The Ending Effect in Investment Decisions: The Motivational Need for an Emotionally Rewarding Ending

Abstract

The present study examined the power of endings on risky decision making. With four experiments, the changes in the individuals’ risk-taking tendencies were examined as the end of an investment decision task approached; the role of motivational shift toward emotional satisfaction in the ending effect was also explored. As predicted, participants who knew they were working on the last round of an investment task were more risk seeking than those who did not know (i.e., ending effect, Experiment 1). Experiments 2 through 4 examined the motivational mechanism of the ending effect. The results supported the notion that the motivation to pursue an emotionally rewarding ending leads to the ending effect. The present research complements existing motivational accounts of risk taking and suggests a new research direction of integrating factors associated with time perception of an approaching ending into existing models of risky decision making.

Keywords

ending effect motivation risky decision making emotional satisfaction socioemotional selectivity theory

Remembering that I’ll be dead soon is the most important tool I’ve ever encountered to help me make the big choices in life. . . Remembering that you are going to die is the best way I know to avoid the trap of thinking you have something to lose.

—Steve Jobs

Risk is known as the variance in possible outcomes. Risk taking can be defined as behaviors that involve some potential harm or loss, but also provide an opportunity to obtain rewards (Leigh, 1999). Psychologists and economists believe that people are generally risk averse (Benartzi & Thaler, 1995; Larrick, 1993; Pratt, 1964). However, there is less of a consensus on when and why people engage in risk taking. Several factors have been identified that contribute to risk taking, such as being in a loss frame of mind (Kahneman & Tversky, 1979, 1984) and macroeconomic inequality (Mishra, Hing, & Lalumiere, 2015). In recent years, a growing body of research suggests that motivational factors also play an important role in individuals’ risk taking (Florack, Keller, & Palcu, 2013; Kluger, Stephan, Ganzach, & Hershkovitz, 2004; Kühberger & Wiener, 2012; Larrick, Heath, & Wu, 2009; Mishra, 2014; Mishra & Fiddick, 2012; Sacchi & Stanca, 2014; Scholer, Zou, Fujita, Stroessner, & Higgins, 2010; Schwager & Rothermund, 2012; Zou, Scholer, & Higgins, 2014).

As an example, studies examining the role of self-regulatory goals in risk-seeking behavior showed that, promotion motivation could predict the individuals’ risk-seeking tendency in the domain of gains. Prevention motivation predicts the individuals’ choices between risky and conservative options in decision making when losing, depending on which option offers the possibility to return to the status quo (Scholer et al., 2010; Zou et al., 2014).

The present study attempts to contribute to existing motivational accounts of risk taking by examining how motivation shifts as individuals are approaching an ending may lead to increased risky decision making. Recent studies have suggested that individuals’ reactions tend to change toward the end of a series of tasks (Effron, Bryan, & Murnighan, 2015; Li & Epley, 2009; McKenzie et al., 2016; O’Brien & Ellsworth, 2012). In the domain of risky decision making, it has been suggested that individuals show increased risk taking at the end of a series of gambles (e.g., McKenzie et al., 2016); however, the underlying mechanism of this phenomenon remains unknown. In the current study, we first attempt to replicate a previous finding in which the perception that an ending is approaching would lead to increased risk taking; we name this phenomenon the “ending effect.” Second, we propose that socioemotional selectivity theory (SST, Carstensen, 2006; Carstensen, Isaacowitz, & Charles, 1999), which focuses on the effect of endings on basic human processes (i.e., motivation, cognition, and emotion), may provide a valuable motivational framework to help understand the underlying mechanism of this “ending effect.”

SST posits that when time perception shrinks, the individuals’ goal selection shifts from knowledge acquisition, which could bank for the future, to the regulation of emotional states, which benefit the immediate moment. Based on previous empirical studies examining the effect of perceived endings and findings from the current research, we suggest that this “ending effect” in risk taking may be caused by the increased motivational need for an emotionally rewarding ending, as predicted by SST.

Increased Risk Taking at the Ending: The Last Race Effect

Previous studies examining the betting phenomenon of horse races have suggested that individuals tend to be more risk taking toward the end of the racing day (Ali, 1977; Asch, Malkiel, & Quandt, 1982; Kopelman & Minkin, 1991; McGlothlin, 1956; Metzger, 1985). Other studies additionally showed that betting on long shots increases only in the last race (McGlothlin, 1956; Metzger, 1985), a condition referred to as “the last race effect” (McKenzie et al., 2016). More recently, this phenomenon was replicated in a set of laboratory experiments with a group of university students. The experiments used computer-based risky decision tasks simulating a day of horse racing (Experiment 3) and non-horse-race-based risky decision tasks (Experiments 1 and 2, McKenzie et al., 2016).

Taken together, there is robust evidence that individuals’ risk-taking tendencies significantly increases during the last round of a risky decision task, and most work has focused on the domain of horse races. We suspect that this phenomenon could be generalized to other non-horse-race-based risky decision tasks and we refer to this phenomenon as the “ending effect.” By manipulating participants’ perception of whether a task is or is not the end of a set of risky decision tasks, the present study’s first experiment attempted to replicate that perceived endings would increase risk taking in a risky decision task that do not involve a horse-racing scenario.

Underlying Mechanisms of the Ending Effect: Motivational Shift at the Ending

Assuming that the ending effect does exist in risky decision tasks beyond horse racing, why would the risk-taking tendency increase toward the end? McKenzie and colleagues (2016) conducted three well-controlled laboratory experiments to examine whether the last race effect is reference dependent: whether individuals show increased risk taking toward the ending depends on whether they were relatively higher or lower than their starting point. Participants were presented with risky decision tasks on a computer, and then made eight or 10 rounds of choices between two options: a riskier option and a safer option. The results showed that, regardless of previous wins or losses before the last round, participants in all three experiments consistently showed an increased preference for the riskier options in the last round.

Therefore, there might be other reasons, rather than previous wins and losses, that could account for the ending effect. Although McKenzie et al. (2016) could not speak to what kind of reference independent mechanism was at play, it was speculated that people might like to “go out with a bang” in which they save the most exciting or emotionally involving event for the last. This idea aligns with the predictions made by SST.

According to SST (Carstensen, 2006; Carstensen et al., 1999), the perception of remaining time plays a crucial role in motivation. When people are facing an ending, their perception of remaining time shifts from expansive to limited. This change in the time horizon leads to systematic changes in goal selection. Specifically, when time perception is limited, individuals are more motivated toward goals that could be realized in the short term. Under such conditions, individuals emphasize their emotional satisfaction at the present moment, and tend to be motivated toward emotion regulation. Consistent with this assertion, previous studies examining SST have demonstrated that motivation toward emotional satisfaction tends to increase toward the ending. In contrast, when individuals perceive their remaining time to be unlimited, they are more motivated toward goals that they could bank for the future, such as expanding breadth of knowledge (Carstensen et al., 1999) and connecting to more social partners (Carstensen, Gross, & Fung, 1997).

This motivational shift induced by changes in the time perspective could also lead to changes in cognitive processes. Both behavioral data and brain imaging studies suggest that individuals show attentional and memory preferences for positive stimuli over negative stimuli, a phenomenon referred to as the age-related “positivity effect” (for reviews, see Reed, Chan, & Mikels, 2014; Samanez-Larkin & Carstensen, 2011). Specifically, people with a limited future time perspective attend earlier and more to emotionally satisfying stimuli (Isaacowitz, Wadlinger, Goren, & Wilson, 2006a, 2006b; Mather & Carstensen, 2003, 2004, 2005). They also remember more positive stimuli than negative stimuli (Kennedy, Mather, & Carstensen, 2004; Schryer & Ross, 2014). Studies using brain imaging techniques to study patterns of neural activity have also yielded results consistent with the behavioral data above (e.g., Mather et al., 2004; for a review, see Samanez-Larkin & Carstensen, 2011). In short, shifts in the subjective sense of remaining time could cause a wide range of changes in people’s motivational and cognitive processes.

The motivational shift toward emotional satisfaction caused by the limited time perspective has been found among different age groups, and participants in the limited time perception condition included older adults (e.g., Isaacowitz et al., 2006a, 2006b), young people with terminal illnesses (e.g., Carstensen & Fredrickson, 1998), young people facing significant life events (e.g., Fung & Carstensen, 2006), and participants who were induced to perceive their remaining time as limited in laboratory settings (e.g., Fredrickson & Carstensen, 1990; Fung, Carstensen, & Lutz, 1999).

