# Investment performance and emotions: an international study

Haim Levy (School of Business Administration, The Hebrew University of Jerusalem, Jerusalem, Israel)

ISSN: 1086-7376

Article publication date: 30 May 2019

Issue publication date: 21 June 2019

## Abstract

### Purpose

The purpose of this paper is to expand the peer effect analysis to investments in the stock market, where neither direct competition nor interaction with other investors exists.

### Design/methodology/approach

A total of 772 subjects dwelling in six countries completed a questionnaire about their satisfaction with the performance of their hypothetical investment in the stock market. They were informed about the performance of the local stock market and the performance of their peer group, referred to in the questionnaire as their “friends.”

### Findings

Only 5 per cent of subjects are indifferent to their friends’ investment performance, as advocates by expected utility paradigm. Most subjects are happier when their friends earn lower rather than higher returns. On average, subjects are better off losing rather than gaining money as long as their friends lose more money, which violates the univariate monotonicity axiom. A negligible number of subjects exhibit a consistent favorable response, which is a necessary condition for pure economic altruism. Hostility is greater in less-wealthy countries. No link is found with regard to economic inequality.

### Originality/value

This paper shows that when a conflict between absolute wealth and relative wealth arises, the latter dominates, even when the comparison is not with an opponent or a colleague but with the subject’s friends. The astonishing result is that subjects prefer having less wealth as long as their friends lose more, despite no direct competition between subjects as in ultimatum games and despite the performance being equal to market performance.

## Citation

Kaplanski, G. and Levy, H. (2019), "Investment performance and emotions: an international study", Studies in Economics and Finance, Vol. 36 No. 1, pp. 32-50. https://doi.org/10.1108/SEF-11-2017-0311

## Publisher

:

Emerald Publishing Limited

## 1. Introduction

As generally people prefer to have more wealth, it is common in economics to assume that the first derivative of the utility function is positive, i.e., U′(w) ≥ 0, where w stands for wealth and U for the utility function. In actual situations, however, the utility function may be a function of several factors, depending apart from wealth on other factors such as health and moods. When wealth is considered, several studies suggest the what is relevant is the relative wealth rather than the absolute wealth, hence the utility function type is U (w/wÆ) or the bivariate utility function U(w, wÆ), where w stands for the individual wealth and wÆ stands for the average wealth in the economy. Relying on these two variables may have economic justification also in the classic expected utility paradigm: Suppose the one’s income increases by 5 per cent while the average income in the economy increases by, say, 20 per cent, which increases demands inducing an inflation of 10 per cent. In such a case, the individual’s wealth increases in nominal terms but decreases in real terms, justifying having a utility function which depends on relative terms. In this paper we add additional factor, the emotions factor denoted by e, which may be affected by a wide spectrum of emotions such as ego and self-esteem and altruism tendency. Thus, we have the multivariate preference U (w, wÆ, e), where e stands for emotions. In order to concentrate on the effect of emotions on choices, which cannot be rationalized by economic reasons, in this study w/wÆ is held constant such that the effect of average wealth in the economy is neutralized. Thus, the multivariate preference is reduced to the bivariate preference U (w, e).

In the main experiment, 496 subjects dwelling in six countries completed a questionnaire about their satisfaction with the performance of their hypothetical investment in the stock market. They were informed about the performance of the local stock market and the performance of their peer group, referred to in the questionnaire as their “friends.” As explained above, the macroeconomic factor is neutralized by equating the performance of the subjects to the performance of the local stock market index such that w/wÆ is held constant. This way we isolate the impact of the investment performance of the subjects’ friends on their happiness with their own investment performance. We examine whether more wealth resulting from the investment in the stock market (by both the individual and the average investor) is preferred when emotions also affect welfare. As we shall show, in this framework we may have U (w2, e2) < U (w1, e1), where w2 > w1 and e2 ≠ e1, i.e. the subject is happier with less rather than more wealth implying that emotions is a dominating factor affecting welfare.

The effect of other people’s wealth on one’s own welfare is well known. Numerous experimental studies have shown that in some situations people make economic decisions that take others’ wealth into consideration, particularly the wealth of the peer group with which people interact socially or economically (Xiong et al., 2016). Preference that is also a function of the peer group is well documented in relation to social preference (Loewenstein et al., 1989; Charness and Rabin, 2002), the “habit formation consumption” literature (Sundaresan, 1989; Constantinides, 1990) and “catching up with the Joneses” preferences (Abel, 1990). In ultimatum games with competition between the players, emotions such as jealousy and social values such as inequality aversion are found to be sufficiently strong to induce people to give up money to influence the wealth level of the opponent in the game.

The main contribution of this study to the existing literature is by expanding the analysis of the peer effect to satisfaction from the investment performance in the stock market. In particular, we analyze the subjects’ welfare owing to investment performance relative to the investment performance of their friends, where the stock market performance is neutralized. Unlike ultimatum games, with stock market investment there is neither direct competition nor strong interaction with other investors. Moreover, the comparison is not to the opponent in a game or a colleague but rather to the subject’s friends, to examine the allegedly positive perspective of the peer effect. Therefore, our null hypothesis is that investors are mainly self-interested with regard to their wealth resulting from their investment performance and the performance of their friend does not affect their welfare.

