1 Introduction

Research has shown that decision-making can be influenced by a variety of factors such as norms, framing, and cognitive biases. One such bias proposed by Kahneman and Tversky’s [1] prospect theory is loss aversion, which refers to the finding that individuals are more averse to losses than they are attracted to equivalent gains. This bias has often been examined in the context of risky choices in which individuals are reluctant to accept a symmetric bet with an equal chance of winning or losing a fixed dollar amount because the attraction to the gain offered is not enough to outweigh the aversion to the risk of incurring a loss. Individuals have even been shown to avoid bets that involved a loss, but had positive expected values [2]. Individual differences exist in the extent to which people display loss aversion [3]. One variable that could moderate the size of the loss aversion bias is symptomatology associated with Attention-Deficit/Hyperactivity Disorder (ADHD). ADHD is a neurodevelopmental disorder characterized by inattention, hyperactivity, and impulsivity [4]. Studying loss aversion in individuals with varying levels of ADHD symptomatology will increase knowledge of how these symptoms are related to decision-making, which could further our understanding of why those affected by the disorder engage in risky behavior and inform interventions to improve outcomes.

1.1 Individual differences in loss aversion

Researchers have examined which factors predict the strength of the loss aversion bias. Demographically, those middle aged or older exhibited significantly greater loss aversion than people in younger age categories, and the more education individuals had, the less loss averse they were [5, 6]. Personality and psychopathology moderators have also been examined. Individuals high in conscientiousness reported a greater loss aversion effect than those low in conscientiousness [7]. Loss aversion was also higher in individuals diagnosed with Major Depressive Disorder compared to controls [8] and in unmedicated individuals with Obsessive Compulsive Disorder compared to both medicated patients and controls [9]. In contrast, low loss aversion has been found to be an independent risk factor for cigarette smoking, particularly among young adults [10], and has been associated with alcohol and drug use [11,12,13], pathological gambling [12], and internet gaming disorder [14].

Tom et al. [15], who found evidence of neural loss aversion (i.e., reduced activation to losses in the dorsal and ventral striatum and ventromedial prefrontal cortex was greater than the increased activation to equivalent gains in the same areas), speculated that individual differences in loss aversion could relate to variation in individuals’ dopamine systems, based on their findings in key dopaminergic regions. Trémeau et al. [16] used this speculation to predict that individuals with schizophrenia, a disorder typically associated with dopamine dysfunction, would be less loss averse than non-patient controls. Their data revealed that loss aversion was absent in those with schizophrenia. Currie et al. [17] replicated findings of blunted loss aversion in schizophrenia and found that severity of psychotic symptoms was negatively correlated with loss aversion. Even in a non-clinical university sample, loss aversion was significantly lower in the group with high negative symptoms (e.g., avolition, reduced emotional expression), which are associated with various mental disorders such as schizophrenia, compared to the group with low negative symptoms [18].

1.2 Links between loss aversion and ADHD

Another clinical disorder associated with dopamine dysfunction is ADHD. Although the biological processes underlying ADHD are not fully known, there is evidence that variations in genes responsible for dopamine receptor regulation were related to ADHD risk and alterations in brain structure and function in dopamine-rich regions have been found in those with ADHD compared to non-affected controls (see [19] for a brief review). Thus, it is possible that dopamine dysfunction could be a common link between ADHD and loss aversion.

A small body of research has examined punishment sensitivity in ADHD, which could have implications for decisions about gambles involving loss. Studies in children using tasks that include monetary response cost have found decreased sensitivity to punishment in those with ADHD compared to controls [20]. However, other research has indicated that children with ADHD have a greater sensitivity to monetary losses than controls [21, 22]. Still, other findings suggest that children with ADHD were equally as sensitive to punishments as controls, but only when punishments were frequent [23].