Although SST mainly makes predictions about how approaching the end of significant life phases affects motivation and cognition, some other evidence suggests that shifts in motivation and cognition could also be induced by more trivial phases, including the end of a series of laboratory tasks. In a recent study (O’Brien & Ellsworth, 2012), participants were asked to eat five chocolates one by one, each with a different flavor. Participants were randomly assigned to two conditions. In one condition, participants were aware which chocolate was the last one; whereas in the other condition, participants were not aware of the last chocolate. It was found that participants who knew they were eating the last chocolate enjoyed it more than those who did not know.

It appears that, regardless of the nature of the events, the perception that one is approaching the end of the current event is the most relevant factor to induce shifts in the motivational and cognitive processes (Davis & Hicks, 2013). Consistent with the findings above, an earlier study (Pennebaker et al., 1979) found that people rated opposite sex individuals who they met as more attractive as the time to decide whether or not to interact with them approached. It might be possible that the motivational shift to pursue an emotionally rewarding ending as predicted by SST led the participants to rate the final alternatives (chocolate or opposite sex individuals, respectively) as more desirable. Although this idea has not been directly examined, findings from both studies are consistent with this notion.

Overview of the Present Research and Hypotheses

There are two goals of the present research. First, we attempt to replicate the ending effect suggested in previous studies through another decision task paradigm that is associated with real-life consequences. It is hypothesized that participants would show increased risk taking toward the end of the decision task (Hypothesis 1). Second, the current study examines the underlying mechanism of the ending effect. We propose that the ending effect in risky decision making is caused by the individuals’ motivation to pursue an emotionally rewarding ending (Hypothesis 2).

Previous studies examining SST have demonstrated that individuals are more motivated toward emotional satisfaction as they are approaching an ending. In a risky decision task, individuals need to choose between a risky option and a safe option. According to SST, participants would be more likely to choose the option that could meet their goal for emotional satisfaction at the end of a series of risky decision tasks. The risky option is associated with a certain possibility to gain a large monetary reward, which leads to emotional satisfaction. Therefore, the risky option provides the possibility to meet the goal of emotional satisfaction. In contrast, the safe option is associated with a sure outcome with a small gain or no gain. Compared to the risky option, the safe option is relatively less able to meet the individuals’ need for emotional satisfaction. Thus, it is hypothesized that the ending effect is caused by the increased emotional motivation toward the ending, in which the emotional motivation is defined as the motivation for emotional satisfaction. The emotional motivation is heightened at the end of the task due to the shift in time perspective.

In the present research, four experiments were conducted to test the two hypotheses. The decision task paradigm was adapted from Shiv, Loewenstein, Bechara, Damasio, and Damasio (2005). Participants worked on a risky task that included 20 rounds of investment decision making. Changes in their risk-taking tendency were examined as they worked through the task. The first experiment tested the first hypothesis: Perceived endings could lead to increased risk taking. More specifically, it is predicted that participants who know that they are working on the last round of investment decisions would be more risk seeking than those who are not aware of the ending (Hypothesis 1).

Experiments 2 to 4 examined the second hypothesis regarding the underlying mechanism of the ending effect. If the ending effect is driven by the individuals’ motivation to pursue an emotionally rewarding ending, then participants would select the choice option which could realize this goal. When no option is possible to provide a rewarding result, then the ending effect should disappear. Therefore, the emotional motivation account predicts that the ending effect would disappear when options with gains are no longer available (Hypothesis 2a, Experiment 2). Although Experiment 2 predicts that the ending effect is probably driven by potential gains associated with the risky option, it could not distinguish whether the effect is driven (a) by the potential monetary rewards only or (b) by the potential emotional rewards associated with the monetary gains. Experiment 3 measured participants’ motivation for monetary rewards and emotional rewards. It is predicted that the motivation to pursue emotional rewards (rather than the motivation for monetary rewards) before the last round could predict increased risk tasking in the last decision round (Hypothesis 2b).

Experiment 4 directly tested the motivational account by manipulating participants’ motivation as they were approaching the last decision round. The current study proposes that the ending effect is caused by the motivation for emotional satisfaction, and the awareness of an approaching ending increases motivation for emotional satisfaction, thus leading to the ending effect. Therefore, it is predicted that, compared with priming participants to maximize their monetary outcome (utilitarian motivation priming) and no priming, priming participants to maximize their emotional satisfaction while working on the last round of the decision task (emotional motivation priming) would lead to increased risk taking in the last round (Hypothesis 2c).

Experiment 1

Method

Participants and design

Forty-six undergraduate students (13 male and 33 female, M_age = 21.3, SD_age = 2.16) from a university in Beijing, China, were recruited in Experiment 1. All participants completed 20 decision rounds; each of them received 5 to 44 Chinese Yuan (1 Yuan ≈ 0.16 US$) based on their performance. They were randomly assigned to the informed condition and the noninformed condition. In the informed condition, participants were told that they were going to make 20 decision rounds, whereas in the noninformed condition, participants did not know how many decisions they were going to make until they had finished all decisions.

To determine the total sample size of Experiment 1, we conducted a power analysis by G-Power software (Faul, Erdfelder, Buchner, & Lang, 2009; Faul, Erdfelder, Lang, & Buchner, 2007; Kelley, Eastwick, Harmon-Jones, & Schmeichel, 2015). The effect size was based on a previous study (ϕ = 0.13; see McKenzie et al., 2016). If the ending effect is true, rounds would not be equivalent, thus violating the sphericity assumption. One convention holds that a “nonsphericity correction” value should be at least 0.75 or higher in repeated measure ANOVAs. Accordingly, we inputted 0.75 for this parameter. Results showed that at least 38 participants were required to achieve an adequate amount of statistical power, 0.80; see Cohen, 1988, 1992. The current sample of 46 participants reached the target sample size.

Procedure

At the beginning, participants were given 10 tokens and were told that (a) they would receive cash for the amount they had remaining after the experiment; (b) if they lost, they could either pay the amount they lost in cash or eliminate their loss by translating a two-page English article into Chinese (similar procedures have been used in previous studies to convince the participants that they could really lose money; Scholer et al., 2010). In fact, those who were left with fewer than five tokens were paid a minimum of five Chinese Yuan. Within each round, participants decided whether to invest or not. If they decided to invest, they were asked to decide on the amount of tokens to invest. They were allowed to invest no more than five tokens within each round. The experimenter then tossed a die. If the outcome was “1” (1/6 chance), then the participants won; if the outcome was not “1” (5/6 chance), the participants lost. If the participants invested and won, they would receive 7 times the token(s) they invested; if the participants invested and lost, they would lose the token(s) invested. If the participants decided not to invest, the experimenter still tossed the die, and the experiment advanced to the next round after the participants knew the outcome of the toss, but no change happened to the participants’ account. This paradigm was adapted from Shiv et al. (2005).

After completing two practice trials, participants advanced to the 20 decision rounds; the experimenter reminded the participants which round they were in at the beginning of each round. As a manipulation check, after the decision-making session, participants reported if they were aware of the ending as they worked on their 20th decision. They then completed a demographic information sheet (age, gender, etc.) before they were paid, debriefed, thanked, and dismissed.

Results

All participants passed the manipulation check. Two repeated measures ANOVAs were performed, with investment likelihood (percent of rounds invested) and amount (average investment amount per round) as the dependent variables. The independent variables were task round and experimental conditions. The results showed significant interaction effects for both dependent variables: investment likelihood: F_likelihood(19, 836) = 1.80, p = .019, $η_{p}^{2}$ = 0.039, 95% confidence interval (CI) = [0.001, 0.045], and investment amount: F_amount(19, 836) = 3.18, p = .000, $η_{p}^{2}$ = 0.067, 95% CI = [0.021, 0.081]. As shown in Figure 1, during the last round, participants in the informed condition were more likely to invest (M = 87%, SD = 34%) and invested more tokens (M = 1.43, SD = 1.38) than those in the noninformed condition: investment likelihood: M = 61%, SD = 50%, t(44) = 2.06, p = .045, d = 0.60; investment amount: M = 0.7, SD = 0.64, t(44) = 2.34, p = .024, d = 0.67. Next, separate analyses were performed within each of the two conditions.

Figure 1.

Average likelihood to invest and average investment amount during the 20 rounds of the risk decision task in Experiment 1.

