https://orcid.org/0000-0001-6710-8575
Abstract
Contextually driven decision making is multidimensional, as individuals need to contend with prioritizing both competing and complementary demands. However, data is limited as to whether temporal discounting rates vary as a function of framing (gains vs loss) and domain (monetary vs social) in middle-to-older aged adults. It is also unclear whether socioaffective characteristics like social isolation and loneliness are associated with temporal discounting.
Temporal discounting rates were examined across monetary gain, monetary loss, social gain, and social loss conditions in 140 adults aged 50–90 during the Omicron stage of the pandemic. Self-report measures assessed loneliness and social isolation levels.
Results found evidence of steeper temporal discounting rates for gains as compared to losses in both domains. Social outcomes were also more steeply discounted than monetary outcomes, without evidence of an interaction with the framing condition. Socioeconomic and socioaffective factors were unexpectedly not associated with temporal discounting rates.
Community-dwelling middle-to-older aged adults showed a preference for immediate rewards and devalued social outcomes more than monetary outcomes. These findings have implications for tailoring social and financial incentive programs for middle to later adulthood.
Decision making is a complex process drawing on one’s abilities to reason, plan, and weigh the costs and benefits of outcomes. Increasing evidence suggests that decision-making abilities change as a function of normal age-related changes in cognitive function (for review, see Hess et al., 2015). For example, changes in frontal lobe function and age-related decrements in frontal-subcortical circuits appear to hinder decision-making capabilities in older adults by way of difficulties with executive functions such as cognitive control and reward valuation (see Lighthall, 2020). Additionally, age-related affective changes (e.g., shifts in priorities, values, and goals; for review, see Carstensen & Mikels, 2005) along with potential changes in social landscapes characterized by losses (e.g., death of loved ones, ailing health, and geographical barriers) may collectively and uniquely affect decision making by promoting a greater desire for more immediate outcomes.
Given the significance of decisions that are made by older adults (e.g., retirement, navigating complex medical care, financial investments, spending enough time with family and friends), and that there are often competing demands (e.g., time spent at work vs family), better understanding how older adults respond to choices can inform public health messages and identify ways to help older adults optimize their choices. However, while there is extensive research investigating monetary reward paradigms in older adults, no research to our knowledge has examined potential differences in affective framing (i.e., gains vs loss) and different functional domains. Hence, as a first step, this study aimed to better understand how middle-to-older adults respond to affective framing within social as compared to financial decision-making tasks.
Decision-Making Theories
It is well established that individuals behave differently based on how decisions are framed relative to a reference point (Kahneman & Tversky, 1979) as well as past experiences and affective reactions (Betsch, 2005). Specifically, individuals tend to view outcomes as either gains or losses from a subjective reference point (Kahneman & Tversky, 1979), as well as discount, or devalue, outcomes as a function of time (Mazur, 1987). This tendency, known as temporal discounting, results in individuals more frequently choosing smaller, immediate gains over larger, delayed gains; but larger, delayed losses over smaller, immediate losses. This latter preference may be a product of loss aversion, which describes the greater emotional impact an individual feels from a perceived or actual loss as opposed to a comparable perceived or actual gain (Kahneman & Tversky, 1979).
Temporal Discounting Rates
Much of the temporal discounting research in older adults has focused on monetary gain scenarios and yielded mixed results (for review, see Seaman et al., 2022). Although this offers important insight into reward-based behavior and preferences, little is known regarding decision-making preferences in loss conditions. This remains a problematic gap given many real-life decisions involve losses. Additionally, waning time horizons associated with age impose limited opportunities for remedying potential consequences of poor financial decision making, particularly as they relate to losses. Furthermore, decisions are rarely unidimensional, and later life typically requires making critical monetary and social decisions that involve weighted decisions and sometimes sacrificing one domain over another (e.g., time spent with loved ones vs time making more money). Of the few studies that have examined loss paradigms in older adults (Halfmann et al., 2013; Löckenhoff et al., 2011; Sparrow & Spaniol, 2018), primarily monetary scenarios are used that suggest shallow discounting for losses (e.g., the tendency to prefer a smaller, immediate loss as compared to a larger, and delayed loss).
