Exp Econ DOI 10.1007/s10683-015-9445-0 ORIGINAL PAPER
The influence of investment experience on market prices: laboratory evidence Ju¨rgen Huber1 • Michael Kirchler1,2 Thomas Sto¨ckl1
•
Received: 16 July 2014 / Revised: 7 May 2015 / Accepted: 12 May 2015 Ó Economic Science Association 2015
Abstract We run laboratory experiments to analyze the impact of prior investment experience on price efficiency in asset markets. Before subjects enter the asset market they gain either no, positive, or negative investment experience in an investment game. To get a comprehensive picture about the role of experience we implement two asset market designs. One is prone to inefficient pricing, exhibiting bubble and crash patterns, while the other exhibits efficient pricing. We find that (i) both, positive and negative, experience gained in the investment game lead to efficient pricing in both market settings. Further, we show that (ii) the experience effect dominates potential effects triggered by positive and negative sentiment generated by the investment game. We conjecture that experiencing changing price paths in the investment game can create a higher sensibility on changing fundamentals (through higher salience) among subjects in the subsequently run asset market. Keywords Experimental finance Asset market Bubble Mispricing Information Experience
Electronic supplementary material The online version of this article (doi:10.1007/s10683-015-94450) contains supplementary material, which is available to authorized users. & Thomas Sto¨ckl
[email protected] Ju¨rgen Huber
[email protected] Michael Kirchler
[email protected] 1
Department of Banking and Finance, Innsbruck University School of Management, Universita¨tsstrasse 15, 6020 Innsbruck, Austria
2
Centre for Finance, University of Gothenburg, P.O. Box 600, 40530 Gothenburg, Sweden
123
J. Huber et al.
JEL Classification
C92 D84 G10
1 Introduction The words ‘‘bubble’’, ‘‘crash’’ and ‘‘crisis’’ have dominated headlines of newspapers worldwide since 2007. The bursting of a bubble often affects the wider economy, causing unemployment and recession (see, e.g., Kindleberger 2011). Thus, understanding causes of bubbles and crashes is an eminently important challenge for practitioners, regulators and academics alike. Bubbles have arguably existed since the creation of modern financial markets, but reasons for their development are still far from fully understood. There is ample evidence that various forms of experience significantly affects subjects’ behavior in financial markets. By taking a historical perspective Malmendier and Nagel (2011) show that experiencing macroeconomic shocks influences financial risk taking. Gong et al. (2013) conducted laboratory experiments with real investors during a boom and later during a crash of the Shanghai stock exchange. They report that, compared to those in the crash-treatment, subjects in the boom-treatment were much more active in the laboratory markets and preferred to hold more shares than cash. In a very general study not directly related to financial markets, Lejarraga (2010) shows that experience sampling can increase subjects’ accuracy in non-lottery tasks and thus has a positive effect on subjects’ decisions. Taking a step towards financial markets there is evidence from lab and field experiments showing that experienced investors are less likely to exhibit behavioral biases like the disposition effect (Shapira and Venezia 2001; Feng and Seasholes 2005) or the endowment effect (List 2003). However, some studies also find that experienced investors show similar or even more attenuated behavioral biases or deviations from Expected Utility Theory compared to non-experienced subjects. Abdellaoui et al. (2013) report that a sample of private bankers and fund managers clearly behave according to Prospect Theory and thus violate utility maximization. The professionals are risk-averse for gains, risk-seeking for losses and exhibit loss aversion for mixed gambles (though less pronounced than assumed in the literature). Furthermore, Haigh and List (2005) show that professional CBOT-traders actually show a higher degree of myopic loss aversion compared to a students sample pointing at a detrimental effect of experience on decision quality. By using empirical data Greenwood and Nagel (2009) study the behavior of mutual fund managers during the Tech-stock bubble. They show that younger (less experienced) managers, compared to their more experienced colleagues, were more exposed to tech stocks and exhibited trend chasing behavior. The success of younger managers in the build-up phase of the bubble generated a massive inflow of money into these funds—a fact that may have additionally spurred the bubble. Earlier studies by Loewenstein and Lerner (2003) and Nguyen and Noussair (2014) suggest that positive past investment experience influences mood (e.g., feelings like anger and happiness) and this can in turn affect the readiness to take risks. Furthermore, there is evidence that the kind of experience, either positive or
123
The influence of investment experience on market prices...
negative, also influences behavior and market prices (Lakonishok and Smidt 1988; Shiller 2000; Hirshleifer 2001; Hirshleifer and Shumway 2003; Edmans et al. 2007; Gong et al. 2013). For instance Kaustia and Knu¨pfer (2008) and Chiang et al. (2011) show that successful investments in IPOs increase investors’ likelihood of participating in future auctions. Intuitively, these findings can be rationalized as follows: positive past experience might result in unduly optimistic expectations as good experiences bolster investors’ confidence (Weinstein 1980; Van den Steen 2004). If many investors develop the same optimistic view about the future based on past experience, markets might exhibit undesirable up and down swings. However, in Chiang et al. (2011) institutional investors do not exhibit such behavior. Additionally, experimental asset market studies also contribute to this strand of literature. Here, studies on bubbles and crashes were pioneered by the seminal model of Smith et al. (1988, henceforth SSW). They report substantial mispricing in their markets, although traders know the distribution of all future dividends, and thus know the expected fundamental value (FV) in advance (see Palan 2013 for a review of studies based on the SSW design). Experience, gained through multiple repetition of the same market, is an important factor that moderates mispricing in SSW settings (Smith et al. 1988; Van Boening et al. 1993; Dufwenberg et al. 2005; Haruvy et al. 2007; Lei and Vesely 2009; Huber and Kirchler 2012; Sutter et al. 2012; Cheung et al. 2014). Corgnet et al. (2010) explore the impact of releasing public messages with different levels of reliability on asset prices. They find that messages can play a significant role in bubble abatement, or rekindling. Hussam et al. (2008) find that the result is fragile to minor changes in the setup and consequently, bubbles can be rekindled even with experienced subjects. However, there is hardly any evidence whether experience from trading on real financial markets impacts price efficiency in the laboratory. There is only anecdotal evidence from one market in the classical study of Smith et al. (1988) that price efficiency is not improved when markets are populated by financial professionals instead of students. Hence, the effect of prior investment experience on subjects’ behavior and consequently on market prices is not unequivocally clear and seems to depend on the kind of experience gained and the market context used. To reduce this research gap we explore how different levels of investment experience (positive, negative, no experience) in an individual investment game influence price formation in subsequent laboratory asset markets. Here we analyze whether different levels of investment experience gained from one particular source of experience (i.e., an investment game) influence price efficiency. Thus, our first treatment variable is the level of investment experience subjects gain before the main experiment. We vary experience in three ways: no, positive or negative experience. Subjects either do not play or do play an investment game first. Those that collect investment experience in the game observe and earn either predominantly positive or predominantly negative returns. In the asset market that follows only subjects with identical experience (no, positive or negative) interact. To get a comprehensive picture on the influence of investment experience on markets, we introduce the salience of the fundamental value as second treatment variable. Both market settings are based on SSW but we vary their proneness to