Research on punishment sensitivity in adults is similarly conflicting. Braaten and Rosen [24] found that undergraduates high in ADHD symptoms reported significantly lower subjective emotional reactions in response to punishment situations compared to a low-symptom group. Adults with ADHD were shown to discount future losses as much as gains, whereas control participants discounted future losses less than gains, possibly due to greater loss aversion [25]. Tanaka et al. [25] also found lower activity in the caudate in adults with ADHD compared to controls, but only in the loss condition. Interestingly, Stoy et al. [26] found that adult males who had childhood ADHD but were never treated with medication showed less activity than control participants in the left insula when they were able to avoid loss on a task, but those same participants showed greater activation in the left/right insula to the outcome of loss compared to controls. Stronger neural responses to cues of loss have also been found in adults with ADHD compared to controls [27].

1.3 The present study

Examining loss aversion in the context of ADHD is important because it has implications for understanding the relationship between ADHD and engagement in risky behavior. Individuals with ADHD have been shown to be more likely to engage in lab-based risk-taking as well as a variety of real-life risk-taking behavior such as dangerous driving, substance use, risky sexual activity, and problematic gambling [28]. Furthermore, higher ADHD symptoms (both inattention and hyperactivity/impulsivity) measured dimensionally in a general sample of adults were found to be correlated with a higher likelihood of engaging in risky behavior across a variety of domains such as health, recreation, finances, and ethics [29]. Pollak et al. [28] suggested that there are decision-making factors that could help explain the relationship between ADHD and risky behavior such as loss amounts or probabilities being underweighted. We already know that lower loss aversion is associated with a variety of risky behaviors such as substance use and pathological gambling [12]. Establishing a relationship between loss aversion and ADHD is an important next step in determining whether or not this bias could be an explanatory mechanism for the association of ADHD and risk-taking behavior.

In the present study, we examined the extent to which college students with high ADHD symptomatology exhibited loss aversion compared to individuals with low ADHD symptomatology by asking participants to rate their likelihood of accepting bets and to choose between bets involving losses and sure values as measures of loss aversion. Since many items on the survey involved choices between an offer with certainty and a bet with different possible outcomes, participants’ responses may have partially depended on individual willingness to take risks, which has been shown to be higher in adults with ADHD [30], or on their preference for gambles. In order to determine whether or not these potential individual differences between high-symptomatology participants and low-symptomatology participants were contributing to observed effects, we included both the Balloon Analogue Risk Task (BART) [31] and a measure of gambling preference as potential covariates.

One study examining the relationship between ADHD and decision making about losses has used a task that assesses loss aversion more directly by asking participants to choose between certain and risky alternatives, albeit while controlling for the expected value of both [32]. No differences between adolescents with ADHD and controls were found on gambles assessing loss aversion, but sample size was small (n per group < 38), possibly indicating an underpowered study. The conflicting findings on punishment sensitivity in individuals with ADHD reviewed above make it unclear whether affected individuals would be more or less loss averse than non-affected individuals. If those with high ADHD symptomatology are more sensitive to punishment and weigh losses more heavily in making choices, they may exhibit a stronger loss aversion bias than those low in ADHD symptomatology. However, if those with greater symptoms are less sensitive to punishment they may be less loss averse than those with fewer symptoms. Therefore, we hypothesized that individuals with high and low ADHD symptomatology would differ in terms of the loss aversion bias, but we did not specify a direction for the difference. We also anticipated that controlling for baseline riskiness and gambling preference would attenuate, but not negate, the findings.

2 Method

2.1 Participants

A total of 98 undergraduates (73% women, 2% non-binary) from a small private institution in New York who were between the ages of 18 and 22 years (M = 19.6, SD = 1.2) participated in the present study on loss aversion. The majority of participants (63%) were White, with Asian/Pacific Islander (18%), Hispanic (6%) and Black/African American (5%) also represented. A priori power analyses indicated that a minimum of 90 participants were needed in order to detect medium-sized effects with a power of 0.80. Students were self-selected during recruitment and were compensated for their participation with extra credit in their psychology course or $5 in cash.