Figure 1 summarizes the main results. Figure 1(a) presents the cumulative distribution functions (CDFs) of the reported satisfaction from two investment scenarios: 10 per cent gain and 10 per cent loss of both the subject and the local stock market. Generally, a shift to the right implies higher degree of satisfaction. As the two CDFs do not cross, for any satisfaction level the percentage of subjects who selected this satisfaction level or a higher one is higher for the 10 per cent gain than for the 10 per cent loss. Thus, with no information on the subject’s friends the results conform to the monotonicity axiom as the utility of the typical subject increases with wealth, i.e.

U($27,500)>U($22,500).

This indicate also that the subjects have read carefully the questionnaire and did not filled it randomly, which increase the reliability of the results.

Figure 1(b) presents the CDFs of reported satisfaction from the two investment scenarios with additional information of either 50 per cent gain or 50 per cent loss for the subject’s friends, respectively. Curve 1 is located to the left of Curve 2, implying that the subjects are generally happier with losing 10 per cent rather than gaining 10 per cent, as long the performance of their friends is worse. Thus, with the information on the subject’s friends the typical result is:

U($22,500, wF=$12,500)>U($27,500, wF=$37,500)
where wF stands for the friends’ wealth resulting from the same investment amount ($25,000). This means that the typical subject is happier with having$22,500 rather than $27,500 as long her friends’ wealth decreases, which implies a violation of the monotonicity axiom once emotions are incorporated. Consistent with the previous literature (Easterlin, 1995), we find that both absolute wealth and relative wealth determine welfare. In particular, here the market average wealth is neutralized and what left is relativeness of wealth to the closer group of friends. We show that when a conflict between absolute wealth and wealth relative to friends arises, the latter dominates. The astonishing (and we must add sad) result revealing that subjects generally prefer having less wealth as long their friends lose more than they lose, despite no direct competition between subjects as in ultimatum game and despite the subjects’ performance being equal to market performance. Unfortunately, even in this case it seems that hostility and jealousy dominate, whereas potential altruism is negligible. Alternatively, the subjects may be less happy with the relative success of their friends simply because of low self-esteem, as their friends have better invested. Although we are unable to disentangle between jealousy and low self-esteem, the strong role that emotions play in determining the welfare from achieved wealth is dramatic. In the univariate expected utility paradigm people are found to be irrational as they more satisfied with less over more wealth. To further characterize the peer group effect and check for robustness, we compare the results in different countries. In line with studies arguing that competition for resources affects altruism (West et al., 2002; Giron et al., 2004), we find that the hostility (or self-esteem effect) of the subjects toward their friends is greater in countries that are less rich. No similar link is found with regard to economic inequality. The structure of this paper is as follows. Section 2 presents the experiment and the sample. Section 3 reports the main results. Section 4 analyzes the differences across countries and reports a few robustness checks. Section 5 concludes. The questionnaire is presented in the Appendix. ## 2. The experiment In the main investment experiment, several hundred business students dwelling in the USA, Germany, Hong Kong, Israel, Turkey and China reported on their welfare (degree of satisfaction or utility) in various investment performance situations. The subjects faced hypothetical financial gains or losses on investments in the stock market, either in isolation or with additional information on the gains or losses faced by their friends. The subjects reported their degree of satisfaction on a scale of 1 (low satisfaction) to 7 (high satisfaction). We assume that higher reported satisfaction is associated with higher welfare. As in the experiment reported below the subjects do not make choices but rather report their degree of satisfaction (from 1 to 7), it is impossible to have a payoff which is directly related to their answers to the questionnaire. However, as in experimental studies there are always some subjects who improperly fill the questionnaire without even reading it (“just to dispense with the survey”) we use the following procedure to eliminate those subjects from our study: we provide the subjects two simple alternate scenarios of gaining 10 per cent (Scenario G) and losing 10 per cent (Scenario L). Any subject reporting a higher level of satisfaction with a loss than with a gain would fall into the inattentive group. In the analysis below, we first excluded 67 subjects (out of 563 subjects) who reported greater satisfaction with a loss than with a gain. However, as the answers corresponding to those 67 subjects in the other tasks (which are relevant to our study) reveal random patterns, the main results of this study are robust to the inclusion of them in the analysis. Table I reports the descriptive statistics of the 496 business students included in the main analysis, as well as the additional 276 subjects surveyed in the robustness tests. The average age in the main sample is 26, with a relatively high average age of subjects from Israel and Turkey (28 and 34, respectively). The subjects are generally spread across genders, with the exception of a majority of female subjects in Hong Kong. To account for age and gender differences across countries, we conducted a multivariate regression analysis. In the questionnaire, two questions examined the subjects’ response to a gain scenario and two other questions examined the response to a loss scenario, one with and another without additional information on the performance of the investments made by the subject’s friends and by the stock market index. Table II summarizes the four scenarios, which are described below: • Scenario G (Gain): The value of the subject’s portfolio increases from$25,000 to $27,500 (+10 per cent), and the main stock market index in the subject’s country increases at the same rate. • Scenario GS (Gain Smaller than the gain of the subject’s friends): The value of the subject’s portfolio and the main stock market index increase exactly as in Scenario G, whereas the portfolios of friends appreciate, on average, by 50 per cent, from$25,000 to $37,500. • Scenario L (Loss): The value of the subject’s portfolio decreases from$25,000 to $22,500 (–10 per cent), and the main stock market index decreases at the same rate. • Scenario LS (Loss Smaller than the loss of the subject’s friends): The value of the subject’s portfolio and the market main stock index decrease exactly as in Scenario L, whereas the subject’s friends’ portfolios depreciate, on average, by 50 per cent, from$25,000 to $12,500. The various scenarios allow us to explore and control the following questions: • Do investors in the sample prefer having higher rather than lower wealth levels with no information about their friends? This trivial question enables us to establish the reliability of our experiment. • Is there a peer effect (negative or positive) in case of investment beyond the general stock market performance effect when there is no interaction, let alone direct competition, between the subject and the peer group? • Are investors happier with an increase in their friends’ wealth with no change in their own wealth and no expectation of direct or indirect reciprocal reward? The answer to this question will shed light on the proportion of subjects who may support the existence of pure altruism. • What is the proportion of subjects who adhere to the classic univariate utility paradigm and, at the other extreme, what is the proportion of the subjects who violate the classic univariate expected utility’s monotonicity axiom, as they are willing to lose money as long as their so-called friends lose more money? These questions will shed light on the importance of the performance of the peer group in the case of investments. • As behavioral factors such as altruism may be associated with genetic (Cesarini et al., 2009) and cultural factors (Bell et al., 2009), as well as potentially being affected by the degree of competition for resources (West et al., 2002; Giron et al., 2004), we also explore whether the results vary across countries depending on economic wealth, measured by the gross domestic product (GDP) per capita, and economic inequality, measured by the GINI index for inequality per capita. To obtain unambiguous answers to the above questions, we must disentangle the impact of the information concerning the gains/losses of the subjects’ friends from other potential effects. Therefore, the study also controls for the following factors: • In the experiment there is neither kin selection nor direct or indirect reciprocal reward, which allows us to examine pure altruism and to distinguish it from reciprocal altruism. • Anonymity guarantees that the questionnaire does not affect the social status of the subject. • The questionnaire concealed its goals by apparently focusing on investments. Thus, the subject is not affected by self-image, which is probably biased toward virtues such as altruism. • In the main tests, investments are made hypothetically by a professional investment manager. Thus, the subject cannot affect the results and does not have explicit responsibility for them. This mitigates the possibility that self-image and ego issues related to the subject’s own investment abilities affect the reported level of satisfaction. The low degree of satisfaction with relatively low performance of the subject, accompanied with relatively high performance of the subject’s friend, may reflect anger on the portfolio manager or low self-esteem over choosing the wrong portfolio manager. However, in a separate test in which investments are made hypothetically by the subjects themselves, we find that the results do not depend on whether investments are self-made or made by a third party. Therefore, the role of the portfolio manager is not important in determining welfare and as explained above we conclude that self-esteem, jealousy or both are the main explanation for the obtained results. • To neutralize the effect of gains and losses relative to the benchmark, in all scenarios the subject’s portfolio performance is equal to the performance of the benchmark market portfolio in the subject’s country. Thus, in all cases the subjects keep their share in the average welfare as measured by the stock market main index. • The gain and loss questions were deliberately mixed up and there were two versions of the questionnaire with the questions in a different order (G, LS, GS, L and LS, G, L, GS). This mitigates the possibility that the order of questions will affect the results and makes it more difficult to follow the relations between the questions. • The questionnaire included a fifth “logical” investment question given in the middle of the questionnaire. This control question separated similar questions and concealed the exact goal of the questionnaire. Finally, in separate robustness checks, in addition to the 496 subjects, we include four control groups as follows: • Neutralizing the impact of the questionnaire structure: The first group consisted of 85 additional subjects who completed only a two-scenario partial questionnaire (G, LS or GS, L). Those subjects were not affected by the questionnaire structure and the complementary scenarios (i.e. with or without the information on friends), as they did not see the complementary questions at all. • Neutralizing the impact of the portfolio manager: This group consisted of 53 subjects who faced the same scenarios as in the main questionnaire, only this time they made their own investment decisions with no portfolio manager. As we obtain very similar results with and without a portfolio manager, we rule out anger on the portfolio manager as the explanation for the results. • Generalizing the results for non-business students: This group consisted of 63 non-business students (logistics studies or education studies). • Check for the impact of financial employment experience: This group consisted of 86 financial professionals who work in the financial industry. After excluding the 5, 5 and 6 subjects from groups 2, 3 and 4 above, respectively, who reported higher satisfaction from a loss than from a gain, we have a sample of 53 self-made investors, 58 non-business students and 80 subjects with financial employment experience (see Table I). Finally, there was no financial payoff in this experiment. The main reason for that is that the subjects report on the degree of satisfaction from various investment scenarios but did not make any choice. Therefore, one cannot attach monetary payoff as no choices are made. To overcome this difficulty, we added a simple question in which the subjects face either 10 per cent gain or 10 per cent loss. All subjects who reported higher satisfaction with a loss than with a gain (83 subjects out of 772 subjects) probably did not pay attention to the tasks and therefore are eliminated from the study. We stress that the strong result that subjects are happier with a loss as long as their friends’ loss is larger is obtained with no financial payoff, a case where the subjects can be generous and reveal altruism as they do not loss real money. Therefore, we suspect that with financial payoff the results would be even stronger. ## 3. Results Table III summarizes the main results. As expected, in a Robinson Crusoe economy, the average satisfaction is significantly higher with a gain (4.69) than with a loss (3.01) and the difference is highly significant (t = 24.4; p < 0.001). This confirms that most subjects paid attention to the questionnaire and did not complete it just to get rid of it. The first column shows that the information about the friends’ gains significantly reduces the average satisfaction from the subject’s own gain, from 4.69 to 3.14 (t = 19.86; p < 0.001). This is despite the fact that the subject faces identical gains in the two scenarios, which are also equal to the return on the market portfolio. The other columns show that the satisfaction of 76 per cent of subjects decreases owing to friends’ gains. Only 13.9 per cent of subjects are indifferent, as the classic univariate utility paradigm advocates, and 10.1 per cent exhibit a favorable response of higher satisfaction because of the gains of friends. In a similar manner, but in the opposite direction, information about the friends’ losses significantly increases the satisfaction from the subject’s own loss, from 3.01 to 3.96 (t = 12.49; p < 0.001). In the case of a loss, the satisfaction of 56.5 per cent of the subjects increases owing to the friends’ losses and 26.4 per cent of subjects are indifferent. Finally, 17.1 per cent exhibit a favorable response toward their friends as their satisfaction decreases when learning about their friends’ losses. Note that inequality aversion cannot explain the results for losses as according to this explanation we would expect to have a lower mean satisfaction in Scenario LS than in Scenario L because of the increase in inequality. The striking result that emerges from Panel C of Table III is that only 3.2 per cent of the subjects exhibit a favorable response under both gain and loss outcomes. Their satisfaction increases with the gain of their friends and it decreases with their loss. Those subjects constitute the maximum percentage of pure economic altruists in our sample. Nevertheless, they are not necessarily pure economic altruists as they do not donate anything to their friends. The other subjects cannot be pure altruists as their welfare either decreases or remains unchanged with an increase in their friends’ wealth, even though the increase in the friends’ wealth is not taken from their own wealth. Therefore, their welfare would have decreased if wealth had been transferred to their friends. Another interesting result is that only 5 per cent of subjects are indifferent in both gain and loss scenarios, i.e. their welfare is determined solely by their own wealth, as advocated by the classic univariate expected utility paradigm. According to the monotonicity axiom, satisfaction from a gain should be trivially higher than satisfaction from a loss. Indeed, with no information about the performance of the peer group, the average satisfaction in Scenario G is significantly higher than that in Scenario L. The most interesting result in Table III is that the impact