123
J. Huber et al.
mispricing by varying the degree of FV-salience. Either the FV is shown only on the history screen (low salience treatment) or it is also displayed on the trading screen (high salience treatment). With our 32 design with the treatment variables ‘‘Experience’’ and ‘‘FV-salience’’—outlined in Table 1—we are able to test which of two effects, investment experience or mood, dominates. If mood dominates, we should see high (low) prices after positive (negative) investment experience. If experience dominates, we should see efficient prices irrespective of whether the prior investment experience was positive or negative. With our treatment design we are also able to compare the experience effect of the investment game (positive or negative) on price efficiency in a setting that is prone to bubbles with one that is usually efficient, but where bubbles might be triggered through prior positive investment experience. We find that (i) both, positive and negative, experience gained in the investment game lead to more efficient pricing. We conjecture that experiencing changing price paths in the investment game creates a higher sensibility on changing fundamentals and prices among subjects in the subsequently run asset market, as the fundamental value information is more salient. Further, we show that (ii) the experience effect dominates potential effects triggered by positive and negative sentiment generated by the investment game. This finding is remarkable as we show that positive and negative investment experience strongly influence subjects’ mood. Thus, we conjecture that experience and a related stronger focus on the exogenous FVprocess dominates sentiment effects on mispricing.
2 The experiment In each market ten subjects trade an asset for experimental currency (Taler) in a sequence of 15 periods of 150 s each. At the beginning of each market each subject is endowed with 40 shares and 3600 Taler. Valued at the initial FV of 45, which is identical in all treatments, each subject starts with an initial wealth of 5400 Taler. In total we conduct six treatments and two robustness check treatments, which differ in the precision of subjects’ FV-information and in their pre-market investment experience. 2.1 Treatments The market design is based on the basic setup of the SSW-setting Smith et al. (1988) and is similar to the setting outlined in Sto¨ckl et al. (2015). The asset pays a dividend of either 0, 3, or 6 Taler with equal probability at the end of each period. The FV in period 1 of a 15-period market is 45 Taler and decreases by the expected dividend of 3 Taler each period. After 15 periods the asset expires worthless and the terminal value of the asset is zero, which is public knowledge (see Online Appendix D for instructions and screen shots to the market experiments). Table 1 outlines the different treatments. In Treatment T1 the FV is explained in the instructions and displayed on the history screen for 15 s after each trading period
123
The influence of investment experience on market prices... Table 1 Treatment overview
Salience of fundamental value Low
High
Experience None
T1
T2FV
Positive
T3POS
T5FVPOS
Negative
T4NEG
T6FVNEG
of 150 s.1 Subjects do not play the investment game prior to the market opening. In Treatment T2FV we build on T1, with the only difference that the FV of a given period is additionally displayed on the trading screen throughout trading to increase the salience of fundamental information. Based on earlier experimental evidence we argue that this treatment will lead to more efficient prices. More specifically, Kirchler et al. (2012) and Huber and Kirchler (2012) report that the reduction of confusion about the fundamental value and different framing of the traded asset helps to reduce mispricing. Lei and Vesely (2009) outline the importance of a very careful explanation of the FV. These studies show that increasing the salience of FV-information increases price efficiency. We implement this treatment variation to compare the experience effect of the investment game (positive or negative) on price efficiency in a setting that is prone to bubbles with one that is usually efficient, but where bubbles might be triggered through prior positive investment experience. We run two robustness checks for treatments T1 and T2FV to rule out that pure anchoring on the displayed information on the FV is the major reason for efficiency. In Treatment T7MAXFV we build on T1, but we display the maximum possible dividend value of a given period on the trading screen—i.e., the maximum FV when a dividend of 6 is realized in all periods. With this treatment we test whether any information which is a good proxy for the FV influences prices and mispricing in the market. In Treatment T8NOISE we test whether non-relevant information (noise) has an impact on mispricing. Here, the average period trading price of another, unrelated, experimental market is displayed on the trading screen.2 With this treatment we test whether in Treatment T2FV subjects simply anchor on the displayed information because it serves as a focal point upon which to coordinate or whether it really reduces confusion about the fundamental value of the asset. In the four main treatments T3POS , T4NEG , T5FVPOS and T6FVNEG subjects play an investment game (inspired by, Lohrenz et al. 2007) before trading on the market. The game is not a market, but subjects make individual decisions and have no influence on prices. Subjects have to decide which percentage of their wealth to 1
In all treatments the FV of period p is disclosed on a history screen after trading ended in period p. In prior studies the development of the FV is mostly displayed in a table in the experimental instructions (e.g., see the experimental instructions of Noussair et al. 2001, Dufwenberg et al. 2005, Huber and Kirchler 2012, and Kirchler et al. 2012). We find no statistical differences between our baseline treatment T1 and the comparable baseline treatment in Kirchler et al. (2012) which uses the FV-table in the instructions of the experiment. See also Sto¨ckl et al. (2015) as a reference study for displaying the FV on the history screen instead of displaying it in the experimental instructions.