Socioaffective Factors and Decision Making
Careful consideration of the environment and other contextual factors is crucial for interpreting the appraisal of decision-making outcomes (Mata et al., 2012). Evidence suggests that older adults may exhibit specific patterns of temporal discounting rates as a product of unique age-related socioaffective changes such as increases in prosocial behavior, accumulated experiences, and waning time horizons (Carstensen, 2006; Löckenhoff & Rutt, 2015; Löckenhoff & Samanez-Larkin, 2020; Seaman et al., 2016; Sparrow & Spaniol, 2018; Sparrow et al., 2019). Socioaffective factors can play an important role in decision making (for review, see Kensinger & Gutchess, 2017) and are also integral to well-being and health (Holt-Lunstad et al., 2010).
Evidence suggests that older adults show a greater preference for immediate social rewards such as spending time with a close friend (Seaman et al., 2016), and make more altruistic intertemporal monetary choices (Sparrow & Spaniol, 2018) than younger adults. However, the impact of negative valence states, such as loneliness, on temporal discounting behavior among older adults is relatively unknown. Loneliness is a negative emotional state related to perceived social isolation, which evolutionarily may have evolved to promote survival and maintain social connections (Hawkley & Cacioppo, 2010). As such, loneliness would likely increase sensitivity to delay periods, particularly for socially based decision making. This tendency may be particularly pronounced in older adults given their increased risk for loneliness and social isolation (see Malcolm et al., 2019). There is significant heterogeneity regarding the pattern of temporal discounting among older adults within the literature, in which examining potential differences across both framing (e.g., gain vs loss) and domain (e.g., money vs social) may provide contextual clarity.
Objectives
Decision making relative to loss and gains within monetary and social conditions remains understudied and inconclusive in older adult populations. To address current gaps in the literature, this study comprehensively investigated decision-making preferences as a function of time, framing, and functional domain. A two-way ANOVA was used to investigate whether temporal discounting rates differed as a function of framing (gains vs loss) and domain (social vs financial). We specifically aimed to determine whether there was an interaction between condition and framing. We hypothesized that middle-to-older aged adults would discount social rewards more strongly than monetary rewards based on prior work (Seaman et al., 2016), and extended this hypothesis to losses. We also expected that there would be evidence of a main effect of framing condition that would reflect steeper discounting for gains as compared to loss. Last, we explored whether temporal discounting rates were associated with social isolation and loneliness, given their relationship with socioaffective experience, and that this study occurred in the context of the coronavirus disease-2019 (COVID-19) pandemic’s mandated social distancing procedures.
Method
Participants
Participants (n = 140; 50–90 years) were recruited as part of the second wave of the Maine-Aging Behavior Learning Enrichment (M-ABLE) study conducted at the University of Maine. Initial power analyses using G*Power (Faul et al., 2009) indicated that a sample of 134 participants was sufficiently powered (0.80) to detect within-group differences between decision-making conditions based on an expected effect size of 0.25. As temporal discounting and framing paradigms effect sizes range from small to large (e.g., Seaman et al., 2022), a 0.25 effect size was chosen based on practically significant effects commonly observed in the social sciences (Ferguson, 2009). The study used community-based recruitment methods to enhance the recruitment of a socioeconomically diverse sample of middle-to-older aged adults. Participants were screened for eligibility via phone, and study visits were conducted via Zoom and telephone following Center for Disease Control (CDC) COVID-19 guidelines during the Delta and Omicron periods of the COVID-19 pandemic (November 2021 to May 2022). Study inclusion criteria were intentionally wide to improve the generalizability of findings and included: the ability to participate via Zoom given the modality of the decision-making paradigm, being 45 years of age or older, and willingness to complete assessment measures. Exclusion criteria included: a history of a neurodegenerative disorder, physical limitations (e.g., hearing impairments) that would preclude completion of study measures, or currently receiving treatment for a dementia disorder.