2.2 Measures

2.2.1 ADHD symptomatology

Given that there are pros and cons to several available measures of ADHD symptomatology in adults that are based solely on DSM criteria [e.g., 33], we used a combination of items that included all 18 DSM-IV-based items from the Adult ADHD Self-Report Scale [34] with nine additional items from the Adult Rating Scale [35] (impulsive, difficulty establishing/maintaining routine, difficulty delaying gratification, inconsistent school/work performance, impatient, explosive temper, physically daring activities/reckless, getting into trouble with the law, performance in school/work below competence), and eight items recommended by Barkley et al. [36] as important to the diagnosis of adult ADHD: thinking before acting, difficulty stop** activities/behavior, rushing through tasks, poor follow through on promises/commitments, driving motor vehicle faster than others, resisting temptation, starting project without reading/listening to directions, doing things in proper order/sequence. Students rated these 35 items on a scale from 1 = Not at all to 4 = Very Often. Of the 35 items, seven were reverse worded. Reliability analyses were conducted separately for each class year and means were computed for each participant for both inattentive (n = 15 items) and hyperactive/impulsive (n = 20 items) subscales. Cronbach’s alphas were 0.77 or higher for both subscales across all class years.

2.2.2 Loss aversion

Items measuring loss aversion (see Online Resource 1) were based on findings from previous research or used in previous studies [e.g., 37, 38]. Six items asked participants to make a forced-choice between $0 with certainty and a 50/50 bet involving losses. These items were combined into one score by summing the number of times the participant engaged in loss aversion by choosing the $0 with certainty option (Cronbach’s α = 0.71). Two items asked participants how likely it was that they would accept a specific bet. Participants rated their answers on a scale from 1 = very unlikely to 5 = very likely. We included these items in order to vary response format and detect differences that may be missed by forcing participants to choose between two options. Responses to both of these rating-scale items were highly correlated with the summary score from the forced-choice items: for 100/100 bet, r(95) = − 0.57, p < 0.001 [− 0.69, − 0.41]; for 60/25 bet, r(95) = − 0.27, p = 0.008 [− 0.44, − 0.07], suggesting the two methods for measuring loss aversion were tap** into the same construct.

2.2.3 Risky behavior

Risky behavior was measured with a modified BART [31]. The BART is a computer task that involves participants earning money by pum** up 30 balloons using the cursor. Participants pumped a single balloon in each trial with each pump equaling $0.05. Participants could stop pum** at any time, bank the money they earned, and move on to the next trial. If participants continued pum** a balloon until it exploded (participants were told the balloons could explode at any time), they lost all the money earned within that trial. Riskiness on the BART is measured by the adjusted average number of pumps on unexploded balloons with higher scores indicating more risky behavior. The BART has been shown to have adequate test–retest reliability [39] as well as support for convergent and discriminant validity [31].

In the original BART participants were given the money they earned by the end of the task. We modified the BART such that participants were told the goal of the task was to try to make as much money as possible and we would record their total. Because of budgetary constraints, participants knew they were playing with hypothetical money. Previous research has modified the BART in this manner and still found support for its convergent validity [e.g., 40].

2.2.4 Gambling preference

Five questionnaire items included bets that did not involve losses and served as a measure of an individual’s preference for either an offer with certainty or a gamble (see Online Resource 1). A composite measure of preference for gambles that equaled the number of times the participant chose the bet for these five items was not reliable (Cronbach’s α = 0.22). A principal components analysis with oblimin rotation indicated a three-factor structure with a single item (choose between $2500 with certainty and a bet with a 50% chance of winning $3000 and a 50% chance of gaining $1000) accounting for the most variance (26.94%). Compared to the other four items, this single item had the largest difference between the expected value of the gamble and the value being offered for sure, which means that those who chose the gamble on this item (n = 26) likely had a strong preference for gambles regardless of the outcome. This single item was also the only item out of the five that was significantly related to ADHD-symptomatology group status, Χ2(1, N = 98) = 9.50, p = 0.022, Φ = 0.31. Therefore, we chose to use this single item as our measure of gambling preference.