Within the informed condition, repeated contrasts were conducted to compare the mean of each round with the subsequent round. The results revealed that the average investment amount in the last round (M = 1.44, SD = 0.29) was significantly different from the previous round (19th round, M = 0.83, SD = 0.15), F(1, 22) = 6.82, p = .016, $η_{p}^{2}$ = 0.24. Deviation contrasts showed that the average likelihood to invest and the amount of investment in the last round was significantly larger than the grand mean of all rounds, Fs(1, 22) = 17.23, 10.05, ps = .000, .004, $η_{p}^{2} s$ = 0.44, 0.31, respectively. These results were consistent with Hypothesis 1.

In contrast, participants in the noninformed condition were not more likely to invest in the last round (M = 63%, SD = 50%) than the average of all rounds, F(1, 22) = 0.08, ns. Neither did they invest more tokens (M = 0.70, SD = 0. 635) in the last round than the grand mean of all rounds, F(1, 22) = 2.46, ns. Their likelihood to invest and their investment amount in the last round did not differ from the previous round (19th round, M = 61%, SD = 50%, M = 0.70, SD = 0. 635), Fs(1, 22) = 0.00, 0.00, respectively, ns.

The ending effect score

Next, we calculated the ending effect scores of the investment amount and the likelihood by subtracting the average investment (amount or likelihood) during the first 19 rounds from the investment (amount or likelihood) during the last round. A higher ending effect score reflected more increased risk taking toward the end of the risky decision task. This ending effect score provided a more straightforward indicator to assess the existence of the ending effect. As expected, the ending effect score of the investment likelihood under the informed condition (M = 0.28, SD = 0.33) was significantly higher than that of the noninformed condition (M = −0.02, SD = 0.35), F(1, 44) = 9.209, p = .004, $η_{p}^{2}$ = 0.173, 95% CI = [0.020, 0.360]. Similarly, the ending effect score of the investment amount under the informed condition (M = 0.76, SD = 1.15) was significantly higher than that of the noninformed condition (M = −0.14, SD = 0.45), F(1, 44) = 12.411, p = .001, $η_{p}^{2}$ = 0.220, 95% CI = [0.042, 0.405].

Next, one-sample t test compared each of the ending effect scores with zero. It was found that in the informed condition, the ending effect score of both the investment likelihood, t(22) = 4.15, p = .000, d = 0.848, and the investment amount, t(22) = 3.17, p = .004, d = 0.661, were significantly higher than zero. In contrast, the ending effect scores in the noninformed condition were not different from zero, likelihood: t(22) = −0.279, ns, amount: t(22) = −1.567, ns.

Discussion

Results from Experiment 1 clearly supported Hypothesis 1. Participants who knew they were working on the last round took more risks than those who did not know. The perception of an approaching ending led to increased risky decision making. Next, we examined the possible underlying mechanism of the ending effect. With three experiments, the motivational account of the ending effect was examined.

Experiment 2

In Experiment 2, we modified the task paradigm used in the first experiment in two ways. First, while Experiment 1 used a positive-expected value gamble, so that taking risks led to a higher expected outcome than not taking a risk, Experiment 2 used an equal-expected value gamble task, so the expected values of the risky and safe option were equal. Therefore, Experiment 2 could access the participants’ risk taking more effectively. Second, the risky decision task was modified into two different versions: the gain-framed version and the loss-framed version. In the loss-framed version of the task, neither of the two options provides the possibility to gain anything: one option is associated with the sure loss of a small amount and the other option is associated with the small chance to lose a larger amount. Assuming that the ending effect is driven by the motivation to gain rewards, it should disappear when the possibility to gain rewards is removed from the choice options. In contrast, in the gain-framed version of the task, one choice option is associated with the sure gain of a small amount and the other option is associated with the small chance to gain a larger amount. It was expected that the ending effect would not be observed in the loss-framed version of the task, though it would remain significant in the gain-framed version of the task (Hypothesis 2a). Support for this hypothesis would be consistent with the motivational mechanism.

Method

Participants and design

Forty-six undergraduate students (14 males and 32 females, M_age = 21.74, SD_age = 1.71) from a university in Beijing, China, were recruited. Participants were randomly assigned to the two conditions: the loss condition (n = 21) and the gain condition (n = 25). Participants in the loss condition worked on 20 rounds of the loss-framed version of the decision task, whereas those in the gain condition completed 20 rounds of the gain-framed version of the decision task. In both conditions, participants knew at the beginning that there were 20 rounds of decision tasks in the decision-making session. Sample size selection of Experiment 2 through Experiment 4 was based on the effect size coefficient from Experiment 1 ( $η_{p}^{2}$ = 0.039). Power analyses showed that a sample of at least 36 participants was required for Experiment 2. The current sample of 46 participants reached the target sample size.

Decision paradigm and procedures

The two versions of decision tasks used in Experiment 2 were modified based on the paradigm used in Experiment 1. Both versions were an equal-expected value gamble. In the loss-framed version of the task, participants were given 30 tokens at the beginning of the decision-making session. Within each trial, the experimenter tossed a die and the participants decided whether or not to take a risk. If the participants decided not to take a risk, they would lose one token. If they decided to take a risk and the result was not “1” (5/6 chance), they would lose nothing. If the result was “1” (1/6 chance), they would lose six tokens. For the gain-framed version of the task, participants were given no tokens at the beginning; they would receive one token within each trial if they decided not to take a risk. If they decided to take a risk, they would receive nothing if the result was not “1” (5/6 chance), but if the result was “1” (1/6 chance), they would receive six tokens.

The procedure in Experiment 2 was similar to the procedure in the first experiment, except that the participants in Experiment 2 completed one of the two different task versions.

Results

The task paradigm used in Experiment 2 only asked participants to decide whether or not to take a risk in each round, it did not allow participants to decide how many tokens to invest. Therefore, Experiment 2 used only one dependent variable: the participants’ decision as to whether to take a risk within each round (risk-taking likelihood). First, a repeated measures ANOVA was conducted in which the participants’ risk-taking likelihood served as the dependent variable. The main effect of the experimental condition was significant: participants in the loss condition (M = 79.5%, SD = 5.5%) were more risk seeking than those in the gain condition (M = 43.2%, SD = 50.0%), F(1, 44) = 23.84, p = .000, $η_{p}^{2}$ = 0.35, 95% CI = [0.133, 0.521]. This is consistent with previous findings that individuals are more risk seeking in the loss domain than in the gain domain (Tversky & Kahneman, 1981; for reviews, see Levin, Gaeth, Schreiber, & Lauriola, 2002; Levin, Schneider, & Gaeth, 1998; for a meta-analysis, see Kühberger, 1998). The main effect of rounds was also significant, F(19, 836) = 2.30, p = .001, $η_{p}^{2}$ = 0.05, 95% CI = [0.008, 0.059]. Participants were more likely to take a risk when the gambling task was drawing to a close, and the series affected risk taking. As shown in Figure 2, there was a significant interaction effect between task round and conditions (loss-framed vs. gain-framed), F(19, 836) = 1.765, p = .023, $η_{p}^{2}$ = 0.039, 95% CI = [0.0003, 0.044]. Next, separate analyses were performed to analyze data within each experimental condition.

Figure 2.

Average risk-taking likelihood during the 20 rounds of the risky decision task in Experiment 2.

In the gain condition, repeated contrasts revealed that the last round (M = 0.68, SD = 0.48) was significantly different from the round before the last round (19th round, M = 0.40, SD = 0.50), F(1, 24) = 6.68, p = .016, $η_{p}^{2}$ = 0.218. Deviation contrasts also confirmed that the last round was significantly higher than the grand mean of all rounds, F(1, 24) = 7.98, p = .009, $η_{p}^{2}$ = 0.25. Therefore, the ending effect remained significant among participants who completed the gain-framed version of the decision task. In the loss condition, repeated contrasts revealed that the last round (M = 71%, SD = 46%) was not significantly different from the 19th round (M = 67%, SD = 48%), F(1, 20) = 0.19, ns. Deviation contrasts also confirmed that the last round was not different from the grand mean of all rounds, F(1, 20) = 0.882, ns. As expected, the ending effect was not observed when participants completed the loss-framed version of the decision task. There was a quadratic effect in the gain condition, B = 0.003, t = 3.498, p = .003, but no quadratic effect was found in the loss condition, B = 0.000, t = 0.333, ns.

Next, the ending effect score of the risk-taking likelihood was calculated for each condition. The ending effect score of the gain condition (M = 0.26, SD = 0.46) was significantly higher than that of the loss condition (M = −0.09, SD = 0.42), F(1, 44) = 7.012, p = .011, $η_{p}^{2}$ = 0.137, 95% CI = [0.007, 0.322]. Whereas the ending effect score in the gain condition was significantly higher than zero, t(24) = 2.824, p = .009, d = 0.565, the ending effect score in the loss condition was not different from zero, t(20) = −0.939, ns

Discussion

The results supported Hypothesis 2a: The ending effect disappeared when the option that provides gains was no longer available. This finding provides initial support for the motivational account. As the ending effect was no longer observed when no option could enable the participants to gain rewards, it is reasonable to argue that the ending effect may be driven by the individuals’ desire to gain rewards.