Participants were first screened for eligibility and underwent informed consent procedures approved by the University of Maine Institutional Review Board. Study visits were scheduled in the morning to lessen the effects of the time of day on cognitive performance unless the participant reported having later sleep-wake cycles, in which the time of day was adjusted to account for this factor (i.e., 11 am or 12 pm appointments). Participants were compensated with a US$25 gift card to a large regional grocery store for completion of the study.
Measures
Demographic information
Demographic information (e.g., age, sex, and years of education) was obtained through a clinical interview. Participants’ approximate family income (including wages, disability payment, retirement income, and welfare) was obtained via a confidential survey. Annual income levels were organized into nine categories along with the percentage of participants who comprised each category: (1) <US$10,000 (1%), (2) US$10,00–$19,999 (5%), (3) US$20,000–$29,999 (11%), (4) US$30,000–$39,999 (10%), (5) US$40,000–$49,999 (16%), (6) US$50,000–$59,999 (14%), (7) US$60,000–$69,999 (10%), (8) US$70,000–$100,000 (13%), and (9) >US$100,000 (19%).
Decision-making paradigm
Temporal discounting was assessed with a task adapted from a well-validated decision-making paradigm (Du et al., 2002) and implemented using E-prime Version 3.0. The decision-making task had four counterbalanced conditions (monetary gains, monetary losses, social gains, and social losses). In each condition, a series of hypothetical choices were presented that asked participants to choose between a smaller outcome available immediately or a larger outcome available after a specified delay period. The delay periods consisted of 1 week, 1 month, 6 months, and 1 year. The number of delayed outcomes stayed fixed across trials within each condition. The amount of the immediate outcome adaptively changed across trials, such that the options presented in the subsequent trial were contingent upon the choice made in the previous trial. Participants completed five practice trials for each condition before moving on to test trials to ensure comprehension of the task. Instructions for all conditions were adapted from Estle et al. (2006). Each participant made a total of 112 choices (7 trials at each delay period × 4 delay periods × 4 conditions).
The first trial presented in the monetary gain condition was a choice between gaining US$50 “now” and US$100 “later” in 1 week. Choices for each subsequent trial were adaptively generated based on the previously mentioned iterative process (Frye et al., 2016). The size of each subsequent adjustment across trials decreased with each trial until seven choices were made within each delay level, allowing for convergence upon the participant’s indifference point. This process was then repeated for each delay level, such that four separate indifference points were derived for each condition. For the monetary loss condition, the dollar amounts, and delay levels were identical to the monetary gain condition. The same adaptive procedure was also used, with one notable difference being the direction in which the adjustments across trials were made.
The social gains condition was adapted from a previous study (Seaman et al., 2016). The first trial presented in this condition was a choice between gaining 1.5 h “now” and 3 h “later” in 1 week. Participants were asked to imagine engaging in a social activity they enjoyed with a person they wished they could spend more time with, such as a loved one or a friend. The same adaptive process and the iterative algorithm used in the monetary gain condition were used in the social gain condition. For the social loss conditions, the amounts of time and delay levels were identical to the social gain condition, except participants were now asked to imagine losing time doing a social activity. The same adaptive procedure used in the monetary loss condition was used. See Figure 1 for a visual representation of the discounting tasks.
Example trials from the monetary gains, monetary loss, social gains, and social loss conditions. In all conditions, participants were presented with hypothetical situations in which they were instructed to make a choice between “now” and “later.” In the monetary trials, choices were made about either gaining or losing money. In the social trials, choices were made about either gaining or losing time spent with a loved one. The immediate and delayed choices randomly switched sides of the screen and participants communicated their choice to the examiner. Participants were reminded there were no right or wrong answers and to simply respond with their preference.