of the information regarding the performance of the subjects’ friends is sufficiently strong to induce a violation of the monotonicity axiom. The average satisfaction in Scenario LS (3.96) is significantly higher than that in Scenario GS (3.14), despite the fact that Scenario LS represents a 10 per cent loss whereas Scenario GS represents a 10 per cent gain. Thus, on average, subjects are happier having$22,500 rather than $27,500, provided that their friends lose more money. This average difference in satisfaction (0.82) is highly significant (t = 6.86; p < 0.001). While similar phenomena have been reported in situations such as ultimatum games, to the best of our knowledge, no such finding has been reported within the investment framework, where the success of one party does not affect the success of the other party. More important, this phenomenon encompasses the majority of subjects: 58.3 per cent of the subjects reveal strict violation of the monotonicity axiom as they report higher satisfaction in Scenario LS than in Scenario GS, and an additional 12.3 per cent reveal weak violation, reporting the same satisfaction in both Scenarios LS and GS (see the bottom of Table III). ## 4. International comparison We next compare the results across subjects from different countries. In addition to robustness checks, this comparison sheds some light on the factors that affect the observed peer effect in investments. Table IV summarizes the results for each country separately, as well as the combined results of the three countries with higher past decade GDP (Hong Kong, the USA and Germany) and the three countries with past decade lower GDP (China, Israel and Turkey), where GDP is measured as the average GDP per capita in the last 10 years. Although, as we shall show below, the GDP variable is found to distinguish significantly between subjects’ characteristics across countries, we stress at the outset that this variable may be a proxy for other variables related not only to the economy but also to other factors such as culture. Consistent with the previous results, in all six countries a gain for the subject’s friends significantly decreases the average satisfaction from the subject’s own gain (see the reduction in satisfaction between Columns 1 and 2). For example, in the USA. The average satisfaction in Scenario G of 5.16 decreases to 3.39 in Scenario GS (t = 7.96; p < 0.001). Consistently, a loss incurred by the subject’s friends’ increases average satisfaction (see Columns 3 and 4). This increase is significant in all countries but Hong Kong. While the direction of the peer effect is similar in all counties, the intensity of this effect differs across countries. Indeed, in all scenarios the ANOVA analysis, at the bottom of the table, rejects the hypothesis of equal means in all countries. In four countries (the USA, China, Turkey and Israel), the satisfaction in Scenario LS (Column 4) is higher than that in Scenario GS (Column 2), implying a violation of the monotonicity axiom. The differences in satisfaction from Scenarios LS and GS, reported in Column 5, are not significantly different from zero in the case of Hong Kong/USA/Germany, but are positive and highly significant in the case of China/Turkey/Israel. Notice that in the case of Hong Kong/USA/Germany, the insignificant difference in satisfaction from gaining or losing 10 per cent also implies a violation of the monotonicity axiom in terms of the classic univariate utility paradigm. To check the association between countries and the violation of the monotonicity axiom, we use Scheffe’s test for homogeneous subsets. As can be seen from Column 5, the monotonicity violation in Hong Kong/USA/Germany and China/Turkey/Israel may indeed come from two separate homogeneous subsets (a and c, respectively). Consistently, under any partitioning, Turkey and Israel on the one hand and Hong Kong on the other hand belong to two different subsets. However, household inequality as measured by the GINI index (see the columns on the left-hand side) cannot explain the differences across countries because this index is particularly large in China, Israel, Hong Kong and the USA, which do not belong to the same subset. In line with the Hong Kong/USA/Germany and China/Turkey/Israel partition, the assumption of equal mean across the two groups of countries is significantly rejected in all scenarios except Scenario L (see the last row). The last column in Table IV reports the percentage of subjects exhibiting a favorable response toward their friends in both gain and loss scenarios, which is a necessary condition for pure altruism. The percentage of subjects who are potentially pure altruists in Hong Kong/USA/Germany is very small and in China/Turkey/Israel it is close to zero. As subjects from Israel and Turkey are older, on average, than the other subjects and as the proportion of female subjects is the greatest in Hong Kong, these demographic differences may account for the differences across countries. To control for gender and age biases across countries, and to verify that those who did not answer the control question correctly and might not have paid sufficient attention to the questionnaire do not induce the results, we run the following probit regression: (1) ECON_ALTRi=β1COUNTRYi+β2GENDERi+β3AGEi+β4ATT_QUESTIONi+εi where ECON_ALTRi is an indicator of economic altruism; COUNTRYi are dummies for the countries equal to one if subject i comes from that country and zero otherwise; GENDERi is a dummy for a female; AGEi is the age of subject i in years; ATT_QUESTIONi is a dummy for a correct answer to the attention question. ECON_ALTRi is defined in two alternative ways. According to the positive definition, it is defined as “possibly an economic altruist” and it is equal to one if subject i exhibits a favorable response to his or her friends either in the case of a gain or in the case of a loss. This is a soft definition in two respects. First, this definition requires a favorable response in only one scenario rather than in both scenarios. Second, as previously explained, a favorable response does not guarantee that the subject is an economic altruist, but rather that he or she may be an economic altruist in some situations. According to the alternative strict negative definition of “definitely not an economic altruist,” it is equal to one if subject i exhibits a violation of the monotonicity axiom. If the subject’s adverse response is sufficiently strong to induce irrationality in terms of the classic univariate utility paradigm, this subject is definitely not an economic altruist. Table V reports the regression results. The dependent variable in the first model is “possibly an economic altruist.” The Hong Kong/USA/Germany coefficients are negative and two of them are significant at the 5 per cent level. The China/Turkey/Israel coefficients are negative, highly significant (p < 0.001) and larger in absolute terms than those corresponding to Hong Kong/USA/Germany. Indeed, the null hypothesis of equal country coefficients is rejected at highly significant degree (p < 0.001). Thus, the probability of a favorable response, which allows this subject to be an economic altruist in some situations, is significantly smaller in China/Turkey/Israel. This difference is further confirmed in the second test, in which the joint dummy coefficient corresponding to subjects from China/Turkey/Israel is negative and highly significant. The results are even more profound in the last two tests, corresponding to the second model, in which the dependent variable is “definitely not an economic altruist.” While the Hong Kong/USA/Germany coefficients are either significant at the 5 per cent level or not significant, the China/Turkey/Israel coefficients are about twice as large and highly significant (p < 0.01). The null hypothesis of equal country coefficients is also rejected (p < 0.001). Thus, the probability of an adverse response that is sufficiently large to induce a violation of the monotonicity axiom (i.e. the subject is definitely not an economic altruist) is significantly larger in China/Turkey/Israel. This result is confirmed in the last joint test, for which the China/Turkey/Israel dummy coefficient is highly significantly positive (p < 0.001). The results in Table V show that the probability of a subject potentially (but not necessarily) being an economic altruist is significantly smaller in China/Turkey/Israel, and this result