10: More than 25.6. The bars represent the proportion of accounts who did not reenter the market.
Error bars are corresponding 95% confidence intervals.
38
Figure 4. Investor Withdrawal Rate and Bull vs. Bear Markets
(a) Finnish stock market prices
This graph plots the movement of Finnish stock market index price from January 1995 to January 2003. The
y-axis is HEX closing price and x-axis is the date.
(b) Investor withdrawal and the exit year
The bars represent the withdrawal rate for different years. For each year, two bars are plotted for Loss and Gain
separately. Error bars are corresponding 95% confidence intervals computed by the bootstrap (1,000 resamples).
39
(c) Investor withdrawal and the exit period: bull vs. bear
The bars represent the withdrawal rate. For each group of investors (Exit-in-bull or Exit-in-bear), two bars
are separately plotted for Loss and Gain. Error bars are corresponding 95% confidence intervals computed
by the bootstrap (1,000 resamples).
40
Figure 5. Investor Withdrawal Rate and Investor Age
(a) Investor withdrawal rate for different age group
The bars represent the withdrawal rate. For each age group, two bars are separately plotted for Loss and
Gain. Error bars are corresponding 95% confidence intervals computed by the bootstrap (1,000 resamples).
(b) Difference in investor withdrawal between loss and gain for different age groups
The bars represent the difference between Loss and Gain investors in the withdrawal rate for different age
groups. Error bars are corresponding 95% confidence intervals computed by the bootstrap (1,000
resamples).
41
Figure 6. Matched Sample Results; (i) Helsinki vs. Non-Helsinki, and
(ii) Single-stock vs. Multiple-stock
(a) Investor withdrawal rate for matched sample: Helsinki vs. non-Helsinki
The bars represent the withdrawal rate for investors living in Helsinki vs. investors living outside Helsinki.
Two samples are matched based on age, gender, and investment duration. Error bars are corresponding 95%
confidence intervals computed by the bootstrap (1,000 resamples).
(b) Investor withdrawal rate for matched sample: single vs. multiple
The bars represent the withdrawal rate for investors trading a single stock vs. investors trading a portfolio of
multiple stocks. Two samples are matched based on age, gender, and investment duration. Error bars are
corresponding 95% confidence intervals computed by the bootstrap (1,000 resamples).
42
Table 1. Descriptive Statistics
This table reports the descriptive statistics of the sample used in the analysis. Panel A presents some characteristics
of investors such as age and gender. First-time investment profit is defined as loss if the realized return is negative,
gain if positive, and zero otherwise. Panel B shows the list of top 10 stock names that are most commonly invested
on by investors who initiate their first-time stock trading. Panel C list the years of exit and re-entry and summary
statistics on main variables are presented in Panel D. Withdrawal equal to one if an individual neither purchase nor
sell any stock within one year since unwinding her first investment in full, and otherwise zero. Loss is a dummy
variable equal to one if first investment position was unwound with a loss, and otherwise zero. MktRet and MktVol is
return and volatility (a standard deviation of daily return) respectively of OMX Helsinki Index, which are measured
for one month period after the investor’s exit. Duration is the length of time in days between entry and exit date.
Female is the dummy variable equal to one if women and zero if men. Age is the continuois variable for investor
age, Nokia is equal to 1 if investor’s first investment in on Nokia stock, zero otherwise, Helsinki is equal to 1 if
investor lives in Helsinki, zero otherwise. Duration(exit-rentry) is the length of time in days between exit and re-
entry date. First time return is the rate of return on investor’s first investment, First time return(loss) is the rate of
return for Loss group investors and First time return(gain) is the rate of return for Gain group investors. Exit-in-
Bear, being equal to one if an account fully unwound the first investment before 30th
of April in 2000, and otherwise
zero.