Socioaffective variables
Social isolation was measured using the National Alzheimer’s Coordinating Center COVID-19 Impact Survey (NACC, 2020). Participants responded to a single-item question that evaluated the extent to which they felt socially isolated from friends and family due to the COVID-19 pandemic on a 5-point Likert scale ranging from 1 (not at all isolated) to 5 (extremely isolated). Loneliness was assessed by the item: “I feel very lonely” rated on a 4-point Likert scale, ranging from 1 (not at all) to 4 (very much).
Preliminary Analyses
Missing data were minimal (less than 0.1% of total values). Series mean replacement was used to replace one missing income value. A comparative value based on the participant’s other condition discounting rates was imputed for one missing social discounting rate value. Descriptive statistics were generated for demographic, sociability, and discounting rate variables and inspected for outliers, skew, and kurtosis. Significant outliers (z-scores exceeding ± 3.29 standard deviations from the mean) were winsorized (Tabachnick & Fidell, 2007).
Discounting data was examined for validity by identifying participants who demonstrated a nonsystematic responding (NSR) pattern. NSR occurs when participants respond in such a way that their subjective values of delayed outcomes increase and decrease in a nonsensical way. Twenty-two participants were identified as NSR according to a variation of Criterion 1 of the algorithm proposed by Johnson and Bickel (2008), which meant that if any indifference point (starting with the second delay) was greater than the preceding indifference point by a magnitude greater than 30% of the larger later reward (i.e., US$30 for monetary trials and 0.9 h for social trials) it was considered NSR. The percentage of systematic data in the current sample was within the expected range (see Smith et al., 2018). Based on prior outlined recommendations (Johnson & Bickel, 2008; Smith et al., 2018), analyses were conducted using all available data.
Statistical Analyses
Discounting rates were derived using a partial hyperbolic function [V = A/(1 + kD)] posited to best represent and quantify temporal discounting rates (Mazur, 1987). Within this equation, V represents the subjective value of the delayed outcome, A is equal to the objective value of the delayed outcome, k is the discounting rate, and D is the magnitude of the delay period. Nonlinear curve-fitting analyses determined discounting rates. As expected, k-values were significantly skewed, non-normally distributed, and nonamenable to transformations. Therefore, to address issues of non-normality, a second well-established method for deriving discounting rates, known as area under the curve (AUC; Myerson et al., 2001) was conducted. Briefly, the AUC is calculated based on observed subjective values, rather than the values predicted by a particular theoretical equation. It is posited to have a normal distribution making it ideal for skewed data (Myerson et al., 2001). The AUC is normalized so that it ranges from 0.0 (steepest discounting) to 1.0 (no discounting; Myerson et al., 2001). AUC values met assumptions of normality for within-group comparison tests and a repeated measures design was used to assess the impact of framing (gain vs loss) and domain (monetary vs social) on AUC values.
A two-way repeated measures ANOVA evaluated the effect of framing (gain vs loss) and domain (monetary vs social) on AUC values. Bootstrapped hierarchical multiple regressions investigated the influence of relevant covariates on the decision-making outcomes. Univariate and multivariate assumptions of normality were examined, and appropriate nonparametric robust methods were used for hierarchical regressions. All tests of significance were two-tailed. Bonferroni corrections were applied within families of planned contrasts and adjusted p values are presented. Because p values can fluctuate based on which iteration of the bootstrap analysis is provided, variables were determined to be significant if, in addition to p < .05, bootstrapped bias-corrected accelerated 95% confidence intervals did not include the value zero. Partial eta squared served as a measure of effect size. Nonlinear curve-fitting analyses were performed using GraphPad Prism Version 9.0 software. All other statistical analyses were performed using SPSS Version 28.
Results
Descriptive Statistics
Table 1 shows the demographic and socioaffective characteristics of the sample. A broad range of both years of education (11–20 years) and annual income levels (range: <US$10,000–>US$100,000) was achieved. There was a larger proportion of women as compared to men. The sample was primarily White, which reflects the 94.2% non-Hispanic white demographics of Maine (U.S. Census Bureau, 2022). In an exit survey, 40% of participants reported some level of loneliness. Approximately two-thirds of participants endorsed feeling some degree of isolation from friends and family due to the COVID-19 pandemic.