2
Specifically, the price path was taken from Sto¨ckl et al. (2015), market 1 of treatment R3(=).
123
J. Huber et al.
invest in the stock market, with the price development externally given. Specifically, the price development is taken from the real historical development of weekly prices of the DJIA and the S&P500 between January 1928 and December 2012. This is known to subjects. Subjects play six independent rounds and each round uses a unique time series that spans over a period of 72 weeks (periods). To ensure comparability of rounds, index values are adjusted to 100 in the starting week of the selected time period. At the beginning of each round, each subject receives an initial endowment of 1000 Eurocent (10 Euro) and they see the time series development of the index for the first 24 weeks. Starting with period 24 subjects can invest a fraction (integer values ranging from 0 to 100) of their current wealth in the development of the index over the upcoming four weeks. After entering their decision they are informed about the development of the index in these four weeks (four data points are added to the chart) and their current wealth is adjusted accordingly. This procedure is repeated 12 times in each round until the time series ends in week 72. In total they make six (rounds) times 12 = 72 investment decisions. Subjects have 20 s for each investment decision, otherwise the invested fraction is simply set to zero for that period. The initial endowment in each market is set to 10 Euro and one of the six rounds is selected randomly for payout.3 What subjects did not know is that we have pre-selected six predominantly increasing and six predominantly decreasing price paths. To create positive or negative investment experience we provide all subjects of the same market either with a set of increasing or a set of decreasing prices. While subjects received identical instructions, they did not know that all subjects in a market would have experienced the same index realizations. We chose this design to avoid common knowledge of investment experience, as e.g. given a bullish investment experience speculative behavior might be fostered, possibly favoring bubbles (Cheung et al. 2014). The variable ‘‘Experience’’ is either ‘‘POS’’ (positive) or ‘‘NEG’’ (negative). The distinction in predominantly positive and negative experience is important, as it allows us to measure the impact of positive and negative investment experience on prices in the subsequently run markets.4 To test whether this investment experience triggers positive or negative emotions/mood we test subjects’ emotional state with the well-established SAMprocedure (self-assessment manikin by, Bradley and Lang 1994) and with the emotional scale of Izard (1991) before the asset market opens and after the market ends. With the non-verbal pictorial SAM-procedure mood is elicited on a 9-point scale. The SAM measures pleasure, arousal, and dominance associated to a person’s affective reaction to the experiences in the investment game and the asset market. Emotions elicited according to the 7-point scale of Izard (1991) include positive and 3 Subjects earned 12.13 Euro in the investment game preceding T3POS , 8.96 Euros in the investment game preceding T4NEG , 12.05 Euros in the investment game preceding T5FVPOS , and 8.53 Euros in the investment game preceding T6FVNEG with an average of 10.66 Euros. Average earnings in the subsequent laboratory market were 13.50 Euros, hence the investment game on average made up some 44 percent of total earnings from these two experiments. 4
See Online Appendix B for instructions and Online Appendix C for information on the price (index) paths used in the NEG and POS experience treatments of the investment game.
123
The influence of investment experience on market prices...
negative emotions such as pleasure, interest, anger, and fear. By eliciting subjects’ emotions prior to the market opening we test whether positive or negative mood induced by the investment game influences price efficiency. In treatments T3POS and T4NEG subjects trade in the classic bubble-prone setting of SSW without FV display on the trading screen. Treatments (T3POS and T4NEG ) differ in the experience gained in the investment game as subjects in T3POS experience bullish price paths and subjects in T4NEG experience bearish price paths. Instead, in treatments T5FVPOS and T6FVNEG subjects get the FV displayed on the trading screen as in Treatment T2FV . In Treatment T5FVPOS subjects enter the market with positive experience (POS) as they experience six increasing price paths in the investment game. Subjects in Treatment T6FVNEG , however, start trading with negative experience (NEG) as they are confronted with six decreasing price paths before the market starts. 2.2 Market architecture In all treatments subjects trade in a continuous double auction with open order books. All orders are executed according to price and then time priority. Market orders have priority over limit orders and are always executed instantaneously. Any order size and the partial execution of limit orders are possible. Posted limit orders can be cancelled without costs. Taler and asset holdings are carried over from one period to the next. No interest is paid on Taler holdings and there are no transaction costs. Shorting assets and borrowing money is not allowed. 2.3 Implementation of the experiment Six sessions were run for each of the six treatments and the two robustness checks.5 All 52 laboratory sessions were conducted at the Innsbruck EconLab with a total of 520 students (bachelor and master students in business administration and economics). In part 1 of treatments T3POS , T4NEG , T5FVPOS , and T6FVNEG subjects had ten minutes to read the instructions of the investment game on their own and questions were answered privately. Part 1 took roughly 45 minutes. In part 2 of treatments T3POS , T4NEG , T5FVPOS , and T6FVNEG (part 1 in treatments T1, T2FV , T7MAXFV , and T8NOISE ), subjects had 15 minutes to read the instructions of the market experiment on their own and questions were answered privately. Afterwards, the trading screen was explained in detail, followed by two trial periods to allow subjects to become familiar with the trading interface. In all markets the traded asset is worthless after the last period. Thus, only Taler holdings were converted at a known exchange rate of 400 Taler = 1 Euro. This part lasted around 60 minutes. Each subject participated in only one session of this study. We especially took care that subjects did not participate in earlier asset market experiments of 5
The robustness checks are presented in greater detail in Appendix A. Note that we ran 10 markets for Treatment T3POS .
123
J. Huber et al.
comparable designs. The investment game and the markets were programmed and conducted with z-Tree 3.2.8. by Fischbacher (2007). Subjects were recruited using ORSEE by Greiner (2004).