Experiment 3

Experiments 1 and 2 showed that the ending effect remained significant as long as the two choice options provided the possibility to gain rewards. It appears that the anticipation to gain extra rewards is a prerequisite to induce the ending effect. However, when individuals receive extra monetary rewards, they also experience emotional satisfaction. It is possible that the ending effect is driven by the individuals’ motivation to gain monetary rewards, it is also likely that the ending effect is driven by the individuals’ motivation to pursue emotional satisfaction. Both motivational accounts might explain the ending effect. Experiment 3 attempted to distinguish between these two competing motivational accounts.

In Experiment 3, we directly measured the strength of the participants’ motivation toward monetary gains (utilitarian motivation) and emotional satisfaction (emotional motivation). This allowed us to further distinguish between the roles of the two motivations in the ending effect. In the present work, the emotional motivation refereed to the motivation to pursue emotional satisfaction and the utilitarian motivation referred to the motivation to gain monetary rewards. The source of emotional satisfaction in risk taking comes from the prospect of monetary gain associated with the risky task. As SST predicted, the emotional motivation is heightened at the end of the task. Thus, it was expected that, when facing the last round of the decision task, the emotional motivation would increase compared with the utilitarian motivation. Therefore, it was predicted that the emotional motivation, but not the utilitarian motivation, could predict increased risk taking in the last decision round (Hypothesis 2b).

Method

Participants and procedure

Fifty undergraduate students (10 males and 40 females, M_age = 20.40, SD_age = 1.69) from a university in Beijing, China, were recruited. Based on the effect size coefficient from Experiment 1 ( $η_{p}^{2}$ = 0.039), power analyses showed that a sample of at least 48 participants was required for the design and planned analyses of Experiment 3. Each participant completed 20 rounds of equal-expected-value decision tasks. Just before the last round of investment, all participants were reminded that they were going to work on the last round. To avoid potential demand characteristics, we did not ask participants to report their motivation before the decision session or before they worked on the last round. Instead, after the last decision-making session, participants were asked to report the strength of their motivations on an 8-point Likert-type scale. Specifically, participants indicated to what extent they agreed with the statement “The motivation for my decision was to satisfy my emotional (financial) need” while they were making their decision in the final round, and during the entire decision session. Finally, participants were paid, debriefed, and thanked.

Results

Replication of the ending effect

As shown in Figure 3, results in Experiment 3 replicated the ending pattern found in previous experiments. Repeated measures ANOVA showed that the participants’ investment likelihood and investment amount in the last round (M_likelihood = 62%, SD_likelihood = 49%; M_amount = 0.90, SD_amount = 0.89) were significantly different from the grand mean of all previous rounds, M_likelihood = 45%, SD_likelihood = 27%; M_amount = 0.54, SD_amount = 0.36; F_likelihood(1, 49) = 5.382, p = .025, $η_{p}^{2}$ = 0.099; F_amount(1, 49) = 9.225, p = .004, $η_{p}^{2}$ = 0.158, and were also significantly different from the 19th round, M_likelihood = 44%, SD_likelihood = 50%; M_amount = 0.56, SD_amount = 0.73; F_likelihood(1, 49) = 6.975, p = .011, $η_{p}^{2}$ = 0.125; F_amount(1, 49) = 16.447, p = .000, $η_{p}^{2}$ = 0.251.

Figure 3.

Average likelihood to invest and average investment amount during the 20 rounds of the risky decision task in Experiment 3.

Emotional versus utilitarian motivation

Correlation analyses showed that emotional motivation did not correlate with utilitarian motivation for either the entire session (r = −.086, ns) or for the last round (r = .063, ns). Next, a repeated-measures ANOVA was conducted with two within-group independent variables: type of motivation (emotional vs. utilitarian) and time of motivation (last round vs. entire rounds). The dependent variable was the strength of motivation participants reported at the end of the decision session. An interaction effect was found, F(1, 49) = 11.210, p = .002, $η_{p}^{2}$ = 0.186. Simple effect analysis showed that whereas utilitarian motivation (M = 5.50, SD = 1.67) was significantly higher than emotional motivation during the entire session (M = 3.64, SD = 1.69), F(1, 49) = 30.269, p = .000, difference between these two types of motivations (M_utilitarian = 4.68, SD_utilitarian = 2.13; M_emotional = 4.10, SD_emotional = 2.14) disappeared immediately before the last round, F(1, 49) = 2.070, ns

Correlations between motivation and the ending effect

The ending effect scores of investment likelihood and investment amount were calculated. Consistent with Hypothesis 2b, correlation analyses revealed that last round emotional motivation significantly correlated with both ending effect scores of investment likelihood (r = .369, p = .008) and investment amount (r = .456, p = .001). In contrast, last-round utilitarian motivation did not correlate with either ending effect scores of investment likelihood (r = −.176, ns) or investment amount (r = −.101, ns).

Regression analyses

Next, to further determine which kind of motivation was driving the ending effect, a linear multiple regression was performed in which the ending effect scores served as the dependent variable. Consistent with Hypothesis 2b, when the emotional motivation and utilitarian motivation were both included in the same model, only emotional motivation was positively predictive of the ending effect scores of both investment likelihood and investment amount (Table 1). The correlation between the utilitarian motivation and the ending effect scores were not significant. Similar results were obtained from separate regression models.

Table 1.

Emotional Motivation, but Not Utilitarian Motivation, Predicts Increased Risk Taking in the Last Round in Experiment 3.

Predictor	Outcome
	Ending effect score of amount				Ending effect score of likelihood
	B	p	CI	R ²	B	p	CI	R ²
Multiple linear regressions
EM in Round 20	0.187	.001	[0.085, 0.290]	.231	0.095	.005	[0.031, 0.160]	.184
UM in Round 20	−0.061	.240	[–0.164, 0.042]	.231	−0.054	.104	[–0.119, 0.011]	.184
Separate linear regressions
EM in Round 20	0.181	.001	[0.078, 0.283]	.208	0.090	.008	[0.024, 0.155]	.136
UM in Round 20	−0.040	.485	[–0.155, 0.075]	.010	−0.043	.220	[–0.113, 0.027]	.031

Note. EM = emotional motivation; UM = utilitarian motivation.

The moderating role of emotional motivation on the ending effect

Next, we examined Hypothesis 2b by examining whether emotional or utilitarian motivation for the last round moderated the ending effect. First, participants were divided into a lower last round emotional motivation group and a higher last round emotional motivation group by their self-reported emotional motivation before the last round. Then, repeated measures ANOVAs were conducted. The two independent variables were the strength of the participants’ emotional motivation (high vs. low), round of risky task (previous 19 rounds vs. last round), and the dependent variables were the participants’ investment likelihood and investment amount, respectively. The interaction between emotional motivation and round was significant for both investment likelihood, F(1, 48) = 7.602, p = .008, $η_{p}^{2}$ = 0.137, and investment amount, F(1, 428) = 10.510, p = .002, $η_{p}^{2}$ = 0.180. Simple effect analyses revealed that, as expected, the ending effect only emerged among participants who reported higher last round emotional motivation, investment likelihood: F(1, 48) = 13.71, p = .001; investment amount: F(1, 48) = 21.41, p = .000, but not among those with lower last round emotional motivation, investment likelihood: F(1, 48) = 0.00, ns; investment amount: F(1, 48) = 0.11, ns.

Furthermore, between group comparisons examined the participants with higher and lower last round emotional motivation, and it was found that their average investment in the first 19 rounds did not differ, likelihood: F(1, 48) = 0.46, ns; amount: F(1, 48) = 0.08, ns. In contrast, in the last round, participants with a higher last round emotional motivation (M_likelihood = 86%, SD_likelihood = 35%; M_amount = 1.32, SD_amount = 0.84) invested more than those with lower last round emotional motivation (M_likelihood = 43%, SD_likelihood = 50%; M_amount = 0.57, SD_amount = 0.79); likelihood: F(1, 48) = 11.85, p = .001; amount: F(1, 48) = 10.43, p = .002. Next, participants were again divided into a lower last round utilitarian motivation group and a higher last round utilitarian motivation group by their self-reported utilitarian motivation before the last round. Repeated measures ANOVAs showed that the interaction effect between utilitarian motivation and round was not significant for investment amount, F(1, 48) = 2.162, ns, or for investment likelihood, F(1, 48) = 1.982, ns.