Panel A. Investor characteristics
No. of Accounts Proportion
Gender
Male 8,148 71%
Female 3,395 29%
Age
0-19 1,082 9%
20-39 5,411 47%
40-64 4,258 37%
65- 792 7%
First-time investment
Profit
Loss 2,451 21%
Zero 86 1%
Gain 9,006 78%
Total 11,543 100%
Panel B. Breakdown of the first-time investment (Top 10)
ISIN Company Name Num. accounts Proportion Cumulative proportion
FI0009000681 NOKIA CORP 4,125 0.36 0.36
FI0009007371 SONERA OYJ 1,143 0.10 0.46
FI0009007264 BITTIUM CORP 516 0.04 0.50
FI0009002943 RAISIO PLC 454 0.04 0.54
FI0009000053 MERITA LTD 378 0.03 0.57
FI0009005987 UPM-KYMMENE CORP 362 0.03 0.60
FI0009801310 F-SECURE CORP 243 0.02 0.63
FI0009900070 HARTWALL OYJ 219 0.02 0.64
FI0009006738 ELCOTEQ SE 213 0.02 0.66
FI0009005961 FORTUM OYJ 197 0.02 0.68
Others 3,693 0.32 1.00
Total 11,543
43
Panel C. Exit and Re-entry by Year
Panel D. Summary Statistics of Variables
N Mean Std Dev Minimum Maximum
Withdrawal 11,543 0.5398 0.4984 0.0000 1.0000
Loss 11,543 0.2125 0.4091 0.0000 1.0000
MktRet 11,543 0.0101 0.1188 -0.3196 0.3940
MktVol 11,543 0.0257 0.0104 0.0055 0.0521
Duration 11,543 209.4173 288.0535 1.0000 2,652.0000
Female 11,543 0.2943 0.4557 0.0000 1.0000
Age 11,543 38.4988 16.2457 0.0000 105.0000
First-time return 11,543 0.2956 0.9554 -0.9849 29.0000
First-time return(loss) 2,451 -0.2664 0.2443 -0.9849 -0.0000
First-time return(gain) 9,006 0.4472 1.0172 0.0000 29.0000
Exit-in-bear 11,543 0.4984 0.5000 0.0000 1.0000
Nokia 11,543 0.3578 0.4794 0.0000 1.0000
Helsinki 11,543 0.2447 0.4300 0.0000 1.0000
Duration(exit-re-entry) 6,229 190.8208 353.4944 1.0000 3,146.0000
1995 1996 1997 1998 1999 2000a 2000b 2001 2002 2003
1995 30 8 - - - - - - - - 59 97
1996 85 35 - - - - - - - 375 495
1997 149 40 - - - - - - 394 583
1998 275 123 - - - - - 585 983
1999 621 356 89 - - - 957 2,023
2000a 708 243 15 - - 632 1,598
2000b 879 281 - - 1,053 2,213
2001 691 162 - 1,179 2,032
2002 430 94 995 1,519
30 93 184 315 744 1,064 1,211 987 592 94 6,229 11,543
No
re-entry
Sub total
Sub totalYear of the re-entry (within a year)
Year
of
the e
xit
44
Table 2. HSE: Baseline Regression
This table reports the estimated coefficients, standard errors, and p-values for the following logit regression:
Withdrawali,t+1 = β0 + β1Lossi,t + β2MktReti,t + β3MktVoli,t + β4Yeari,t + β5Durationi,t + 𝜖𝑖,𝑡
The dependent variable, Withdrawal is equal to one if an individual neither purchase nor sell any stock within one
year since unwinding her first investment in full, and otherwise zero. Loss is a dummy variable equal to one if first
investment position was unwound with a loss, and otherwise zero. MktRet and MktVol is return and volatility (a
standard deviation of daily return) respectively of OMX Helsinki Index, which are measured for one month period
after the investor’s exit. Duration is the length of time in days between entry and exit date. Year is a year dummy
variable with year 1995 a reference level, representing the year in which an investor exit the market. Due to the
dramatic changes in stock market conditions during the year of 2000, two time dummies, 2000a and 2000b are used
respectively for the year before and after the busrt of dot-com bubbles (that is, 30th
of April). ∗∗∗, ∗∗, and ∗ denote
statistical significance at 1%, 5%, and 10%.