3 Results Figure 1 pictures the evolution of individual market prices (grey lines), of mean treatment prices (bold line with circles), and of FVs (bold line). To quantify mispricing and overvaluation we follow Sto¨ckl et al. (2010) and calculate RAD (relative absolute deviation) and RD (relative deviation). The average bid-askspread (SPREAD), share turnover (ST), and the standard deviation of log-returns (VOLA) are provided to give further market statistics. Additionally, we calculate the standard deviation of period mean prices (DEV) as a proxy for the variability of price paths within a treatment.6 Table 2 outlines details on the calculation of these measures and Table 3 provides treatment averages for all measures. Differences in treatment averages along with results from pairwise two-sided Mann–Whitney U tests are presented in Table 4.7 Individual market results (except for DEV) can be found in Online Appendix B. Treatment T1 (Fig. 1, top left) exhibits the classic pattern for SSW-markets with prices remaining high while the FV falls and a price crash towards the end of the experiment. In particular, RAD of 43.2 percent and RD of 39.9 percent are significantly higher than in T2FV (Fig. 1, top right) making T1 stand out as a treatment with very high mispricing and very high overvaluation. T2FV ’s values for RAD and RD are significantly lower than T1’s with RAD of 16.1 percent and RD of 11.8 percent. We attribute this to the higher salience the FV has in T2FV . This result is in line with Huber and Kirchler (2012). We run two robustness checks for treatments T1 and T2FV to rule out that pure anchoring on the displayed information is the major reason for the observed increase in efficiency. The results (see Appendix A for details) are as follows: displaying a useful proxy of the FV (Treatment T7MAXFV ) on the trading screen seems to reduce mispricing and overvaluation, marginally insignificant though. Displaying other, non-relevant, information on the trading screen (Treatment T8NOISE ) does not significantly influence price efficiency compared to Treatment T1. Thus, we can rule out that simple anchoring on the provided information drives our results. Before we turn to results of the other four treatments we discuss the impact of the investment game on subjects’ mood. The interesting question to be answered is, whether mood spills over to the market, as claimed by Andrade et al. (2012), or whether the effect of experience dominates mood. The top panels of Fig. 2 show that subjects’ emotions before the market opening are significantly different between the POS- and the NEG-experience treatments. In particular, subjects show significantly 6
The calculation of DEV yields one observation per treatment.
7
Additionally we provide results from two-sample Kolmogorov–Smirnov tests to test for differences in distributions. Tests for treatment differences in DEV are based on period values. Note, however, that these values do not constitute independent observations.
123
75 60 45 30 15
Mean price and fundamental value
T2(FV)
0
15
30
45
60
75
90
T1
0
Mean price and fundamental value
The influence of investment experience on market prices...
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
1
2
3
4
5
6
7
period FV
10
11
12
13
14
15
Individual markets
15
30
45
60
75
T4(NEG)
0
15
30
45
60
75
T3(POS)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
1
2
3
4
5
6
7
period
9
10
11
12
13
14
15
Indiv. markets
FV
Indiv. markets
45
60
75
90
T6(FVNEG)
0
0
15
30
45
60
75
Mean price and fundamental value
T5(FVPOS)
Mean price
30
Mean price
90
8
period
15
FV
Mean price and fundamental value
9
Mean price
90
Individual markets
Mean price and fundamental value
90
Mean price
0
Mean price and fundamental value
FV
8
period
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
1
2
3
4
period FV
Mean price
5
6
7
8
9
10
11
12
13
14
15
period Indiv. markets
FV
Mean price
Indiv. markets
Fig. 1 Fundamental value (FV, bold line), mean prices (bold line with circles) and volume-weighted mean prices for individual markets (grey lines) as a function of period of T1 (top left), T3POS (middle left), T5FVPOS (bottom left), T2FV (top right), T4NEG (middle right), and T6FVNEG (bottom right)
higher levels of positive emotions (pleasure, dominance) in the POS-conditions (Mann–Whitney U tests, p values\0.05, N = 280) and significantly higher values in negative emotions (anger, sorrow) in the NEG-conditions (Mann–Whitney U tests, p values\0.05, N = 280). From the bottom panels of Fig. 2 it is evident that emotions are almost identical in both conditions after the market has ended. As is visible in Fig. 1 (middle and bottom panels), all four treatments with premarket investment experience are very efficient with levels of mispricing (overvaluation) ranging from 9.2 (6.4) to 27.0 (19.7) percent. There are no significant differences between the four treatments, which indicates that positive and negative investment experiences have an almost identical impact on price efficiency (see Table 4). To check whether this result is not driven by a lack of power due to the small sample size, we pool the two treatments under each type of experience (T3POS and T5FVPOS for positive and T4NEG and T6FVNEG for negative
123
J. Huber et al. Table 2 Price efficiency measures and formulae Measure
Calculation P RAD ¼ N1 Np¼1 Pp FVp =jFVj P RD ¼ N1 Np¼1 ðPp FVp Þ=jFVj P SPREAD ¼ N1 Np¼1 SPREADp =FVp . P ST ¼ Np¼1 VOLp =TSO ffi P qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi PT 2 1 VOLA ¼ N1 Np¼1 T1 t¼1 ðRETt RETÞ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi P PM 2 1 DEVTR ¼ N1 Np¼ 1 M1 m¼1 ðPp PM Þ
Relative absolute deviation Relative deviation Bid-ask spread Share turnover Standard deviation of log returns Standard deviation of period mean prices