2.3 Procedure

Participants completed the ADHD symptomatology measure online via Qualtrics the summer before they entered college as part of a broader Institutional Research data collection effort that targeted entire first-year classes. Data were collected from first-years who started in the fall of 2010 through 2014, 2019, and 2022. An average of 211 students per year completed the symptomatology measure across those 7 years. Quartiles were calculated for the distribution of inattention and hyperactivity/impulsivity mean scores in each class year such that students were in the high-symptomatology group if they were at or above the 75th percentile for both inattention and hyperactivity/impulsivity and students were in the low-symptomatology group if they were at or below the 25th percentile for both subscales as compared to their peers in the same class year. Participants in the high- (n = 50) and low- (n = 48) symptomatology groups from each class year were recruited through email at a later date [average of 4.0 (SD = 2.4) semesters after completing the symptomatology measure] to complete an in-person study on “cognitive processes during decision-making.”

Upon entering the lab room, participants sat at a computer and completed the BART and a Qualtrics questionnaire in a random order. The online questionnaire included the loss aversion and gambling preference items as well as nine filler items about making estimations (e.g., “How many bristles are on a toothbrush?”). The loss aversion and gambling preference items were ordered so that participants were not asked related or similar items in succession. Both parts of the study (ADHD symptomatology measure and loss aversion study) were approved by the Institutional Review Board and all participants provided informed consent.

3 Results

Data [41] were analyzed using IBM SPSS Statistics 27. Four participants skipped one loss aversion item each and the BART score was lost for another participant. Three of these participants were in the high-ADHD-symptomatology group and two were in the low-symptomatology group. Descriptive statistics for all major continuous variables are reported in Table 1. The distributions for the ADHD symptomatology variables and the loss aversion scale item with the symmetric $100 bet were slightly positively skewed, whereas the loss aversion summary score was slightly negative skewed. None of the variables in Table 1 contained outliers in their distributions (all z-scores were less extreme than ± 3.00). Table 1 also displays mean comparisons on the major continuous variables for the high- and low-symptomatology groups. In addition, the two symptom groups did not differ in gender composition [χ2 (2, N = 98) = 3.96, p = 0.138] or age [t(96) = -1.39, p = 0.167].

Table 1 Descriptive statistics and comparison of means for major study variables as a function of ADHD symptomatology
Full size table

3.1 Is loss aversion related to ADHD symptomatology?

We first examined group differences for the three loss aversion variables (see Table 1). Those with low ADHD symptomatology more frequently chose $0 with certainty than those with high ADHD symptomatology on the summary measure of the forced-choice loss aversion items. We then examined the rating-scale items assessing loss aversion. Those high and low in ADHD symptomatology did not differ in their likelihood to accept a bet with a 50% chance to win $60 or lose $25. Those with high ADHD symptomatology were more likely to accept a symmetrical bet with a 50% chance to win or lose $100 than individuals with low symptomatology.

We then wanted to determine whether each ADHD-symptomatology score significantly predicted the loss aversion variables after controlling for the other symptomatology score (e.g., is inattention related to the loss aversion summary score after controlling for hyperactivity/impulsivity). Scatterplots of all pairs of variables did not reveal non-linear relationships or outliers so we used partial correlations for these analyses (see Table 2). Inattention was not significantly related to any of the three loss aversion variables after controlling for hyperactivity/impulsivity. However, greater hyperactivity/impulsivity was related to less loss aversion as measured by the summary score and a greater likelihood to choose the symmetric $100 bet, even after controlling for inattention scores.