Discussion

Results examining the two types of motivation showed that, at the end of the decision session, emotional motivation increased in comparison with utilitarian motivation. Such a change in pattern of the two types of motivation was consistent with previous studies examining SST. Furthermore, utilitarian motivation and emotional motivation did not correlate with one another, suggesting that expectation for emotional satisfaction and desire for monetary gain could be separated from each other.

Results from the regression analysis and moderation analysis consistently supported Hypothesis 2b. Whereas the motivation toward emotional satisfaction before the last round could predict increased risk taking in the last decision round, the motivation toward monetary rewards could not. Consistently, emotional motivation moderated the emergence of the ending effect. Participants who reported higher levels of emotional motivation before the last round showed increased risk taking during the last round, whereas those who reported lower levels of emotional motivation before the last round did not show increased risk taking during the last round. Such a moderating effect was not found in the participants’ utilitarian motivation. Partial evidence was found from the moderation analyses that the participants with lower utilitarian motivation showed the ending effect. However, no predictive effect of utilitarian motivation was found in the regression analyses. To summarize, it appears that increased emotional motivation, but not decreased utilitarian motivation, was driving the ending effect.

The unique predictive power of emotional motivation suggested that, at least in the current study, the participants’ desire for emotional satisfaction was more relevant to the increase of risk taking toward the end. This finding supports the motivational account of emotional satisfaction to explain the ending effect, and it speaks against the motivational account of monetary gains.

One limitation of Experiment 3 is that the participants’ motivation was measured, rather than manipulated. Therefore, the relationship between emotional motivation and risky decision-making revealed in Experiment 3 was susceptible to alternative explanations including a third variable. To address this concern, Experiment 4 examined the causal link between motivation and risky decision making by experimentally manipulating participants’ motivation immediately before the last decision round.

Experiment 4

Experiment 4 used a concurrent double randomization design (Pirlott & MacKinnon, 2016). We manipulated the hypothesized mediator, motivation (M*), by randomly assigning participants to receive emotional motivation priming (emotion condition), monetary motivation priming (utility condition) or no priming (control condition). Simultaneously, we manipulated the independent variable (X), so that participants in each motivation condition were either told (informed) or not told (noninformed) at the beginning of the experiment regarding how many rounds there were in the investment task.

The present work proposes that the ending effect in risky decision making is caused by the individuals’ motivation to pursue an emotionally rewarding ending. Therefore, in the emotion condition, the participants were told to try to maximize their emotional satisfaction while they were working on the last decision round. A competing motivational mechanism is that the ending effect is driven by the motivation to maximize their monetary outcomes. Therefore, in the utility condition, the participants were asked to try to maximize their monetary outcome while they were working on the last round. Participants in the control condition did not receive any motivational priming before the last round.

It was predicted that participants who had received emotional motivation priming would show increased risk taking during the last round, whereas those who had received utilitarian motivation priming and those who had received no motivation priming would not show this pattern of change in the last round, unless they were in the informed condition (Hypothesis 2c). It was also expected that the ending effect found in Experiments 1 to 3 would be replicated in the informed conditions in Experiment 4. No interaction effect between motivational priming and informed conditions was expected for Experiment 4.

Method

Participants and design

One hundred seventy-nine undergraduate students (60 males and 119 females, M_age = 20.18, SD_age = 2.34) from a university in Beijing, China, were recruited. No one had participated in previous studies using the same paradigm. Experiment 4 was a 2 (experimental condition: informed vs. noninformed) × 3 (motivational priming: emotion vs. utility vs. control) between-group design. Power analyses showed that a sample of at least 144 participants was required ( $η_{p}^{2}$ = 0.039). The current sample of 179 participants reached the target sample size.

Procedures

After signing the informed consent form, participants were randomly assigned to one of the six experimental conditions. Half of the participants were informed that they were going to work on 20 rounds of investment tasks (informed condition: n = 89), whereas the other half were not informed of this piece of information (noninformed condition: n = 90). Participants in each condition then worked on 20 rounds of risky decision tasks, which is identical to the task paradigm used in Experiment 3, with the only difference being that participants in the emotion condition and utility condition received an additional instruction (the motivational manipulation) immediately before their last decision round.

The instructions for participants in the emotional (utilitarian) motivation conditions (informed and uninformed) were

In the last (next) round of the decision task, the most important thing for you is to make a decision that could maximize your emotional satisfaction (financial income). In other words, we want you to keep in mind that while working on the last (next) round of the decision task, your most important goal is to meet your emotional need (financial need).

Finally, the participants completed a demographic form before they were paid, debriefed, and thanked.

Results

Investing differences across conditions

Figure 4 illustrates the trial-specific results in Experiment 4. Repeated-measures ANOVAs were conducted. The participants’ investment likelihood and investment amount served as the dependent variables, whereas the independent variables were the informed condition (between-subject: informed vs. noninformed), motivational priming (between-subject: emotional vs. motivational vs. control) and 20 rounds of decisions (within subject). As expected, the three-way interaction was not significant, F_likelihood(38, 3287) = 1.015, ns, F_amount(38, 3287) = 1.047, ns. The informed condition by round interaction was significant, F_likelihood(19, 3287) = 2.603, p = .000, $η_{p}^{2}$ = 0.015, F_amount(19, 3287) = 5.174, p = .000, $η_{p}^{2}$ = 0.029. The motivational condition by round interaction was also (marginally) significant, F_likelihood(38, 3287) = 1.320, p = .091, $η_{p}^{2}$ = 0.015, F_amount(38, 3287) = 1.754, p = .003, $η_{p}^{2}$ = 0.020.

Figure 4.

Average likelihood to invest and average investment amount during the 20 rounds in Experiment 4.

Next, we examined the risk-taking pattern of the participants in the informed condition. The interaction effect between round and emotional conditions was marginally significant for investment amount, F_amount(38, 1634) = 1.403, p = .053, $η_{p}^{2}$ = 0.0320, but not for investment likelihood, F_likelihood(38, 1634) = 1.260, ns. Simple effect analysis of investment amount showed that a significant effect of round emerged in the emotional condition, F_amount(19, 68) = 6.41, p = .000, control condition, F_amount(19, 68) = 1.72, p = .027, and utilitarian condition, F_amount(19, 68) = 2.79, p = .000. These results suggest that, in the three informed conditions, the participants’ risk-taking tendency differed across the 20 rounds.

Then, to examine whether participants showed increased risk taking in the last round across different conditions, the ending effect scores were calculated, following the equation used in Experiment 1 through Experiment 3. An ANOVA was performed to compare the ending effect scores across the six conditions. The dependent variables were the ending effect score of investment likelihood and investment amount, and the independent variables were the informed condition and motivational priming. As shown in Figure 5, a significant main effect of the informed condition emerged for both dependent variables, F_likelihood(1, 173) = 12.327, p = .001, $η_{p}^{2}$ = 0.067, 95% CI = [0.013, 0.147], F_amount(1, 173) = 21.413, p = .000, $η_{p}^{2}$ = 0.11, 95% CI = [0.037, 0.201]. For participants in the informed condition, their ending effect score of both investment likelihood and amount (M_likelihood = 0.178, SD_likelihood = 0.455; M_amount = 0.793, SD_amount = 1.396) were higher than those in the noninformed condition (M_likelihood = −0.661, SD_likelihood = 0.483; M_amount = 0.021, SD_amount = 0.821).

Figure 5.

Ending effect scores of investment likelihood and investment amount across the six conditions in Experiment 4.

More importantly, the main effect of motivational priming was also significant for the ending effect score of investment amount, F_amount(2, 173) = 4.619, p = .011, $η_{p}^{2}$ = 0.051, 95% CI = [0.003, 0.120]. As three groups were equal in variance and unequal in sample sizes, we chose Gabriel’s test to conduct post hoc analysis (Field, 2009). The results showed that, as predicted, the ending effect score of investment amount in the emotional priming condition (M_emotional = 0.764, SD_emotional = 1.52) was significantly higher than in the other two conditions (M_utility = 0.241, SD_utility = 1.004, p = .037; M_control = 0.217, SD_utility = 0.952, p = .025), and there was no difference between the utility condition and the control condition (p = .999). The ending effect score for investment likelihood showed no difference among the three groups, Ms = 0.008, 0.154, 0.005, SDs = 0.483, 0.472, 0.498, F_likelihood(2, 173) = 2.048, p = .132, though descriptive data were in the predicted direction.