is not affected by gender, age or attention to the questionnaire. While there are no differences between males and females, the probability of subjects potentially being economic altruists increases significantly with age. The robustness tests in Table VI verify that the previous results are not affected by the questions’ crossover effects or the involvement of a third party (the portfolio manager) in the process and are not confined to business students. To avoid cross-country biases, these tests are conducted with subjects from the same country (Israel). The first test compares the main results reported above, corresponding to 133 Israeli business students who completed a standard full questionnaire, with the results corresponding to a control group of 85 subjects who completed one of two partial two-scenario questionnaires (G, LS or GS, L), so they cannot be affected by complementary questions. In all the scenarios, the hypothesis of equal means between the two groups of subjects is not rejected (Columns 1 through 4). For example, while the mean satisfaction from Scenario GS in the main group is 3.08, it is 3.18 in case of subjects who completed partial questions. The t-values for equal means is –0.41 indicating the null is not rejected. Furthermore, also in this group the average monotonicity violation (see Columns 5) is positive, large (0.92) and significant (t = 2.7). Thus, we conclude that crossover effects from complementary questions do not account for the main results reported in Table III. The second test compares main results with the results corresponding to 53 business students who faced the same hypothetical investment scenarios, but they are explicitly told that they manage their own portfolios while their friend managed their portfolios. Thus, investments do not involve professional portfolio manager. Once again, the main results are not significantly different from the results in the main test with portfolio manager. For example, the mean satisfaction from Scenario LS is 4.41, in comparison to 4.45 in the main test with a t-values of 1.21 for different means. Also similar, violation is significant and subjects on average prefer less (Scenario LS) over more (Scenario GS) as long as they exceed their friends. Although the violation is smaller in this case it shows that the main results are robust and do not stem from introducing the third-party portfolio manager (the t-values in this case are smaller owing to smaller sample size relative to the main test). This similar result rules out that anger on the portfolio manager as the explanation for the results. Therefore, the emotions which may explain the results that people are happier with less rather than more wealth is ego and self-esteem, jealousy or both. The next two tests in Table VI compare the results corresponding to the main test with 133 business students to those corresponding to the 58 non-business students and 80 subjects with financial industry experience. In all scenarios, the hypothesis of equal means across business students and non-business students is not rejected. For example, the mean satisfaction from Scenario GS in case of non-business students is 3.08 in comparison to 3.52 in the main test with a t-values of 1.88 for different means. Similarly, a significant monotonicity axiom violation exists in all cases (see the last column). Namely, we do not find significant differences between business and non-business students. Nevertheless, for those with financial industry experience, the effect of the investment performance of the subject’s friends is significantly reduced. The average monotonicity violation is significantly smaller (0.48 vs 1.38; t = 3.07) and the monotonicity violation is less significant (t = 2.14 vs t = 7.46). Thus, while the subjects with financial industry experience are also affected by their friends’ success or failure, they are less prone to being affected by the results of others and the extreme result of implied violation of the monotonicity axiom is smaller. ## 5. Concluding remarks Defining the utility function as a function of the individual’s wealth relative to the average wealth in the economy may be economically rationalized. In this study, we assume that the investor wealth increases exactly as the average wealth in the economy, hence we neutralize this effect. However, the individual’s wealth changes relative to the individual’ friends’ wealth. Thus, the peer effect in our study corresponds to the individual’s friends’ wealth. Our study involves tasks dealing with investment in the stock market. As in our hypothetical investment experiment there is neither direct competition nor interaction with the peer group, the null hypothesis is that there is no peer effect and if it exists it is positive, namely, the subject’s welfare increases with the investment success of her friends. Collecting questionnaires from several hundred subjects dwelling in six countries, we find that only 13.9 per cent of the subjects in the positive outcomes domain and 26.4 per cent of the subjects in the negative outcomes domain are indifferent to their fiends’ investment performance, as advocated by the classic univariate expected utility paradigm. Only 5 per cent are indifferent to their friends’ investment performance in both the positive and the negative domains. Unfortunately, most subjects are happier when their friends have lower rather than higher investment returns. Moreover, on average, subjects are better off in terms of satisfaction losing money rather than gaining money as long as their so-called friends lose more money, which implies a violation of the univariate monotonicity axiom. Consistently, the number of subjects who exhibit a consistent favorable response toward their friends, which is a necessary condition for pure economic altruism, is negligible. International comparison shows a very similar effect in all countries. However, in line with the existing theory, hostility (or low self-esteem) is greater in countries that were less wealthy in the last decade. Quite surprisingly, no clear link is found with regard to economic inequality. Although past GDP per capita may be a proxy for other variables, this suggests that the differences between countries are probably related more to competition for resources than to culture. In sum, the results are very negative: the welfare of most subjects is negatively correlated with the investment success of their friends. Although we are unable to determine whether jealousy or self-esteem dictates these strong results, we show that the effect of emotions on welfare from investment is very strong. While in isolation the subjects typically prefer more over less wealth with emotions this expected monotonicity is strongly violated. ## Figures #### Figure 1. Violation of the monotonicity axiom ## Table I. Sample Main sample (business students) Control groups Groups USA Germany Hong Kong Israel Turkey China Total Separate questions Subjects invest by themselves Non-business students Financial experience Total Number of valid questionnaires 57 73 20 133 27 186 496 85 53 58 80 772 Non-valid questionnaires 5 4 6 11 2 27 67 0 5 5 6 83 Average age 23 23 23 28 34 25 26 28 28 33 36 28 Males (%) 67 34 15 57 63 39 47 29 49 59 29 45 Notes: The table reports descriptive statistics for the sample, which consists of 563 undergraduate and graduate business administration students from six countries, as well as additional 292 subjects in four control groups: 85 business students who completed partial questionnaires, 58 subjects who completed a questionnaire with the same hypothetical investment scenarios but with the assumption that they invest by themselves rather than by a portfolio manager, 63 non-business students and 86 subjects working in the financial industry. Valid questionnaires do not include those in which subjects reported higher satisfaction from a loss (Scenario L) than from a gain (Scenario L), implying that they did not pay attention while completing the questionnaire ## Table II. Questionnaire 1. Scenario G (10% gain) 2. Scenario GS (10% gain, friends’ 50% gain) 3. Scenario L (10% loss) 4. Scenario LS (10% loss, friends’ 50% loss) Portfolio value (thousand US$) at Start End Start End Start End Start End
Market portfolio in the subject’s country 25 27.5 25 27.5 25 22.5 25 22.5
Subject’s portfolio 25 27.5 25 27.5 25 22.5 25 22.5
Friends’ portfolios No information 25 37.5 No information 25 12.5
Notes:

The table reports the four scenarios presented to the subjects in the questionnaire: Scenario G (10% gain), Scenario GS (10% gain in comparison to friends’ 50% gain), Scenario L (10% loss) and Scenario LS (10% loss in comparison to friends’ 50% loss)

## Table III.

Response to information on friends’ performance

Response to information on friends’ wealth
Scenario Mean satisfaction Adverse response Indifference Favorable response
Gain A. Response to friends’ gainsa
Scenario G (10% gain) 4.69 Number of subjects 377 (76.0%) 69 (13.9%) 50 (10.1%)
Scenario GS (10% gain, 50% friends’ gains) (paired t-value for G and GS equal means) 3.14 (t = 19.86**)
Loss B. Response to friends’ lossesb
Scenario L (10% loss) 3.01 Number of subjects 280 (56.5%) 131 (26.4%) 85 (17.1%)
Scenario LS (10% loss, 50% friends’ losses) (paired t-value for L and LS equal means) 3.96 (t = 12.49**)
Combined results C. Same response to friends’ gains and lossesc
(paired t-value for G and L equal means) (t = 24.40**)
Number of subjects 234 (47.2%) 25 (5.0%) 16 (3.2%)
Violation of the monotonicity axiom
Average violation (mean satisfaction in Scenario LS less than in GS) (paired t-value for violation) 3.96-3.14 = 0.82 (6.86**)
Number of subjects revealing strict violation 289 (58.3%)
Number of subjects revealing weak violation 61 (12.3%)
Notes:

The table reports the average satisfaction and percentage (out of 496 subjects) of adverse, indifferent and favorable responses to the information about friends’ wealth. The last three rows report the difference in mean satisfaction between Scenario LS and Scenario GS, which, if positive, represents a violation of the monotonicity axiom, and the t-values for non-violation;

* and ** denote significance at the 5% and 1% levels, respectively;

a

If friends’ gains decrease satisfaction, it is defined as an adverse response and vice versa;

b

If friends’ losses increase satisfaction, it is defined as an adverse response and vice versa;

c

The other 44.6% reveal different responses to the information on friends’ gains or losses

## Table IV.

Response to the information on friends’ performance by country

Mean satisfaction Monotonicity axiom violation
Country 10-year mean GDP per capita (‘000 US$) 10-year mean GINI inequality index per capita 1. Scenario G (10% gain) 2. Scenario GS (10% gain, friends’ 50% gains) (Paired t-value) 3. Scenario L (10% loss) 4. Scenario LS (10% loss, friends’ 50% losses) (Paired t-value) 5. Average violation (Scenario LS less GS) (Paired t-value for violation) % of favorable responses in both gain and loss scenarios (%) USA 44.3 41.11 5.16 3.39 (7.96**) 3.05 3.77 (3.47**) 0.39abc (0.96) 5.3 Germany 37.0 31.32 4.90 3.33 (6.47**) 2.62 3.08 (2.18*) −0.25ab (−0.76) 5.5 Hong Kong 28.9 41.20 5.20 4.25 (2.33*) 3.50 3.60 (0.38) −0.65a (−1.03) 15.0 Combined upper half 5.04 3.47 2.90 3.41 −0.06 (0.26) 6.7 Israel 24.7 41.94 4.83 3.08 (12.53**) 3.18 4.45 (9.35**) 1.38bc (7.46**) 0.8 Turkey 7.9 40.08 4.48 2.07 (7.94**) 2.74 4.15 (4.51**) 2.07c (4.22**) 0.0 China 3.0 42.44 4.3 3.08 (10.47**) 3.01 4.03 (7.81**) 0.95abc (4.90**) 2.7 Combined lower half 4.53 3.00 3.00 3.05 4.20 1.20 (9.06**) 1.7 ANOVA F for countries’ equal means (5.80**) (4.66**) (2.87*) (6.35**) (7.48**) Student’s t-test for two halves’ equal means (−4.27**) (−2.76**) (1.24) (4.64**) (4.67**) Notes: Columns 1-4 report the mean satisfaction in the four investment scenarios. Column 5 reports the mean difference between Scenarios LS and GS, which, if positive, represents a violation of the monotonicity axiom. The superscripts a,b and c denote homogeneous subsets of countries according to Scheffe’s test. The last column reports the percentage of subjects, revealing a consistently favorable response to the information on friends’ wealth, which is a prerequisite for pure altruism. The lower rows report the ANOVA results for different means across countries and student’s t-test for the upper and lower halves’ GDP per capita countries’ equal means; * and ** represent significance at the 5 and 1% levels, respectively ## Table V. Regression analysis across countries Dependent variable Model 1: potentially an altruist Model 2: definitely not an altruist Variable Coeff Sig Coeff Sig Coeff Sig Coeff Sig Constant −0.89 0.03 0.58 0.12 Upper GDP USA −1.34 0.00 0.83 0.03 Germany −1.00 0.02 0.80 0.05 Hong Kong −0.94 0.06 0.59 0.21 Lower GDP Israel −1.96 0.00 1.55 0.00 Turkey −2.57 0.00 1.71 0.00 China −1.38 0.00 1.13 0.01 Lower GDP dummy −0.48 0.00 0.50 0.00 Demographic variables Gender (female) −0.04 0.80 0.07 0.57 −0.02 0.84 −0.12 0.30 Age (years) 0.05 0.00 0.03 0.00 −0.03 0.01 −0.01 0.24 Correct attention question −0.24 0.08 −0.29 0.03 −0.13 0.27 −0.14 0.26 Notes: The table reports the results of the following probit regression: ECON_ALTRUISMi=β1COUNTRYi+β2GENDERi+β3AGEi+β4ATT_QUESTIONi+εi, where ECON_ALTRUISMi is an indicator of economic altruism; COUNTRYi are dummies for the country of subject i; GENDERi is a dummy for a female; AGEi is the age of subject i in years; ATT_QUESTIONi is a dummy for a correct answer to the attention question. In Model 1, the dependent variable, ECON_ALTRUISMi, is positively defined as equal to one if subject i reveals a favorable response to friends’ wealth, indicating that this subject may be an economic altruist in some situations. In Model 2, it is alternatively negatively defined as equal to one if subject i reveals an implied violation of the monotonicity axiom, indicating that this subject is definitely not an economic altruist. The data include 496 business students from six countries. The significance level is based on robust (Huber–White) standard errors ## Table VI. Robustness tests Mean satisfaction Monotonicity axiom violation Experiment No. of subjects 1. Scenario G (10% gain) 2. Scenario GS (10% gain, friends’ 50% gains) 3. Scenario L (10% loss) 4. Scenario LS (10% loss, friends’ 50% losses) 5. Average violation (Scenario LS less GS) (Paired t-value for violation) Question crossover effects Standard questionnaire 133 4.83 3.08 3.18 4.45 1.38 (7.46**) Partial questionnaires 40 + 45 = 85 4.33 3.18 3.60 4.10 0.92 (2.70**a) (t-value for equal means) (1.96) (−0.41) (−1.76) (1.15) No portfolio manager Standard questionnaire 133 4.83 3.08 3.18 4.45 1.38 (7.46**) Subject invest by themselves 53 5.08 3.58 2.98 4.41 0.53 (1.69*) (t-value for equal means) (1.06) (2.10*) (0.87) (1.21) Non-business students Business students 133 4.83 3.08 3.18 4.45 1.38 (7.46**) Non-business students 58 5.07 3.52 2.95 4.41 0.90 (2.64*) (t-value for equal means) (−1.07) (−1.88) (1.05) (0.14) Financial industry experience Business students 133 4.83 3.08 3.18 4.45 1.38 (7.46**) Financial industry experience 80 4.79 3.61 2.84 4.09 0.48 (2.14*) (t-value for equal means) (0.25) (2.50*) (−1.76) (−1.59) Notes: The table compares the main results to the results obtained for the four control groups: subjects who completed partial questionnaires without complementary questions, subjects who invested by themselves rather than by professional portfolio managers, subjects who were not business students and subjects who worked in the financial industry; * and ** denote significance at the 5 and 1% levels, respectively; a Independent samples ## Table AI. Control question Investment A value of investment (year-end) Probability Investment B value of investment (year-end) Probability$20,000 ¼ $25,000 ½$25,000 ¼