Estimate
Standard
Error p-Value
Loss*** 0.9963 0.0537 <.0001
MktRet -0.1851 0.1985 0.3511
MktVol** -6.4130 2.7644 0.0203
Duration*** 0.0028 0.0001 <.0001
Year1996*** 0.6407 0.2395 0.0075
Year1997 0.0273 0.2334 0.9068
Year1998 -0.0457 0.2249 0.8388
Year1999** -0.5620 0.2216 0.0112
Year2000a*** -0.7266 0.2255 0.0013
Year2000b** -0.4938 0.2246 0.0279
Year2001 -0.3403 0.2231 0.1271
Year2002* -0.3855 0.2240 0.0852
Observation 11,543
45
Table 3. HSE and Magnitude of Loss
This table reports the estimated coefficients, standard errors, and p-values for the following logit regression:
Withdrawali,t+1 = β0 + β1Lossi,t +β1Lossi,t×Magnitude(Loss) i,t
+ β3MktReti,t + β4MktVoli,t + β5Yeari,t + β6Durationi,t + 𝜖𝑖,𝑡
The dependent variable, Withdrawal is equal to one if an individual neither purchase nor sell any stock within one
year since unwinding her first investment in full, and otherwise zero. Loss is a dummy variable equal to one if first
investment position was unwound with a loss, and otherwise zero. Magnitude(Loss) is a continuous variable
measuring the absolute value of negative realized returns on the first investment. MktRet and MktVol is return and
volatility (a standard deviation of daily return) respectively of OMX Helsinki Index, which are measured for one
month period after the investor’s exit. Duration is the length of time in days between entry and exit date. Year is a
year dummy variable with year 1995 a reference level, representing the year in which an investor exit the market.
Due to the dramatic changes in stock market conditions during the year of 2000, two time dummies, 2000a and
2000b are used respectively for the year before and after the busrt of dot-com bubbles (that is, 30th
of April). ∗∗∗, ∗∗,
and ∗ denote statistical significance at 1%, 5%, and 10%.
Estimate
Standard
Error p-Value
Loss*** 0.8785 0.0739 <.0001
Loss×Magnitude(Loss)** 0.5468 0.2408 0.0232
MktRet -0.1983 0.1987 0.3181
MktVol** -6.5837 2.7650 0.0173
Duration*** 0.0028 0.0001 <.0001
Year1996*** 0.6349 0.2392 0.0079
Year1997 0.0273 0.2330 0.9066
Year1998 -0.0487 0.2246 0.8283
Year1999** -0.5611 0.2212 0.0112
Year2000a*** -0.7294 0.2252 0.0012
Year2000b** -0.5019 0.2243 0.0252
Year2001 -0.3549 0.2228 0.1112
Year2002* -0.3964 0.2237 0.0764
Observation 11,543
46
Table 4. HSE and Time of Exit: Bull vs. Bear Markets
This table reports the estimated coefficients, standard errors, and p-values for the following logit regression:
Withdrawali,t+1 = β0 + β1Lossi,t +β1Lossi,t×Exit-in-Bear i,t
+ β3MktReti,t + β4MktVoli,t + β5Exit-in-Beari,t + β6Durationi,t + 𝜖𝑖,𝑡
The dependent variable, Withdrawal is equal to one if an individual neither purchase nor sell any stock within one
year since unwinding her first investment in full, and otherwise zero. Loss is a dummy variable equal to one if first
investment position was unwound with a loss, and otherwise zero. Magnitude(Loss) is a continuous variable
measuring the absolute value of negative realized returns on the first investment. MktRet and MktVol is return and
volatility (a standard deviation of daily return) respectively of OMX Helsinki Index, which are measured for one
month period after the investor’s exit. Duration is the length of time in days between entry and exit date. Year is a
year dummy variable with year 1995 a reference level, representing the year in which an investor exit the market.
Due to the dramatic changes in stock market conditions during the year of 2000, two time dummies, 2000a and
2000b are used respectively for the year before and after the busrt of dot-com bubbles (that is, 30th
of April). Exit-in-
Bear, being equal to one if an account fully unwound the first investment before 30th
of April in 2000, and otherwise
zero. ∗∗∗, ∗∗, and ∗ denote statistical significance at 1%, 5%, and 10%.