Consistent with Hypothesis 2c, neither of the two dependent variables showed a significant interaction effect between motivational priming and the experimental condition, F_likelihood(2, 173) = 0.658, p = .519, F_amount(2, 173) = 0.479, p = .62. In addition, no difference was found between the emotional-noninformed condition (M_likelihood = −0.018, SD_likelihood = 0.488, M_amount = 0.267, SD_amount = 1.178) and the control-informed condition (M_likelihood = 0.126, SD_likelihood = 0.465, M_amount = 0.551, SD_amount = 1.136), F_likelihood(1, 58) = 1.365, ns, F_amount(1, 58) = 0.904, ns. This finding provides further support for Hypothesis 2c.

Discussion

Experiment 4 replicated previous findings of the ending effect in risky decision making. More importantly, by manipulating the participants’ motivation immediately before the last round of the risky decision task, Experiment 4 directly tested the motivational account of the ending effect. The results showed that participants who received motivation priming to maximize their emotional satisfaction showed increased risk taking during the last round. This effect remained significant, even when they were in the noninformed condition. In contrast, the participants who were instructed to maximize their monetary outcome did not differ from the participants in the control condition. Results from Experiment 4 provide direct support for the motivational account of the ending effect: Motivational induction toward emotional satisfaction could lead to increased risk taking toward the end.

No Moderation by Prior Winnings in Experiment 1 to Experiment 4

To examine whether the number of winning outcomes in rounds before the last round influenced the individuals’ risky decision making in the last round, moderation analyses were performed for each experiment. Winning outcomes refer to the rounds in which the participants decided to invest and the result of the die toss was 1. Across all of the four experiments, we found no evidence that the ending effect depended on the number of winning outcomes the participants received before the last round. The results of the moderation analyses were summarized in Table 2. These results are consistent with McKenzie et al. (2016), and both studies speak against the reference dependent explanation of the ending effect.

Table 2.

No Moderation of Previous Winnings on Risk Taking in the Last Round in Experiments 1 Through 4.

Experiment		Dependent variable
		Likelihood		Invested amount
		t (p)	CI	t (p)	CI
1	Prior winning × Condition	−0.379 (.707)	[–.219, .150]	0.076 (.940)	[–.493, .531]
2	Prior winning × Condition	0.997 (.324)	[–.110, .326]	—	—
3	Prior winning × Growth score of EM	0.467 (.643)	[–.052, .083]	1.659 (.104)	[–.018, .191]
3	Prior winning × EM in the Round 20	−0.205 (.838)	[–.065, .053]	0.796 (.430)	[–.060, .138]
4	Prior winning × Condition	−1.009 (.314)	[–.143, .046]	0.190 (.849)	[–.229, .278]
4	Prior winning × Priming	1.067 (.287)	[–.027, .092]	0.692 (.490)	[–.105, .218]

Note. EM = emotional motivation.

General Discussion

The present work demonstrates that individuals’ risk-taking tendency is not equivalent as they work through a set of risky decision tasks. With four experiments, we examined the ending effect in risky decision making and tested the underlying mechanism of this effect. The first experiments supported the notion that individuals show increased risk taking when they are approaching the end of a risky decision task. Participants in the current study were university students who are unlikely to be professional betters. Accordingly, the results are consistent. with previous studies that suggest individuals are more likely to bet on the option that is associated with higher risk, at least for novice betters and recreational betters (nonprofessional betters, Ali, 1977; Asch et al., 1982; McGlothlin, 1956; McKenzie et al., 2016).

Experiments 2 through 4 addressed the possible underlying mechanism of this ending effect. Based on the present findings, the ending effect may be best explained by a motivational shift that emerges as individuals are approaching an ending. This motivational shift could be described as individuals becoming more motivated to pursue emotional satisfaction. Experiment 2 demonstrated that when neither of the two options provides the possibility to gain any rewards (i.e., no option could meet the individuals’ need to gain rewards), the ending effect disappeared. This finding provides support for the notion that the motivation to gain rewards is driving the ending effect. However, results from Experiment 2 could not distinguish whether the ending effect is driven by the anticipation to gain monetary rewards or by the desire to pursue emotional satisfaction. This question was addressed by Experiments 3 and 4, and the results from both experiments support the motivational account of emotional satisfaction.

SST (Carstensen, 2006; Carstensen et al., 1999) suggests that the goal of emotional satisfaction gets more salient as individuals are approaching the end of a significant life event. This shift in motivation could lead to changes in cognitive processes, as individuals pay more attention to and remember more positive stimuli than negative stimuli toward the end. The present research revealed that this shift in motivation toward emotional satisfaction might also lead to changes in the risk-taking tendency. Experiment 3 measured changes in the participants’ motivation toward emotional satisfaction and toward monetary rewards before they worked on the last decision round. It was found that only the motivation toward emotional satisfaction, but not the motivation toward monetary rewards, was linked with the ending effect. This result provides additional support for the motivational account of emotional satisfaction to explain the ending effect.

Experiment 4 directly tested the motivational account by manipulating the participants’ motivational state as they were working on the last round. It was found that the participants who were asked to try to maximize their emotional satisfaction as they were working on the last round of the risky decision task showed the ending effect, whereas those who were instructed to maximize their monetary outcome during the last round of the risky task did not show increased risk taking in the last round. Across three experiments (Experiments 2-4), the findings consistently support the idea that individuals are motivated to prioritize goals relevant to emotional satisfaction toward the end. This motivational shift may subsequently lead to increased risk taking toward the end (i.e., the ending effect).

The risk-as-feelings hypothesis posits that emotional reactions to risky situations play an important role in risky decision making (Loewenstein, Weber, Hsee, & Welch, 2001). Factors that influence people’s emotional reactions to risks might not influence their cognitive evaluations of risks. In addition, the influence of cognitive evaluations on risk taking may be mediated, in part, by affective responses. The present study contributes to our understanding of the role of emotional processes in risk taking by demonstrating that considerations of emotional satisfaction tend to play a more important role toward the end of risky decision making.

The present work provides a potential explanation for the discrepant findings between nonprofessional (novice/recreational) gamblers and more professional gamblers regarding their preference for riskier options toward the end (McKenzie et al., 2016). Nonprofessional gamblers have been consistently suggested to bet more on riskier options at the end (Ali, 1977; Asch et al., 1982; McGlothlin, 1956). In these studies, data were recorded in the United States before 1990, when off-track betting was illegal in most regions of the United States, so most of the individuals involved in data analyses were recreational gamblers or novice gamblers (McKenzie et al., 2016). However, this phenomenon was not found among gamblers who are relatively more professional. Snowberg and Wolfers (2010) found no ending effect when analyzing most U.S. horse races from 1992 to 2001, when many areas of the United States had legalized off-track betting and more off-track betters were included in the analyses. In another study conducted in the United Kingdom, off-track betters showed the opposite of the ending effect (Johnson & Bruce, 1993).

Results from the current study suggest that the ending effect may be caused by a motivational shift toward the end. Novice and recreational gamblers play lotteries for fun, and are not expected to receive any training in gambling. In comparison, off-track gamblers are expected to be more serious with their gambling activities. Therefore, compared with off-track gamblers, recreational/novice gamblers may be more likely to be influenced by a motivational state and to subsequently exhibit the ending effect. Examining when, how, and why professional and recreational gamblers differ and the subsequent consequences is certainly a direction that deserves close investigation, and the present study contributes to this question by highlighting the possibility that the difference in risky behavior between these two groups may be stronger at the end of the risky task due to a vulnerability in motivational shifts induced by the perceived ending.

Limitations and Future Directions

Although the present findings enhance our understanding of risky decision making and its underlying mechanism, they should be considered in light of some limitations. One limitation of the current study is that a positive-expected value gamble task was used in Experiment 1 in an attempt to replicate the ending effect. Therefore, participants in Experiment 1 would expect that choosing the risky option would lead to a higher monetary outcome. This might allow for other explanations in explaining the ending effect found in Experiment 1. However, this limitation might be mitigated by two things. First, we compared within subject difference between the participants’ decision in the final round and previous rounds in Experiment 1. Second, Experiments 2 through 4 used an equal-expected value gamble and the ending effect remained significant (except in the loss condition in Experiment 2). Therefore, it appears that the ending effect reported in the present work and in previous studies (Ali, 1977; Asch et al., 1982; McGlothlin, 1956; McKenzie et al., 2016) is a robust phenomenon.