p indexes period, N total number of periods, m indexes market, M total number of markets, P (volumeweighted) mean price, FV fundamental value, FV average fundamental value of the market, SPREAD (volume-weighted) average bid-ask spread evaluated at each transaction, VOL number of shared traded, TSO total number of shares outstanding, RETt ¼ lnðPt =Pt1 Þ, RET mean of log-returns in period p, T number of transactions in period p, PM mean price in period p over all markets of a treatment, TR index treatment. Note that DEV is calculated on treatment level Table 3 Treatment averages for RAD (mispricing in percent of FV), RD (overvaluation in percent of FV), SPREAD (bid-ask spread), ST (share turnover), VOLA (standard deviation of log returns), and DEV (standard deviation of period mean prices) Treatments
RAD (%)
RD (%)
SPREAD (%)
ST
VOLA (%)
DEV
T1
43.2
39.9
16.6
3.67
9.7
7.41
T2FV
16.1
11.8
11.0
3.21
6.1
8.13
T3POS
27.0
19.7
23.0
2.42
9.5
7.59
T4NEG
14.7
7.0
14.1
2.75
8.7
5.32
T5FVPOS
13.3
10.4
19.3
2.72
11.0
3.31
T6FVNEG
9.2
6.4
13.4
2.73
8.1
2.44
experience) to increase the number of observations. We run double-sided Mann– Whitney U tests and two-sample Kolmogorov–Smirnov tests (exact p values computed) on our measures. None of the test statistics turns out significant, hinting at no difference between positive and negative investment experience.8,9 Although there are no differences in RAD and RD between all four treatments with pre8 Mann–Whitney U tests, N = 28 (60 for DEV). RAD: z = -1.393; p value = 0.1637. RD: z = -1.300; p value = 0.1936. SPREAD: z = -1.346; p value = 0.1782. ST: z = 0.511; p value = 0.6096. VOLA: z = -0.650; p value = 0.5157. DEV: z = -1.552; p value = 0.1206. Kolmogorov–Smirnov tests, N = 28 (60 for DEV). RAD: p value = 0.442. RD: p value = 0.442. SPREAD: p value = 0.237. ST: p value = 0.442. VOLA: p value = 0.802. DEV: p value = 0.183. 9
We would like to stress that we have also investigated the correlations between behavior in the investment game (e.g., average percent invested, final Taler holdings, emotions after the investment game) and subsequent trading behavior on the market (e.g., final cash holdings, number of limit or market orders posted, the ratio of shares bought minus shares sold) on an individual level. We find no systematic correlations between behavior in the investment game and subsequent market activity on an individual level. These findings are in line with our major finding that no differences in price efficiency emerge between positive and negative investment experience. Detailed results can be provided upon request.
123
The influence of investment experience on market prices... Table 4 Differences between market averages (column minus row) in percentage points (except ST and DEV) for measures of mispricing (RAD), overvaluation (RD), average bid-ask-spread (SPREAD), share turnover (ST), standard deviation of log returns (VOLA), and the average deviation of period mean prices (DEV). The table only reports test statistics for treatment comparisons reasonable under c.p. conditions T2FV
T3POS
T4NEG
T5FVPOS
T6FVNEG
RAD T1
-27.1*ðÞ
-16.2
-28.5*
–
–
T2FV
–
–
–
-2.8
-6.9
T3POS
–
–
-12.3
-13.7
–
T4NEG
–
–
–
-1.4
-5.5
T5FVPOS
–
–
–
–
-4.1
RD T1
-28.1*
-20.2*
-32.9**
–
–
T2FV
–
–
–
-1.4*ðÞ
-5.4
T3POS
–
–
-12.7
-9.3
–
T4NEG
–
–
–
3.4
-0.6
T5FVPOS
–
–
–
–
-4.0
SPREAD T1
-5.6
6.4
-2.5
–
–
T2FV
–
–
–
8.3
-3.2
T3POS
–
–
-8.9
-3.7
–
T4NEG
–
–
–
5.2
-0.7
T5FVPOS
–
–
–
–
-5.9
ST T1
-0.46
-1.25*ðÞ
-0.92*
–
–
T2FV
–
–
–
-0.49
-0.48 –
T3POS
–
–
0.33
0.30
T4NEG
–
–
–
-0.03
-0.02
T5FVPOS
–
–
–
–
0.01
VOLA T1
-3.6
-0.2
-1.0
–
–
T2FV
–
–
–
4.9**
2.0
T3POS
–
–
-0.8
1.5
–
T4NEG
–
–
–
2.3
-0.6
T5FVPOS
–
–
–
–
-2.9
0.72
0.18
-2.09ðÞ
–
DEV T1 T2FV
–
–
T3POS
–
–
T4NEG
–
–
–
-4.82***
-5.69***ðÞ
-2.27*ðÞ
-4.28***ðÞ
–
–
-2.01***ðÞ
-2.88***ðÞ
–
ðÞ
123
J. Huber et al. Table 4 continued
T5FVPOS
T2FV
T3POS
T4NEG
T5FVPOS
T6FVNEG
–
–
–
–
-0.87ðÞ
*ðÞ , **ðÞ and ***ðÞ denote the 10, 5 and the 1 % significance levels, derived from double-sided Mann–Whitney U tests (two-sample Kolmogorov–Smirnov tests, exact p values computed). Note that Kolmogorov–Smirnov tests indicate difference in distribution. Number of observations equals 12 (30) for RAD, RD, SPREAD, ST, and VOLA (DEV)
9
Emotional state (Izard 1991) − pre market SAM by Bradley and Lang (1994) − pre market ***
anger
***
7
*** ***
fear
*
guilt
*** ***
interest
***
pleasure
5
average value
contempt
***
reluctance
**
3
shame sorrow
***
surprise 1
0 pleasure
arousal
positive experience
1
dominance
2 average value
positive experience
negative experience
3
4
negative experience
Emotional state (Izard 1991) − post market 9
SAM by Bradley and Lang (1994) − post market anger
7
fear
***
guilt interest pleasure
5
average value
contempt
reluctance
***
3
shame sorrow surprise 1
0 pleasure
arousal
positive experience
dominance negative experience
1
2 average value
positive experience
3
4
negative experience
Fig. 2 Emotions of subjects before (and thus after the investment game) and after the market experiment. Left figures Mood is elicited according to the 9-point scale of the non-verbal pictorial assessment technique self-assessment manikin (SAM) of Bradley and Lang (1994) after the investment game (top left) and after the asset market (bottom left). The SAM measures pleasure (left bars), arousal (middle bars) and dominance (right bars) associated to a person’s affective reaction to the experiences in the investment game. Right figures Emotions elicited according to the 7-point emotions scale of Izard (1991) ranging from 0 (no emotion at all) to 6 (highest level of emotion). *, **, and *** represent the 10, 5, and 1 % significance levels of double-sided Mann–Whitney U tests
market investment experience, there exist differences in the variability of price paths in each treatment (DEV). It is evident that treatments without fundamental value saliency, T3POS and T4NEG , show significantly higher variations between markets within a treatment compared to treatments T5FVPOS and T6FVNEG .10 For the other variables outlined in Table 4 no notable statistical differences are detectable. 10 Note that treatments T5FVPOS and T6FVNEG also show significantly lower variability of price paths compared to treatment T2FV . This is due to one outlier market in the latter treatment. When running the analyses without this market differences are insignificant.