Estimate
Standard
Error p-Value
Loss*** 0.6767 0.0857 <.0001
Loss×Exit-in-Bear*** 0.5282 0.1098 <.0001
Exit-in-Bear -0.0509 0.0515 0.3228
MktRet*** -0.7117 0.1910 0.0002
MktVol*** -23.1775 2.1690 <.0001
Duration*** 0.0028 0.0001 <.0001
Observation 11,543
47
Table 5. HSE and Investor Demographics
This table reports the estimated coefficients, standard errors, and p-values for the following logit regression:
Withdrawali,t+1 = β0 + β1Lossi,t +β1Lossi,t×Female(or Age-bin) i,t
+ β3MktReti,t + β4MktVoli,t + β5Female(or Age-bin)i,t + β6Durationi,t + 𝜖𝑖,𝑡
The dependent variable, Withdrawal is equal to one if an individual neither purchase nor sell any stock within one
year since unwinding her first investment in full, and otherwise zero. Loss is a dummy variable equal to one if first
investment position was unwound with a loss, and otherwise zero. Magnitude(Loss) is a continuous variable
measuring the absolute value of negative realized returns on the first investment. MktRet and MktVol is return and
volatility (a standard deviation of daily return) respectively of OMX Helsinki Index, which are measured for one
month period after the investor’s exit. Duration is the length of time in days between entry and exit date. Female is
the dummy variable equal to one if women and zero if men. Age-bin is a categorical variable with 4 levels (0-19, 20-
39, 40-64, and 65+). The bin cosisting of accounts at age 0-19 is the reference level. Year is a year dummy variable
with year 1995 a reference level, representing the year in which an investor exit the market. Due to the dramatic
changes in stock market conditions during the year of 2000, two time dummies, 2000a and 2000b are used
respectively for the year before and after the busrt of dot-com bubbles (that is 30th
of April). ∗∗∗, ∗∗, and ∗ denote
statistical significance at 1%, 5%, and 10%. Panel A. Gender
Estimate
Standard
Error p-Value
Loss*** 1.0171 0.0618 <.0001
Loss×Female -0.0552 0.1208 0.6478
Female*** 0.1410 0.0490 0.004
MktRet -0.1915 0.1987 0.3351
MktVol** -6.4038 2.7686 0.0207
Duration*** 0.0028 0.0001 <.0001
Year Fixed Effect Yes
Observation 11,543
Panel B. Age
Estimate
Standard
Error p-Value
Loss*** 0.7868 0.1682 <.0001
Loss×Age(20-39) 0.1424 0.1830 0.4363
Loss×Age(39-64) 0.2821 0.1917 0.1412
Loss×Age(65-)* 0.5390 0.2782 0.0526
Age(20-39)*** 0.2250 0.0813 0.0057
Age(39-64) 0.0263 0.0827 0.7501
Age(65-) -0.0479 0.1135 0.6727
MktRet -0.1821 0.1992 0.3607
MktVol** -6.3349 2.7747 0.0224
Duration*** 0.0029 0.0001 <.0001
Year Fixed Effect Yes
Observation 11,543
48
Table 6. HSE and Brand Stocks
This table reports the estimated coefficients, standard errors, and p-values for the following logit regression:
Withdrawali,t+1 = β0 + β1Lossi,t +β1Lossi,t×Nokia i,t
+ β3MktReti,t + β4MktVoli,t + β5Nokiai,t + β6Durationi,t + 𝜖𝑖,𝑡
The dependent variable, Withdrawal is equal to one if an individual neither purchase nor sell any stock within one
year since unwinding her first investment in full, and otherwise zero. Loss is a dummy variable equal to one if first
investment position was unwound with a loss, and otherwise zero. Nokia is equal to 1 if investor’s first investment in
on Nokia stock, zero otherwise. MktRet and MktVol is return and volatility (a standard deviation of daily return)
respectively of OMX Helsinki Index, which are measured for one month period after the investor’s exit. Duration is
the length of time in days between entry and exit date. Female is the dummy variable equal to one if women and zero
if men. Age-bin is a categorical variable with 4 levels (0-19, 20-39, 40-64, and 65+). The bin cosisting of accounts at
age 0-19 is the reference level. Year is a year dummy variable with year 1995 a reference level, representing the year
in which an investor exit the market. Due to the dramatic changes in stock market conditions during the year of 2000,
two time dummies, 2000a and 2000b are used respectively for the year before and after the busrt of dot-com bubbles
(that is, 30th
of April). ∗∗∗, ∗∗, and ∗ denote statistical significance at 1%, 5%, and 10%.