Demographic and Socioaffective Characteristics of Sample (N = 140)
| Characteristic | Statistic |
|---|---|
| Age, M (SD) | 71.7 (7.5) |
| Years of education, M (SD) | 16.4 (2.2) |
| Median income range | US$50,000–$59,999 |
| Female, n (%) | 105 (75.0%) |
| Race: identified as White, n (%) | 139 (99.3%) |
| Loneliness, n (%) | |
| Not at all lonely | 84 (60.0%) |
| A Little lonely | 39 (27.9%) |
| Somewhat lonely | 14 (10.0%) |
| Very much lonely | 3 (2.1%) |
| Isolation, n (%) | |
| Not at all isolated | 48 (34.3%) |
| A Little isolated | 47 (33.6%) |
| Somewhat isolated | 25 (17.6%) |
| Very isolated | 15 (10.7%) |
| Extremely isolated | 5 (3.6%) |
| Characteristic | Statistic |
|---|---|
| Age, M (SD) | 71.7 (7.5) |
| Years of education, M (SD) | 16.4 (2.2) |
| Median income range | US$50,000–$59,999 |
| Female, n (%) | 105 (75.0%) |
| Race: identified as White, n (%) | 139 (99.3%) |
| Loneliness, n (%) | |
| Not at all lonely | 84 (60.0%) |
| A Little lonely | 39 (27.9%) |
| Somewhat lonely | 14 (10.0%) |
| Very much lonely | 3 (2.1%) |
| Isolation, n (%) | |
| Not at all isolated | 48 (34.3%) |
| A Little isolated | 47 (33.6%) |
| Somewhat isolated | 25 (17.6%) |
| Very isolated | 15 (10.7%) |
| Extremely isolated | 5 (3.6%) |
Note: SD = standard deviation.
Demographic and Socioaffective Characteristics of Sample (N = 140)
| Characteristic | Statistic |
|---|---|
| Age, M (SD) | 71.7 (7.5) |
| Years of education, M (SD) | 16.4 (2.2) |
| Median income range | US$50,000–$59,999 |
| Female, n (%) | 105 (75.0%) |
| Race: identified as White, n (%) | 139 (99.3%) |
| Loneliness, n (%) | |
| Not at all lonely | 84 (60.0%) |
| A Little lonely | 39 (27.9%) |
| Somewhat lonely | 14 (10.0%) |
| Very much lonely | 3 (2.1%) |
| Isolation, n (%) | |
| Not at all isolated | 48 (34.3%) |
| A Little isolated | 47 (33.6%) |
| Somewhat isolated | 25 (17.6%) |
| Very isolated | 15 (10.7%) |
| Extremely isolated | 5 (3.6%) |
| Characteristic | Statistic |
|---|---|
| Age, M (SD) | 71.7 (7.5) |
| Years of education, M (SD) | 16.4 (2.2) |
| Median income range | US$50,000–$59,999 |
| Female, n (%) | 105 (75.0%) |
| Race: identified as White, n (%) | 139 (99.3%) |
| Loneliness, n (%) | |
| Not at all lonely | 84 (60.0%) |
| A Little lonely | 39 (27.9%) |
| Somewhat lonely | 14 (10.0%) |
| Very much lonely | 3 (2.1%) |
| Isolation, n (%) | |
| Not at all isolated | 48 (34.3%) |
| A Little isolated | 47 (33.6%) |
| Somewhat isolated | 25 (17.6%) |
| Very isolated | 15 (10.7%) |
| Extremely isolated | 5 (3.6%) |
Note: SD = standard deviation.