In Experiment 3, the participants’ motivation was measured at the end of the decision session, after the participants had made 20 decision rounds. This procedure might be subject to demand characteristics. Future studies might consider asking the participants to report their motivations at the beginning of the decision session.

Another limitation is that we only used one task paradigm to test the ending effect. The task used includes one safe option and one long shot (i.e., choice with a lower likelihood to gain a large amount). Future studies need to examine whether the ending effect would hold for other probabilities and payoffs. This is an open question worthy of further study, as it could not only help us understand the boundaries of the ending effect, but also contribute to a better understanding of the underlying mechanism of the ending effect.

The current study measured the participants’ risky behavior on the basis of their decisions in a task in which full information about probabilities and outcomes were provided. This type of decision task, which involves “decisions from description” (Hertwig & Erev, 2009), is different from other tasks involving “decisions from experience.” The latter tasks do not provide explicit information about probabilities, and individuals make decisions on the basis of the experience they have accumulated through feedback (Hau, Pleskac, & Hertwig, 2010; Hertwig & Erev, 2009). Future studies may benefit from including decisions from experience, and further analyzing which task characteristics are crucial in determining whether the ending effect remains significant. This could also help enhance our understanding of the underlying mechanism of the ending effect.

In the current study, the participants’ gender breakdown was disproportionate. There were more female participants than male participants in each of the four experiments. Although previous studies indicate that there are large gender differences in risk tasking, no gender differences in the “last race effect” have been identified. Therefore, findings in the present work are unlikely to be influenced by the unequal gender breakdown. Indeed, re-analyses of the present data revealed no gender differences in risk taking (ps > .07) and no gender by condition interactions (ps > .09). Even so, future studies might benefit from the inclusion of more male participants.

Another direction for future research would be to examine how motivational shifts lead to subsequent changes in risk-taking tendency. Previous studies examining SST showed that motivational shifts lead to changes in attentional (Isaacowitz et al., 2006a, 2016b) and memory processes (Mather & Carstensen, 2003, 2005). Another study examining medical decision making showed that changes in attention to decision-relevant information caused by a motivational shift played a role in subsequent changes in medical decision making (Lockenhoff & Carstensen, 2007). The current study did not examine whether attention and memory played a role in the link between motivational shift and changes in the risk-taking tendency. Examining this question would help clarify how motivational shift influences the individual’s risk-taking tendency.

Despite these limitations, the current study makes important contributions to the existing literature in three ways. First, it complements existing models of risky decision making by identifying one situational factor, perceived ending, which could lead to increased risk taking. More importantly, these studies demonstrate that motivational shifts toward emotional satisfaction induced by perceived ending may play an important role in accounting for people’s increased risk taking toward the end. Second, it expands existing studies examining the power of endings to the realm of risk taking. Third, it corroborates recent findings that everyday endings could also make a meaningful impact on individuals’ thoughts and behaviors.

In sum, the present research demonstrates the ending effect in risky decision making: Individuals tend to be more risk seeking toward the end of an investment task, and motivational shifts induced by perceived ending might play a role. The present study complements existing motivational accounts of risk taking and suggests a new research direction of integrating factors associated with the time perception of an approaching ending into existing models of risky decision making.

Supplemental Material

Xing_onlineappendix – Supplemental material for The Ending Effect in Investment Decisions: The Motivational Need for an Emotionally Rewarding Ending

Supplemental material, Xing_onlineappendix for The Ending Effect in Investment Decisions: The Motivational Need for an Emotionally Rewarding Ending by Cai Xing, Yuqi Meng, Derek M. Isaacowitz, Yue Wen and Zhongxin Lin in Personality and Social Psychology Bulletin

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by National Natural Science Foundation of China Grant 71873133, Fundamental Research Funds for the Central Universities, and the Research Funds 15XNB020 of Renmin University of China awarded to Cai Xing.

Supplemental Material

Supplemental material is available online with this article.

ORCID iD

Cai Xing

References

Ali

M. M.

(1977). Probability and utility estimates for racetrack bettors. Journal of Political Economy, 85, 803-815. doi:10.1086/260600

Asch

Malkiel

B. G.

Quandt

R. E.

(1982). Racetrack betting and informed behavior. Journal of Financial Economics, 10, 187-194. doi:10.1016/0304-405X(82)90012-5

Benartzi

Thaler

R. H.

(1995). Myopic loss aversion and the equity premium puzzle. The Quarterly Journal of Economics, 110, 73-92. doi:10.2307/2118511

Carstensen

L. L.

(2006). The influence of a sense of time on human development. Science, 312, 1913-1915. doi:10.1126/science.1127488

Carstensen

L. L.

Fredrickson

B. L.

(1998). Influence of HIV status and age on cognitive representations of others. Health Psychology Official Journal of the Division of Health Psychology American Psychological Association, 17, 494-503. doi:10.1037/0278-6133.17.6.494

Carstensen

L. L.

Gross

J. J.

Fung

H. H.

(1997). The social context of emotional experience. Annual Review of Gerontology and Geriatrics, 17, 325-352.

Carstensen

L. L.

Isaacowitz

D. M.

Charles

S. T.

(1999). Taking time seriously: A theory of socioemotional selectivity. American Psychologist, 54, 165-181. doi:10.1037/0003-066X.54.3.165

Cohen

(1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Lawrence Erlaum Associates.

Cohen

(1992). A power primer. Psychological Bulletin, 112, 155-159. doi:10.1037/0033-2909.112.1.155

10.

Davis

W. E.

Hicks

J. A.

(2013). Maintaining hope at the 11th hour: Authenticity buffers the effect of limited time perspective on hope. Personality and Social Psychology Bulletin, 39, 1634-1646. doi:10.1177/0146167213500150

11.

Effron

D. A.

Bryan

C. J.

Murnighan

J. K.

(2015). Cheating at the end to avoid regret. Journal of Personality and Social Psychology, 109, 395-414. doi:10.1037/pspa0000026

12.

Faul

Erdfelder

Buchner

Lang

A.-G.

(2009). Statistical power analyses using G*Power 3.1: Tests for correlation and regression analyses. Behavior Research Methods, 41, 1149-1160. doi:10.3758/BRM.41.4.1149

13.

Faul

Erdfelder

Lang

A.-G.

Buchner

(2007). G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39, 175-191. doi:10.3758/BF03193146

14.

Field

(2009). Discovering statistics using SPSS (3rd ed.). Thousand Oaks. CA: SAGE.

15.

Florack

Keller

Palcu

(2013). Regulatory focus in economic contexts. Journal of Economic Psychology, 38, 127-137. doi:10.1016/j.joep.2013.06.001

16.

Fredrickson

B. L.

Carstensen

L. L.

(1990). Choosing social partners: How old age and anticipated endings make people more selective. Psychology and Aging, 5, 335-347. doi:10.1037/0882-7974.5.3.335

17.

Fung

H. H.

Carstensen

L. L.

(2006). Goals change when life’s fragility is primed: Lessons learned from older adults, the September 11 attacks and Sars. Social Cognition, 24, 248-278. doi:10.1521/soco.2006.24.3.248

18.

Fung

H. H.

Carstensen

L. L.

Lutz

A. M.

(1999). Influence of time on social preferences: Implications for life-span development. Psychology and Aging, 14, 595-604. doi:10.1037/0882-7974.14.4.595

19.

Hau

Pleskac

T. J.

Hertwig

(2010). Decisions from experience and statistical probabilities: Why they trigger different choices than a priori probabilities. Journal of Behavioral Decision Making, 23, 48-68. doi:10.1002/bdm.665

20.

Hertwig

Erev

(2009). The description-experience gap in risky choice. Trends in Cognitive Sciences, 13, 517-523. doi:10.1016/j.tics.2009.09.004

21.

Isaacowitz

D. M.

Wadlinger

H. A.

Goren

Wilson

H. R.

(2006a). Is there an age-related positivity effect in visual attention? A comparison of two methodologies. Emotion, 6, 511-516. doi:10.1037/1528-3542.6.3.511

22.

Isaacowitz

D. M.

Wadlinger

H. A.

Goren

Wilson

H. R.

(2006b). Selective preference in visual fixation away from negative images in old age? An eye-tracking study. Psychology and Aging, 21, 40. doi:10.1037/0882-7974.21.1.40

23.

Johnson

J. E. V.

Bruce

A. C.

(1993). Gluck’s second law: An empirical investigation of horserace betting in early and late races. Psychological Reports, 72, 1251-1258. doi:10.2466/pr0.1993.72.3c.1251

24.

Kahneman

Tversky

(1979). Prospect theory: An analysis of decision under risk. Econometrica, 47, 263-291. doi:10.2307/1914185

25.