Estimate
Standard
Error p-Value
Loss*** 0.9836 0.0626 <.0001
Loss×Nokia -0.1270 0.1209 0.2935
Nokia*** -0.2304 0.0468 <.0001
MktRet -0.1137 0.1993 0.5685
MktVol** -6.4877 2.7676 0.0191
Duration*** 0.0028 0.0001 <.0001
Year1996*** 0.6944 0.2401 0.0038
Year1997 0.0083 0.2337 0.9716
Year1998 -0.0284 0.2255 0.8997
Year1999*** -0.5719 0.2221 0.0100
Year2000a*** -0.7674 0.2261 0.0007
Year2000b** -0.4792 0.2251 0.0332
Year2001 -0.3322 0.2235 0.1373
Year2002 -0.3404 0.2246 0.1296
Observation 11,543
49
Table 7. HSE and Investor Location
This table reports the estimated coefficients, standard errors, and p-values for the following logit regression:
Withdrawali,t+1 = β0 + β1Lossi,t +β1Lossi,t×Helsinki i,t
+ β3MktReti,t + β4MktVoli,t + β5Helsinkii,t + β6Durationi,t + 𝜖𝑖,𝑡
The dependaet variable, Withdrawal is equal to one if an individual neither purchase nor sell any stock within one
year since unwinding her first investment in full, and otherwise zero. Loss is a dummy variable equal to one if first
investment position was unwound with a loss, and otherwise zero. Helsinki is equal to 1 if investor lives in Helsinki,
zero otherwise. MktRet and MktVol is return and volatility (a standard deviation of daily return) respectively of OMX
Helsinki Index, which are measured for one month period after the investor’s exit. Duration is the length of time in
days between entry and exit date. Female is the dummy variable equal to one if women and zero if men. Age-bin is a
categorical variable with 4 levels (0-19, 20-39, 40-64, and 65+). The bin cosisting of accounts at age 0-19 is the
reference level. Year is a year dummy variable with year 1995 a reference level, representing the year in which an
investor exit the market. Due to the dramatic changes in stock market conditions during the year of 2000, two time
dummies, 2000a and 2000b are used respectively for the year before and after the busrt of dot-com bubbles (that is,
30th
of April). ∗∗∗, ∗∗, and ∗ denote statistical significance at 1%, 5%, and 10%.
Estimate
Standard
Error p-Value
Loss*** 1.0352 0.0627 <.0001
Loss×Helsinki -0.1829 0.1180 0.1210
Helsinki*** 0.2637 0.0529 <.0001
MktRet -0.1938 0.1987 0.3295
MktVol** -6.9074 2.7696 0.0126
Duration*** 0.0028 0.0001 <.0001
Year1996*** 0.6578 0.2397 0.0061
Year1997 0.0574 0.2337 0.8058
Year1998 -0.0202 0.2252 0.9284
Year1999** -0.5279 0.2219 0.0173
Year2000a*** -0.6947 0.2258 0.0021
Year2000b** -0.4566 0.2249 0.0423
Year2001 -0.3056 0.2234 0.1713
Year2002 -0.3561 0.2242 0.1122
Observation 11,543
50
Table 8. HSE for Non-NOKIA Stocks
This table reports the estimated coefficients, standard errors, and p-values for the following logit regression:
Withdrawali,t+1 = β0 + β1Lossi,t + β2MktReti,t + β3MktVoli,t + β4Yeari,t + β5Durationi,t + 𝜖𝑖,𝑡
The dependent variable, Withdrawal is equal to one if an individual neither purchase nor sell any stock within one
year since unwinding her first investment in full, and otherwise zero. Loss is a dummy variable equal to one if first
investment position was unwound with a loss, and otherwise zero. MktRet and MktVol is return and volatility (a
standard deviation of daily return) respectively of OMX Helsinki Index, which are measured for one month period
after the investor’s exit. Duration is the length of time in days between entry and exit date. Year is a year dummy
variable with year 1995 a reference level, representing the year in which an investor exit the market. Due to the
dramatic changes in stock market conditions during the year of 2000, two time dummies, 2000a and 2000b are used
respectively for the year before and after the busrt of dot-com bubbles (that is, 30th
of April). ∗∗∗, ∗∗, and ∗ denote
statistical significance at 1%, 5%, and 10%.