Temporal Discounting Area Under the Curve
Table 2 shows the group median AUC values for all four conditions. A two-way repeated measures ANOVA was used to test the effect of framing (gain vs loss) and domain (monetary vs social) on AUC values. Results revealed a significant main effect of domain [F (1, 139) = 213.81, p < .001, ηp2 = 0.606] on AUC values, with social outcomes discounted more steeply than monetary outcomes. There was also a significant main effect of framing on AUC values, indicating that participants more steeply discounted gains as compared to losses [F (1, 139) = 92.95, p < .001, ηp2 = 0.401]. There was no significant interaction between frame and domain [F (1, 139) = .53, p = 0.459, ηp2 = 0.004] on AUC values. Overall, these results suggested greater sensitivity to delay periods for gains as compared to losses, as well as for social outcomes when compared to monetary outcomes. As lower AUC values reflect steeper discounting, Figure 2 shows these findings that gains were more steeply discounted than losses within both domains. All within-subject analyses remained significant after Bonferroni corrections.
Temporal Discounting Values
| Variable | Total |
|---|---|
| Monetary gains AUC | 0.73 (0.02) |
| Monetary losses AUC | 0.91 (0.01) |
| Social gains AUC | 0.46 (0.02) |
| Social losses AUC | 0.62 (0.03) |
| Variable | Total |
|---|---|
| Monetary gains AUC | 0.73 (0.02) |
| Monetary losses AUC | 0.91 (0.01) |
| Social gains AUC | 0.46 (0.02) |
| Social losses AUC | 0.62 (0.03) |
Notes: AUC = Area Under the Curve.
Values indicate Median (Standard Error).
Temporal Discounting Values
| Variable | Total |
|---|---|
| Monetary gains AUC | 0.73 (0.02) |
| Monetary losses AUC | 0.91 (0.01) |
| Social gains AUC | 0.46 (0.02) |
| Social losses AUC | 0.62 (0.03) |
| Variable | Total |
|---|---|
| Monetary gains AUC | 0.73 (0.02) |
| Monetary losses AUC | 0.91 (0.01) |
| Social gains AUC | 0.46 (0.02) |
| Social losses AUC | 0.62 (0.03) |
Notes: AUC = Area Under the Curve.
Values indicate Median (Standard Error).
Group median differences in Area Under the Curve (AUC) for all four conditions as a function of domain (monetary vs social) and frame (gain vs loss). A two-way repeated measures ANOVA revealed significant main effects of domain and framing on AUC values such that participants discounted social outcomes more steeply than monetary outcomes and gains more steeply than losses. As lower AUC values reflect steeper discounting, Figure 2 shows that gains were more steeply discounted than losses within both domains; error bars depict standard errors; **p < .001.
Associations Between AUC Values and Socioaffective Characteristics
A series of bootstrapped hierarchical multiple regressions were performed to quantify the independent contributions of education, income, loneliness, and isolation on the AUC values of each decision-making condition. Age, education, and income level were entered for Step 1. Loneliness and isolation levels were entered for Step 2. For the monetary gains condition, the predictors of age (β = .03, p = .737), education (β = .16, p = .070), income (β = .03, p = .741), loneliness (β = .06, p = .475), isolation (β = .14, p = .110) were not significant predictors of AUC values in the final model. Similarly, age (β = .02, p = .821), education (β = .06, p = .529), income (β = .04, p = .702), loneliness (β = .05, p = .602), and isolation (β = .14, p = .120) were not significant predictors of AUC values in the monetary loss condition. In the social gains condition, age (β = .14, p = .116), education (β = −.004, p = .963), income (β = −.02, p = .857), loneliness (β = .04, p = .654), isolation (β = .001, p = .992) were not significant predictors of AUC values in the final model. A similar pattern was observed in the social loss condition in which age (β = −.15, p = .070), education (β = −.07, p = .429), income (β = −.09, p = .308), loneliness (β = −.11, p = .204), and isolation (β = −.11, p = .219) were not significant predictors of AUC values. Table 3 shows the model summaries for each condition.