Kahneman

Tversky

(1984). Choices, values, and frames. American Psychologist, 39, 341-350. doi:10.1037/0003-066X.39.4.341

26.

Kelley

N. J.

Eastwick

P. W.

Harmon-Jones

Schmeichel

B. J.

(2015). Jealousy increased by induced relative left frontal cortical activity. Emotion, 15, 550-555. doi:10.1037/emo0000068

27.

Kennedy

Mather

Carstensen

L. L.

(2004). The role of motivation in the age-related positivity effect in autobiographical memory. Psychological Science, 15, 208-214. doi:10.1111/j.0956-7976.2004.01503011.x

28.

Kluger

A. N.

Stephan

Ganzach

Hershkovitz

(2004). The effect of regulatory focus on the shape of probability-weighting function: Evidence from a cross-modality matching method. Organizational Behavior and Human Decision Processes, 95, 20-39. doi:10.1016/j.obhdp.2004.05.003

29.

Kopelman

R. E.

Minkin

B. L.

(1991). Toward a psychology of parimutuel behavior: Test of Gluck’s laws. Psychological Reports, 68, 701-702. doi:10.2466/pr0.1991.68.2.701

30.

Kühberger

(1998). The influence of framing on risky decisions: A meta-analysis. Organizational Behavior and Human Decision Processes, 75, 23-55. doi:10.1006/obhd.1998.2781

31.

Kühberger

Wiener

(2012). Explaining risk attitude in framing tasks by regulatory focus: A verbal protocol analysis and a simulation using fuzzy logic. Decision Analysis, 9, 359-372. doi:10.1287/deca.1120.0254

32.

Larrick

R. P.

(1993). Motivational factors in decision theories: The role of self-protection. Psychological Bulletin, 113, 440-450. doi:10.1037/0033-2909.113.3.440

33.

Larrick

R. P.

Heath

(2009). Goal-induced risk taking in negotiation and decision making. Social Cognition, 27, 342-364. doi:10.1521/soco.2009.27.3.342

34.

Leigh

B. C.

(1999). Peril, chance, adventure: Concepts of risk, alcohol use and risky behavior in young adults. Addiction, 94, 371-383. doi:10.1046/j.1360-0443.1999.9433717.x

35.

Levin

I. P.

Gaeth

G. J.

Schreiber

Lauriola

(2002). A new look at framing effects: Distribution of effect sizes, individual differences, and independence of types of effects. Organizational Behavior and Human Decision Processes, 88, 411-429. doi:10.1006/obhd.2001.2983

36.

Levin

I. P.

Schneider

S. L.

Gaeth

G. J.

(1998). All frames are not created equal: A typology and critical analysis of framing effects. Organizational Behavior and Human Decision Processes, 76, 149-188. doi:10.1006/obhd.1998.2804

37.

Epley

(2009). When the best appears to be saved for last: Serial position effects on choice. Journal of Behavioral Decision Making, 22, 378-389. doi:10.1002/bdm.638

38.

Lockenhoff

C. E.

Carstensen

L. L.

(2007). Aging, emotion, and health-related decision strategies: Motivational manipulations can reduce age differences. Psychology and Aging, 22, 134-146. doi:10.1037/0882-7974.22.1.134

39.

Loewenstein

G. F.

Weber

E. U.

Hsee

C. K.

Welch

(2001). Risk as feelings. Psychological Bulletin, 127, 267-286. doi:10.1037/0033-2909.127.2.267

40.

Mather

Canli

English

Whitfield

Wais

Ochsner

. . . Carstensen

L. L.

(2004). Amygdala responses to emotionally valenced stimuli in older and younger adults. Psychological Science, 15, 259-263. doi:10.1111/j.0956-7976.2004.00662.x

41.

Mather

Carstensen

L. L.

(2003). Aging and attentional biases for emotional faces. Psychological Science, 14, 409-415. doi:10.1111/1467-9280.01455

42.

Mather

Carstensen

L. L.

(2005). Aging and motivated cognition: The positivity effect in attention and memory. Trends in Cognitive Sciences, 9, 496-502. doi:10.1016/j.tics.2005.08.005

43.

McGlothlin

W. H.

(1956). Stability of choices among uncertain alternatives. The American Journal of Psychology, 69, 604-615. doi:10.2307/1419083

44.

McKenzie

C. R. M.

Sher

Müller-Trede

Lin

Liersch

M. J.

Rawstron

A. G.

(2016). Are longshots only for losers? A new look at the last race effect. Journal of Behavioral Decision Making, 29, 25-36. doi:10.1002/bdm.1873

45.

Metzger

M. A.

(1985). Biases in betting: An application of laboratory findings. Psychological Reports, 56, 883-888. doi:10.2466/pr0.1985.56.3.883

46.

Mishra

(2014). Decision-making under risk: Integrating perspectives from biology, economics, and psychology. Personality and Social Psychology Review, 18, 280-307. doi:10.1177/1088868314530517

47.

Mishra

Fiddick

(2012). Beyond gains and losses: The effect of need on risky choice in framed decisions. Journal of Personality and Social Psychology, 102, 1136-1147. doi:10.1037/a0027855

48.

Mishra

Hing

L. S. S.

Lalumiere

M. L.

(2015). Inequality and risk-taking. Evolutionary Psychology, 13, 1-11. doi:10.1177/1474704915596295

49.

O’Brien

Ellsworth

P. C.

(2012). Saving the last for best. Psychological Science, 23, 163-165. doi:10.1177/0956797611427408

50.

Pennebaker

J. W.

Dyer

M. A.

Caulkins

R. S.

Litowitz

D. L.

Ackreman

P. L.

Anderson

D. B.

. . . McGraw

K. M.

(1979). Don’t the girls’ get prettier at closing time: A country and Western application to psychology. Personality and Social Psychology Bulletin, 5, 122-125. doi:10.1177/014616727900500127

51.

Pirlott

A. G.

MacKinnon

D. P.

(2016). Design approaches to experimental mediation. Journal of Experimental Social Psychology, 66, 29-38. doi:10.1016/j.jesp.2015.09.012

52.

Pratt

J. W.

(1964). Risk aversion in the small and in the large. Econometrica, 32, 122-136. doi:10.2307/1913738

53.

Reed

A. E.

Chan

Mikels

J. A.

(2014). Meta-analysis of the age-related positivity effect: Age differences in preferences for positive over negative information. Psychology and Aging, 29, 1-15. doi:10.1037/a0035194

54.

Sacchi

Stanca

(2014). Asymmetric perception of gains versus non-losses and losses versus non-gains: The causal role of regulatory focus. Journal of Behavioral Decision Making, 27, 48-56. doi:10.1002/bdm.1786

55.

Samanez-Larkin

G. R.

Carstensen

L. L.

(2011). Socioemotional functioning and the aging brain. In Decety

Cacioppo

J. T.

(Eds.), The Oxford handbook of social neuroscience (pp. 507-521). New York, NY: Oxford University Press. doi:10.1093/oxfordhb/9780195342161.013.0034.

56.

Scholer

A. A.

Zou

Fujita

Stroessner

S. J.

Higgins

E. T.

(2010). When risk seeking becomes a motivational necessity. Journal of Personality and Social Psychology, 99, 215-231. doi:10.1037/a0019715

57.

Schryer

Ross

(2014). Does the age-related positivity effect in autobiographical recall reflect differences in appraisal or memory? The Journals of Gerontology, Series B, 69, 548-556. doi:10.1093/geronb/gbt047

58.

Schwager

Rothermund

(2012). Motivation and affective processing biases in risky decision making: A counter-regulation account. Journal of Economic Psychology, 38, 111-126. doi:10.1016/j.joep.2012.08.005

59.

Shiv

Loewenstein

Bechara

Damasio

A. R.

(2005). Investment behavior and the negative side of emotion. Psychological Science, 16, 435-439. doi:10.1111/j.0956-7976.2005.01553.x

60.

Snowberg

Wolfers

(2010). Explaining the favorite–longshot bias: Is it risk-love or misperceptions? Journal of Political Economy, 118, 723-746. doi:10.1086/655844

61.

Tversky

Kahneman

(1981). The framing of decisions and the psychology of choice. Science, 211, 453-458. doi:10.1126/science.7455683

62.

Zou

Scholer

A. A.

Higgins

E. T.

(2014). In pursuit of progress: Promotion motivation and risk preference in the domain of gains. Journal of Personality and Social Psychology, 106, 183-201. doi:10.1037/a0035391

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.16 MB