123
The influence of investment experience on market prices...
Remarkably, investment experience gained in the investment game leads to a significantly higher level of price efficiency at lower trading volume compared to the baseline markets of T1. The differences in mispricing between Treatment T1 and treatments T3POS and T4NEG are large with values of 16.2 and 28.5 % points, respectively. The corresponding numbers are even stronger for overvaluation with values of 20.2 and 32.9 % points, respectively. Results are weaker when positive and negative investment experience is introduced in markets with highly salient FVinformation, as the benchmark (T2FV ) is already very efficient. RAD and RD are still slightly lower in treatments T5FVPOS and T6FVNEG compared to Treatment T2FV , though differences are statistically insignificant.
4 Conclusion and discussion In this paper we investigated whether general investment experience gained in an individual-decision investment game has an effect on price efficiency in subsequently run laboratory markets. To get a comprehensive picture about the role of investment experience in efficient and inefficient markets, we applied a 3 2 design with ‘‘Experience’’ and ‘‘FV-salience’’ as treatment variables. We varied the degree of FV-salience and therefore the degree of price efficiency by displaying the correct FV only on the history screen or by displaying it also on the trading screen. We varied experience in three ways. Subjects either did or did not participate in an investment game first. Those that acquired investment experience in the game received predominantly positive or negative returns and thus had gained experience with either bullish or bearish price paths. We found that (i) both, positive and negative, experience gained in the investment game led to efficient pricing. We conjecture that the mechanism behind this finding could be as follows: In the investment game subjects learn to focus on the exogenously given variable, i.e., the price path that changes constantly. They base their investment decisions solely on this information (as there is no other information to process and as they are price-takers). In the market experiments that follow, the exogenously given variable is the FV, which changes from period to period. Given their past experience, the changing fundamental value seems to be more salient for subjects, who seem to focus more on the FV in the market context and so prices track fundamentals very closely. Thus, experiencing changing price paths in the investment game created a higher sensibility on changing fundamentals (which are more salient) among subjects in the subsequently run asset market. Our findings offer an additional perspective on the role of experience in asset markets. Experiencing the market several times (Smith et al. 1988; Van Boening et al. 1993; Dufwenberg et al. 2005; Haruvy et al. 2007; Lei and Vesely 2009; Cheung et al. 2014; Sutter et al. 2012) or experiencing the dividend stream and the development of the FV before trading fosters understanding of the FV and increases price efficiency (Lei and Vesely 2009; Huber and Kirchler 2012). Our results suggest that investment experience gained from pre-market investment games also improves price efficiency–most likely by increasing salience of the changing fundamental value.
123
J. Huber et al.
Furthermore, we showed that (ii) the experience effect dominated potential effects triggered by positive and negative sentiment generated by the investment game. This finding is remarkable, as we showed that positive and negative investment experience strongly influenced subjects’ sentiment. This result is surprising as subjects’ emotions after the investment game differ strongly between POS- and the NEG-markets. Thus experiences dominate potential sentiment effects on mispricing. Andrade et al. (2012) and Breaban and Noussair (2013) provided first evidence showing that positive mood might lead to stronger bubbles in asset markets. However, we do not find any sentiment to spill over to the asset market which refines the findings of Andrade et al. (2012). A major difference with our study, however, is that in Andrade et al. (2012) mood is induced by non-investment experience as it is generated by video clips before trading opens. We contribute to this emerging field of research with our study. We can relate our results to studies pointing at a relationship between investor experience and behavior. For instance, Shapira and Venezia (2001), List (2003), and Feng and Seasholes (2005) find that increased investor experience reduces behavioral biases such as the disposition effect. Malmendier and Nagel (2011) observe that experiencing macroeconomic shocks affect financial risk taking as well. Gong et al. (2013) showed that investors in an experiment were ready to hold more stocks after experiencing a boom period in the real world than after a crash. We provide further insights on the role of experience as we show that already very basic investment experience can shape individual behavior and consequently price efficiency in markets. Finally, we want to discuss potential future avenues of research in this field. Our study focuses on a very specific type of non-context related investment experience, i.e., the investment game, because within this particular form of experience we were able to analyze different levels of experience (positive experience which triggers positive mood vs. negative experience which triggers negative mood). In contrast to our approach, a very interesting avenue of research would be to analyze the impact of different forms of investment experience, such as investment games with or without interactions of subjects, investment game versus pre-run asset market or experience gained as a real trader versus no experience gained from real markets. However, we leave the research agenda of investigating different forms of investment experience open for future research. Acknowledgments We thank Charles Noussair (the Editor) and two referees for their very constructive and helpful comments. We thank participants of SAET 2011 and Experimental Finance 2011 for helpful comments. Financial support by the Austrian National Bank (OeNB-Grants 12789 and 14953), the Austrian Science Foundation (FWF-Grants 20609 22400, START-Grant Y617-G11), and the University of Innsbruck (Nachwuchsfo¨rderung Sto¨ckl) is gratefully acknowledged.