Table 2 Partial correlations between ADHD variables and loss aversion variables
Full size table

3.2 Controlling for risk taking and gambling preference

Because those with high ADHD symptomatology had a greater adjusted average number of pumps on the BART than those with low ADHD symptomatology (see Table 1), analyses were repeated controlling for this risk variable using partial correlation (see Table 2). Because ADHD group status was dichotomous, all analyses that included this variable used Spearman’s rho matrices (instead of Pearson’s correlations) as the input. ADHD group status continued to be significantly related to the total loss aversion score and rating-scale responses to the 50/50 gamble to win/lose $100, but was not associated with the likelihood to accept a bet with a 50% chance to win $60 or lose $25. Since scatterplots for all pairs of variables did not reveal evidence of non-linear relationships or outliers, we used partial correlations to examine the relationships between each ADHD-symptomatology score and the loss aversion variables after controlling for the other symptomatology score and adjusted average number of pumps on the BART. Inattention was not significantly related to any of the three loss aversion variables after controlling for hyperactivity/impulsivity and scores on the BART. Hyperactivity/impulsivity scores remained significantly associated with the likelihood to choose the 50/50 gamble to win/lose $100 once BART and inattention scores were controlled.

Because the preference for gambles item was related to ADHD-symptomatology group status and there was no significant relationship between the BART and the gambling preference item [t(95) = 1.71, p = 0.09], suggesting they were tap** into different constructs, partial correlation analyses were repeated also controlling for this item (see Table 2). Because this item was dichotomous, all partial correlations used Spearman’s rho matrices as the input. ADHD group status was no longer related to the total loss aversion score after controlling for both the BART and the gambling preference measure. However, ADHD group status remained significantly associated with the scale item assessing loss aversion via a symmetrical bet with a 50% chance to win or lose $100 over and above these two control variables. With regard to symptomatology scores, inattention was not significantly related to any loss aversion variables after controlling for hyperactivity/impulsivity, BART scores, and preference for gambles. However, hyperactivity/impulsivity did continue to be related to likelihood of choosing the symmetrical $100 bet after controlling for inattention, risky behavior, and gambling preference.

4 Discussion

The present study examined the relationship between individual differences in loss aversion in a risky gamble task and ADHD symptomatology. Individuals high in ADHD symptomatology were less loss averse than those low in ADHD symptomatology, even after controlling for general risk taking and gambling preference. These findings are consistent with research on adults with ADHD symptoms that found a decreased sensitivity to punishment and losses [24, 25]. However, it is also possible that those high in ADHD symptomatology were simply more attracted to the possible gains in the gambles measuring loss aversion than those low in symptomatology. Shoham et al. [29, 42] found that perception of the benefits of behavior explained the link between adult ADHD symptomatology and engagement in risky behavior. Therefore, it is unclear if the observed group differences in loss aversion arose because of a decreased aversion to losses, an increased attraction to gains, or both among those with high ADHD symptomatology.

Controlling for risk taking and gambling preference did not completely negate the group differences in loss aversion, which is consistent with research that has shown that what appears as risky decisions in individuals with ADHD is actually sub-optimal decision making [43], which could be due to difficulty understanding the expected value of gambles or generating the necessary mental effort to calculate expected value. Luman et al. [23] showed that children with ADHD were sensitive to the frequency of penalties, but blind to the magnitude of such penalties. The methodology in the present study relied on participants’ sensitivity to probabilities and magnitude, which could be impaired in those with high ADHD symptomatology.

Our finding that hyperactive/impulsive symptoms of ADHD predicted loss aversion responses over and above inattentive symptoms is consistent with Quay’s [44] theory involving the role of an underactive behavioral inhibition system. He postulated that the theory would not apply to individuals with the primarily inattentive type of ADHD, which makes the hyperactive/impulsive symptoms critical to the decreased responsiveness to signals of punishment.