### Question 4 (Scenario GS)

Your portfolio manager (/You) invested a total amount of $25,000 for you. One year later, the value of your investment has increased to$27,500. Over the same year, an investment in the amount of $25,000 in the S&P 500 stock index has also increased to$27,500. Your portfolio manager managed the investment portfolios of most of your friends (/Your friends also invested similar amounts), and their investment to the value of $25,000 has increased over the same year, on average, to$37,500. Rank your satisfaction with the outcome from 1 to 7, where 1 signifies disappointment and very low satisfaction with the financial development described above, and 7 signifies happiness and very high satisfaction.

### Question 5 (Scenario L)

Your portfolio manager (/You) invested a total amount of $25,000 for you. One year later, the value of your investment has decreased to$22,500. Over the same year, an investment in the amount of $25,000 in the S&P 500 stock index has also decreased to$22,500. Rank your satisfaction with the outcome from 1 to 7, where 1 signifies disappointment and very low satisfaction with the financial development described above, and 7 signifies happiness and very high satisfaction.

## References

Abel, A.B. (1990), “Asset prices under habit formation and catching up with the Joneses”, American Economic Review, Vol. 80, pp. 38-42.

Bell, A., Richerson, P. and McElreath, R. (2009), “Culture rather than genes provides greater scope for the evolution of large-scale human prosociality”, PNAS, Vol. 106 No. 42, pp. 17671-17674.

Cesarini, D., Dawes, C.T., Johannesson, M., Lichtenstein, P. and Wallace, B. (2009), “Genetic variation in preferences for giving and risk taking”, Quarterly Journal of Economics, Vol. 124 No. 2, pp. 809-842.

Charness, G. and Rabin, M. (2002), “Understanding social preferences with simple tests”, Quarterly Journal of Economics, Vol. 117 No. 3, pp. 817-869.

Constantinides, G. (1990), “Habit formation: a resolution of the equity premium puzzle”, Journal of Political Economy, Vol. 98 No. 3, pp. 519-543.

Easterlin, R. (1995), “Will raising the incomes of all increase the happiness of all?”, Journal of Economic Behavior and Organization, Vol. 27 No. 1, pp. 35-47.

Giron, D., Dunn, D., Hardy, I. and Strand, M. (2004), “Aggression by polyembryonic wasp soldiers correlates with kinship but not resource competition”, Nature, Vol. 430 No. 7000, pp. 676-679.

Loewenstein, G.F., Thompson, L. and Bazerman, M.H. (1989), “Social utility and decision making in interpersonal contexts”, Journal of Personality and Social Psychology, Vol. 57 No. 3, pp. 426-441.

Sundaresan, S.M. (1989), “Intertemporally dependent preferences and the volatility of consumption and wealth”, Review of Financial Studies, Vol. 2 No. 1, pp. 73-89.

West, S., Pen, I. and Griffin, A. (2002), “Cooperation and competition between relatives”, Science, Vol. 296 No. 5565, pp. 72-75.

Xiong, H., Payne, D. and Kinsella, S. (2016), “Peer effects in the diffusion of innovations: theory and simulation”, Journal of Behavioral and Experimental Economics, Vol. 63, pp. 1-13.

## Corresponding author

Haim Levy can be contacted at: mshlevy@mscc.huji.ac.il