Appendix A: Robustness checks Figure 3 pictures the evolution of individual market prices (grey lines), of mean treatment prices (bold line with circles), and of FVs (bold line). Table 5 provides treatment averages for RAD, RD, SPREAD, and ST. Difference in treatment
123
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15
90 15
30
45
60
75
T2(FV)
0
Mean price and fundamental value
15
30
45
60
75
90
T1
0
Mean price and fundamental value
The influence of investment experience on market prices...
1
2
3
4
5
6
7
period FV
Individual markets
1
2
3
4
5
6
7
8
9
maxFV
10 11 12 13 14 15
10 11 12 13 14 15
Mean price
Individual markets
15
30
45
60
75
90
T8(NOISE)
1
2
3
4
5
6
period FV
9
0
15
30
45
60
75
90
T7(MAXFV)
Mean price and fundamental value
Mean price
0
Mean price and fundamental value
FV
8
period
7
8
9
10 11 12 13 14 15
period
Mean price
Indiv.markets
FV
Noise
Mean price
Indiv.markets
Fig. 3 Fundamental value (FV, bold line), mean prices (bold line with circles) and volume-weighted mean prices for individual markets (grey lines) as a function of period of T1 (top left), T2FV (top right), T7MAXFV (bottom left) and T8NOISE (bottom right) Table 5 Treatment averages for RAD (mispricing in percent of FV), RD (overvaluation in percent of FV), SPREAD, ST (share turnover), VOLA (standard deviation of log returns), and DEV (the average deviation of period mean prices) for robustness check treatments Treatments
RAD (%)
RD (%)
SPREAD (%)
ST
VOLA (%)
DEV
T7MAXFV
33.5
19.3
21.4
2.31
13.6
9.95
T8NOISE
57.0
53.0
43.5
3.12
14.8
10.61
averages along with results from pairwise two-sided Mann–Whitney U tests are presented in Table 6.Treatment T7MAXFV demonstrates that prices are also quite efficient when a reliable proxy, i.e., information closely related to the FV, is shown. Compared to T1 the display of the maximum FV reduces RAD and RD, though insignificantly because of high variance, by 9.7 and 20.6 % points, respectively. At the same time there are no significant differences in RD compared to Treatment T2FV , while visual inspection of Fig. 3 indicates that variations between markets are markedly higher in T7MAXFV compared to T2FV . Thus, an unbiased proxy for the FV has a positive effect by decreasing mispricing, but less so than precise information on the FV. In contrast, displaying irrelevant ‘‘noise’’ by showing average period prices of another cohort in Treatment T8NOISE neither serves as anchor nor leads to more efficient prices. Values of RAD and RD of 57.0 percent and 53.0 percent are significantly higher compared to treatments T2FV and T7MAXFV , and indistinguishable from the baseline T1.
123
J. Huber et al. Table 6 Differences between market averages (column minus row) in percentage points (except ST) for measures of mispricing (RAD), over-valuation (RD), average bid-ask-spread (SPREAD), share turnover (ST), standard deviation of log returns (VOLA), and the average deviation of period mean prices (DEV) T2FV
T7MAXFV
T8NOISE
-9.7
13.8
RAD T1
-27.1*ðÞ
T2FV
–
17.4*
40.9*ðÞ
T7MAXFV
–
–
23.5*
ðÞ
RD T1
-28.1*
-20.6
13.1
T2FV
–
7.5
41.2*ðÞ
T7MAXFV
–
–
33.7*
T1
-5.6
4.8
29.6**ðÞ
T2FV
–
10.4**
32.5***ðÞ
T7MAXFV
–
–
22.1*
SPREAD
ST T1
-0.46
-1.36**
-0.55
T2FV
–
-0.9**
0.09
T7MAXFV
–
–
0.81
T1
3.6
3.9
5.1
T2FV
–
7.5***ðÞ
8.7***ðÞ
T7MAXFV
–
–
1.2
T1
0.72
2.54
3.20*
T2FV
–
1.82
2.48
–
–
0.66
VOLA
DEV
T7MAXFV ðÞ
ðÞ
ðÞ
* , ** and *** denote the 10, 5 and the 1 % significance levels, derived from double-sided Mann–Whitney U tests (two-sample Kolmogorov–Smirnov tests, exact p values computed). Note that Kolmogorov–Smirnov tests indicate difference in distribution. Number of observations equals 12 (30) for RAD, RD, SPREAD, ST, and VOLA (DEV)
To summarize, we observe that displaying precise information about the FV (T2FV ) or useful proxies thereof (T7MAXFV ) on the trading screen eliminates or at least reduces mispricing and overvaluation. These results are in line with Corgnet et al. (2010) who show that qualitative messages about the level of mispricing can
123
The influence of investment experience on market prices...
play a significant role in bubble abatement, or rekindling. Displaying other nonrelevant information on the trading screen (T8NOISE ) does not significantly influence price efficiency. Thus, we can rule out that simple anchoring drives our results.