Estimate
Standard
Error p-Value
Loss*** 0.9820 0.0634 <.0001
MktRet -0.0491 0.2500 0.8444
MktVol** -7.0944 3.3218 0.0327
Duration*** 0.0025 0.0001 <.0001
Year1996 0.5307 0.3268 0.1043
Year1997 -0.2535 0.2953 0.3907
Year1998 -0.2251 0.2910 0.4392
Year1999*** -0.7611 0.2839 0.0073
Year2000a*** -1.0404 0.2874 0.0003
Year2000b*** -0.7594 0.2885 0.0085
Year2001 -0.4350 0.2862 0.1286
Year2002 -0.3869 0.2909 0.1835
Observation 7,455
51
Table 9. HSE for Multiple Stock Entries
This table reports the estimated coefficients, standard errors, and p-values for the following logit regression:
Withdrawali,t+1 = β0 + β1Lossi,t +β1Lossi,t×Multiple i,t
+ β3MktReti,t + β4MktVoli,t + β5Multiplei,t + β6Durationi,t + 𝜖𝑖,𝑡
The dependent variable, Withdrawal is equal to one if an individual neither purchase nor sell any stock within one
year since unwinding her first investment in full, and otherwise zero. Loss is a dummy variable equal to one if first
investment position was unwound with a loss, and otherwise zero. Helsinki is equal to 1 if investor lives in Helsinki,
zero otherwise. MktRet and MktVol is return and volatility (a standard deviation of daily return) respectively of OMX
Helsinki Index, which are measured for one month period after the investor’s exit. Duration is the length of time in
days between entry and exit date. Female is the dummy variable equal to one if women and zero if men. Age-bin is a
categorical variable with 4 levels (0-19, 20-39, 40-64, and 65+). The bin cosisting of accounts at age 0-19 is the
reference level. Year is a year dummy variable with year 1995 a reference level, representing the year in which an
investor exit the market. Due to the dramatic changes in stock market conditions during the year of 2000, two time
dummies, 2000a and 2000b are used respectively for the year before and after the busrt of dot-com bubbles (that is,
30th
of April). Multiple is equal to 1 if investor initate trading with a portfolio of multiple stocks and zero if a single
stock. ∗∗∗, ∗∗, and ∗ denote statistical significance at 1%, 5%, and 10%.
Estimate
Standard
Error p-Value
Loss*** 0.9970 0.0537 <.0001
Loss×Multiple 0.2241 0.1604 0.1624
Multiple -0.0084 0.0802 0.9162
MktRet -0.2189 0.1895 0.2480
MktVol** -6.3948 2.6343 0.0152
Duration*** 0.0029 0.0001 <.0001
Year1996*** 0.6251 0.2382 0.0087
Year1997 0.0188 0.2315 0.9353
Year1998 -0.0339 0.2240 0.8796
Year1999*** -0.5706 0.2206 0.0097
Year2000a*** -0.7417 0.2240 0.0009
Year2000b** -0.5031 0.2233 0.0243
Year2001 -0.3640 0.2220 0.1010
Year2002* -0.3773 0.2230 0.0907
Observation 12,718