In addition to sensitivity to punishment and reward, the observed findings could be explained by a difficulty in anticipating a loss in the future. Affective forecasting is the ability to anticipate a future emotional state, and individuals typically make large affective forecasting errors—in the case of loss aversion, they may anticipate that the loss will be more detrimental than a real loss is when it is experienced [45], which could motivate avoidance of the loss. It could be that those high in ADHD symptomatology failed to anticipate the negative impact of a future loss as there are known deficits in temporal discounting in adults with ADHD [46]. Changing the loss aversion paradigm so that the losses were immediate instead of hypothetical could help tease apart what is driving any differences related to ADHD symptomatology.

Interestingly, both those high and low in ADHD symptomatology were similarly likely to select a gamble when the potential loss was small (i.e., $25), which is consistent with prior research showing that loss aversion is reduced or absent for small sums [47]. In addition, the potential gain for this item ($60) was more than twice the potential loss, which previous research has found is the asymmetry necessary to view the options as equal and overcome loss aversion [32].

4.1 Limitations and future directions

The results of this study should be interpreted in light of several limitations. First, we used a small number of items to measure loss aversion. This study should be replicated with procedures that allow for systematic variation of magnitude and within-subjects statistical analyses [e.g., 47]. One additional limitation was the use of measures of loss aversion that involved theoretical values. The accuracy of participants’ responses may have depended on their ability to place themselves in theoretical situations. Future studies using the endowment effect as a measure of loss aversion could provide more real-life data and could also be less cognitively demanding because they don’t require reasoning about probabilities.

There were also limitations in how we examined ADHD, as we used a self-report measure. Self-report is an important initial step in this research because it is so heavily relied upon for assessment at the adult level, especially when reports from significant others or parents are unavailable. Even so, comprehensive information to diagnose students was not collected, nor was information about prior diagnoses of ADHD or other comorbid psychiatric conditions. However, a dimensional approach is important because both clinical and sub-clinical symptom expression can be impairing and related to similar mechanisms. One may miss important information if only diagnosed individuals were examined.

Finally, there were limitations in how we created our ADHD symptom groups. High ADHD symptomatology was defined as being at or above the 75th percentile compared to one’s peers who entered the same college in the same year. We did not use a norm-based measure for ADHD symptomatology, which means that those in the high-symptom group could have displayed a level of symptoms that was lower than what one would expect for an individual with a diagnosis. We also defined high symptomatology as being elevated on both inattention and hyperactivity-impulsivity, which corresponds to only one of the three possible presentation types of ADHD (i.e., combined presentation) [4]. Future research should examine loss aversion in individuals who present with elevated symptoms in only one of the two ADHD-symptom domains to further understand if and how the two types of symptoms independently predict aversion to losses.

4.2 Conclusions and implications

Despite these limitations, the present study was the first to find a link between ADHD symptomatology and individual differences in loss aversion and has implications for college staff who are creating policies to promote positive behavior. Framing change on a dimension as a loss could be less effective for students high in ADHD symptomatology. For example, framing grading as starting with 100 points and losing points has been shown to be motivating for the average loss averse student [48]. However, framing grading as starting at zero and earning points may be more effective for those who are high in ADHD symptomatology. Moreover, many colleges implement sanctions that involve losses for undesired behavior, such as a lower housing lottery number or restrictions on off-campus study. It may be the case that for individuals high in ADHD symptomatology a more motivating form of behavior management would focus on gains (e.g., a better housing lottery number for desired behavior such as community leadership). Future research will be important to determining any potential differences based on ADHD symptomatology in sensitivities to real-world punishments that involve losses.

Moreover, a decreased aversion to negative outcomes in the form of loss could be one of the mechanisms by which ADHD symptomatology is related to engagement in risk-taking behavior. Understanding the cognitive decision-making mechanisms associated with engagement in risky behavior can contribute to the identification of individuals with ADHD who are most at risk for negative outcomes as well as the development of interventions aimed at reducing such behavior, which could also reduce the likelihood of that behavior turning into a comorbid addiction disorder. Loss aversion has been shown to be malleable using a cognitive regulation strategy [49] and, therefore, might be a fruitful target for intervention should future research determine that it precedes and mediates risk-taking behavior in those with ADHD.