References Abdellaoui, M., Bleichrodt, H., & Kammoun, H. (2013). Do financial professionals behave according to prospect theory? An experimental study. Theory and Decision, 74, 411–429. Andrade, E., Odean, T., & Lin, S. (2012). Bubbling with excitement: An experiment. Working Paper. Bradley, M., & Lang, P. J. (1994). The self-assessment manikin and the semantic differential. Journal of Behavior Therapy and Experimental Psychiatry, 25(1), 49–59. Breaban, A., Noussair, C. N. (2013). Emotional state and market behavior. Working paper. Cheung, S. L., Hedegaard, M., & Palan, S. (2014). To see is to believe—common expectations in experimental asset markets. European Economic Review, 66, 84–96. Chiang, Y.-M., Hirshleifer, D., Qian, Y., & Sherman, A. E. (2011). Do investors learn from experience? Evidence from frequent ipo investors. The Review of Financial Studies, 24(5), 1560–1589. Corgnet, B., Kujal, P., & Porter, D. (2010). The effect of reliability, content and timing of public announcements on asset trading behavior. Journal of Economic Behavior and Organization, 76(2), 254–266. Dufwenberg, M., Lindqvist, T., & Moore, E. (2005). Bubbles and experience: An experiment. The American Economic Review, 95(5), 1731–1737. Edmans, A., Garcia, D., & Norli, O. (2007). Sports sentiment and stock returns. The Journal of Finance, 62(4), 1967–1998. Feng, L., & Seasholes, M. S. (2005). Do investor sophistication and trading experience eliminate behavioral biases in financial markets? Review of Finance, 9, 305–351. Fischbacher, U. (2007). z-tree: Zurich toolbox for ready-made economic experiments. Experimental Economics, 10(2), 171–178. Gong, B., Lei, V., & Pan, D. (2013). Before and after: The impact of a real bubble crash on investors’ trading behavior in the lab. Journal of Economic Behavior & Organization, 95, 186–196. Greenwood, R., & Nagel, S. (2009). Inexperienced investors and bubbles. Journal of Financial Economics, 93, 239–258. Greiner, B. (2004). Forschung und wissenschaftliches Rechnen 2003, An online recruitment system for economic experiments. GWDG Bericht 63. Gesellschaft fuer Wissenschaftliche Datenverarbeitung, Goettingen, pp. 79–93. Haigh, M. S., & List, J. A. (2005). Do professional traders exhibit myopic loss aversion? An experimental analysis. The Journal of Finance, 60, 523–534. Haruvy, E., Lahav, Y., & Noussair, C. (2007). Traders’ expectations in asset markets: Experimental evidence. The American Economic Review, 97(5), 1901–1920. Hirshleifer, D. (2001). Investor psychology and asset pricing. The Journal of Finance, 56, 1533–1597. Hirshleifer, D., & Shumway, T. (2003). Good day sunshine: Stock returns and the weather. Journal of Finance, 58, 1009–1032. Huber, J., & Kirchler, M. (2012). The impact of instructions and procedure on reducing confusion and bubbles in experimental asset market. Experimental Economics, 15, 89–105. Hussam, R. N., Porter, D., & Smith, V. L. (2008). Thar she blows: Can bubbles be rekindled with experienced subjects? The American Economic Review, 98(3), 924–937. Izard, C. (1991). The psychology of emotions. New York: Plenum Press. Kaustia, M., & Knu¨pfer, S. (2008). Do investors overweight personal experience? Evidence from ipo subscriptions. The Journal of Finance, 63(6), 2679–2702. Kindleberger, C. P. (2011). Manias, panics, and crashes: A history of financial crises (6th ed.). New York: Palgrave Macmillan. Kirchler, M., Huber, J., & Sto¨ckl, T. (2012). Thar she bursts—reducing confusion reduces bubbles. The American Economic Review, 102(2), 865–883.
123
J. Huber et al. Lakonishok, J., & Smidt, S. (1988). Are seasonal anomalies real? A ninety- year perspective. Review of Financial Studies, 1(4), 403–425. Lei, V., & Vesely, F. (2009). Market efficiency: Evidence from a no-bubble asset market experiment. Pacific Economic Review, 14(2), 246–258. Lejarraga, T. (2010). When experience is better than description: Time delays and complexity. Journal of Behavioral Decision Making, 23, 100–116. List, J. A. (2003). Does market experience eliminate market anomalies? The Quarterly Journal of Economics, 118(1), 41–71. Loewenstein, G., & Lerner, J. S. (2003). The role of affect in decision making. In R. J. Davidson, K. R. Scherer, & H. Goldsmith (Eds.), Handbook of affective science. New York: Oxford University Press. Lohrenz, T., McCabe, K., Camerer, C. F., & Montague, R. (2007). Neural signature of fictive learning signals in a sequential investment task. Proceedings of the National Academy of Science, 104(22), 9493–9498. Malmendier, U., & Nagel, S. (2011). Depression babies: Do macroeconomic experiences affect risk taking? The Quarterly Journal of Economics, 126, 373–416. doi:10.1093/qje/qjq004. Nguyen, Y., & Noussair, C. N. (2014). Risk aversion and emotions. Pacific Economic Review, 19, 296–312. Noussair, C. N., Robin, S., & Ruffieux, B. (2001). Price bubbles in laboratory asset markets with constant fundamental values. Experimental Economics, 4, 87–105. Palan, S. (2013). A review of bubbles and crashes in experimental asset markets. Journal of Economic Surveys, 27(3), 570–588. Shapira, Z., & Venezia, I. (2001). Patterns of behavior of professionally managed and independent investors. Journal of Banking & Finance, 25, 1573–1587. Shiller, R. J. (2000). Irrational exuberance. Princeton: Princeton University Press. Smith, V. L., Suchanek, G. L., & Williams, A. W. (1988). Bubbles, crashes, and endogenous expectations in experimental spot asset markets. Econometrica, 56(5), 1119–1151. Sto¨ckl, T., Huber, J., & Kirchler, M. (2015). Multi-period experimental asset markets with distinct fundamental value regimes. Experimental Economics, 18, 314–334. doi:10.1007/s10683-014-9404-1. Sto¨ckl, T., Huber, J., & Kirchler, M. (2010). Bubble measures in experimental asset markets. Experimental Economics, 13, 284–298. Sutter, M., Huber, J., & Kirchler, M. (2012). Bubbles and information: An experiment. Management Science, 58, 384–393. Van Boening, M. V., Williams, A. W., & LaMaster, S. (1993). Price bubbles and crashes in experimental call markets. Economics Letters, 41(2), 179–185. Van den Steen, E. (2004). Rational overoptimism (and other biases). American Economic Review, 94, 1141–1151. Weinstein, N. (1980). Unrealistic optimism about future life events. Journal of Personality and Social Psychology, 39, 806–820.
123