Bootstrapped Model Summaries for Hierarchical Regressions*
| Variable | R | R2 | Adj. R2 | ΔR2 | ΔF | p Value |
|---|---|---|---|---|---|---|
| Monetary gains model | ||||||
| 1a | 0.18 | 0.03 | 0.01 | 0.03 | 1.44 | .253 |
| 2b | 0.23 | 0.06 | 0.01 | 0.02 | 1.69 | .189 |
| Monetary losses model | ||||||
| 1a | 0.08 | 0.01 | −0.02 | 0.01 | 0.28 | .839 |
| 2b | 0.17 | 0.03 | −0.01 | 0.02 | 1.46 | .236 |
| Social gains model | ||||||
| 1a | 0.12 | 0.02 | −0.01 | 0.02 | 0.69 | .555 |
| 2b | 0.13 | 0.02 | −0.02 | 0.01 | 0.16 | .855 |
| Social losses model | ||||||
| 1a | 0.20 | 0.04 | 0.02 | 0.04 | 1.95 | .124 |
| 2b | 0.25 | 0.06 | 0.03 | 0.02 | 1.60 | .206 |
| Variable | R | R2 | Adj. R2 | ΔR2 | ΔF | p Value |
|---|---|---|---|---|---|---|
| Monetary gains model | ||||||
| 1a | 0.18 | 0.03 | 0.01 | 0.03 | 1.44 | .253 |
| 2b | 0.23 | 0.06 | 0.01 | 0.02 | 1.69 | .189 |
| Monetary losses model | ||||||
| 1a | 0.08 | 0.01 | −0.02 | 0.01 | 0.28 | .839 |
| 2b | 0.17 | 0.03 | −0.01 | 0.02 | 1.46 | .236 |
| Social gains model | ||||||
| 1a | 0.12 | 0.02 | −0.01 | 0.02 | 0.69 | .555 |
| 2b | 0.13 | 0.02 | −0.02 | 0.01 | 0.16 | .855 |
| Social losses model | ||||||
| 1a | 0.20 | 0.04 | 0.02 | 0.04 | 1.95 | .124 |
| 2b | 0.25 | 0.06 | 0.03 | 0.02 | 1.60 | .206 |
Notes: Adj. R2 = adjusted R2; R2Δ = change in R2; ΔF = change in the F statistic.
aAge, education, and income were variables.
bLoneliness and isolation were variables.
*All values based on 1,000 bootstrap samples.
Bootstrapped Model Summaries for Hierarchical Regressions*
| Variable | R | R2 | Adj. R2 | ΔR2 | ΔF | p Value |
|---|---|---|---|---|---|---|
| Monetary gains model | ||||||
| 1a | 0.18 | 0.03 | 0.01 | 0.03 | 1.44 | .253 |
| 2b | 0.23 | 0.06 | 0.01 | 0.02 | 1.69 | .189 |
| Monetary losses model | ||||||
| 1a | 0.08 | 0.01 | −0.02 | 0.01 | 0.28 | .839 |
| 2b | 0.17 | 0.03 | −0.01 | 0.02 | 1.46 | .236 |
| Social gains model | ||||||
| 1a | 0.12 | 0.02 | −0.01 | 0.02 | 0.69 | .555 |
| 2b | 0.13 | 0.02 | −0.02 | 0.01 | 0.16 | .855 |
| Social losses model | ||||||
| 1a | 0.20 | 0.04 | 0.02 | 0.04 | 1.95 | .124 |
| 2b | 0.25 | 0.06 | 0.03 | 0.02 | 1.60 | .206 |
| Variable | R | R2 | Adj. R2 | ΔR2 | ΔF | p Value |
|---|---|---|---|---|---|---|
| Monetary gains model | ||||||
| 1a | 0.18 | 0.03 | 0.01 | 0.03 | 1.44 | .253 |
| 2b | 0.23 | 0.06 | 0.01 | 0.02 | 1.69 | .189 |
| Monetary losses model | ||||||
| 1a | 0.08 | 0.01 | −0.02 | 0.01 | 0.28 | .839 |
| 2b | 0.17 | 0.03 | −0.01 | 0.02 | 1.46 | .236 |
| Social gains model | ||||||
| 1a | 0.12 | 0.02 | −0.01 | 0.02 | 0.69 | .555 |
| 2b | 0.13 | 0.02 | −0.02 | 0.01 | 0.16 | .855 |
| Social losses model | ||||||
| 1a | 0.20 | 0.04 | 0.02 | 0.04 | 1.95 | .124 |
| 2b | 0.25 | 0.06 | 0.03 | 0.02 | 1.60 | .206 |
Notes: Adj. R2 = adjusted R2; R2Δ = change in R2; ΔF = change in the F statistic.
aAge, education, and income were variables.
bLoneliness and isolation were variables.
*All values based on 1,000 bootstrap samples.
Discussion
This study extends the current literature by demonstrating that community-dwelling middle-to-older adults displayed steeper temporal discounting rates for gains as compared to losses, and for social outcomes as compared to monetary outcomes, with large effect sizes. Within this study, smaller immediate gains as opposed to larger delayed gains were chosen more often within both social and financial domains. Furthermore, there was evidence of steeper discounting within the social as compared to the financial domain without evidence of any interaction. The latter finding is somewhat contrary to expectations as the more emotionally salient future social outcomes were not valued more highly when compared to financial outcomes in middle-to-older adults.
Temporal Discounting Rates as a Function of Framing and Domain
Findings revealed that gains were more steeply discounted than losses within both monetary and social paradigms. The observed gain-loss asymmetry is consistent with the robust finding in prior research indicating steeper discounting rates for gains as opposed to losses, known as the “sign effect” (Frederick et al., 2002). Although this effect has been widely verified in monetary paradigms, emerging research suggests that this phenomenon exists for discounting time (Abdellaoui et al., 2018). There is also evidence that older adults show a greater preference for immediate social rewards as compared to health outcomes than young adults (Seaman et al., 2016). This study extends these findings by showing that middle-to-older adults demonstrated steeper delay discounting for social outcomes than monetary outcomes. These findings highlight a greater preference for immediate social outcomes as compared to financial outcomes.
Importantly, socioemotional selectivity theory and the positivity effect suggest individuals as they get older place greater value on immediate positive experiences compared to delayed positive experiences in order to prioritize achieving emotional gratification (Carstensen & Mikels, 2005; Carstensen et al., 1999). These theories further propose that as individuals perceive their time as more limited, they prioritize emotionally salient outcomes and relationships, focusing more on optimizing emotional experiences and well-being in the present moment. These theories would suggest as people age, they would show more temporal discounting for socioemotional outcomes. Additionally, perceptions of limited time have been posited to underlie older adults’ preferences for immediate gains due to concern that time will run out before the delayed outcome is obtained. Here, however, it is important to acknowledge patterns within the greater literature that show similar findings in young adults, including meta-analytic evidence that young-to-older adults do not significantly differ in their preference for smaller immediate as compared to larger later rewards (Seaman et al., 2022).
An alternative but not mutually exclusive explanation for the well-observed sign effect is that motivationally the affective experience of receiving immediate rewards is more reinforcing from a neurobiological standpoint and that differences in reward types affect the valuation of outcomes. This conceptual framework provides a rationale for the findings of steeper delayed discounting observed across a wide age range and emerging differences noted in reward devaluation (i.e., emotionally salient rewards resulting in a greater preference for immediate gratification) within the literature. It is also interesting to speculate on these findings from a brain-behavior perspective as the functional connectivity between reward areas of the brain and other brain regions has been shown to differ as a function of domain (Wake & Izuma, 2017). Furthermore, cross-sectional differences in neuroanatomical structure and age-related changes in white matter integrity can also affect reward valuation that results in steeper discounting for gains as compared to loss with age (Dhingra et al., 2020; Han et al., 2018). As most of the research to date has been cross-sectional, longitudinal studies are needed to investigate whether the desire to choose more immediate outcomes is motivated by the affective salience of outcomes and/or related to greater uncertainty about the future, which from a developmental perspective also affects younger adults.
Notably, within certain contexts, the preference for immediate gratification can lead to poorer financial (e.g., spending instead of saving/earning money) and social (e.g., prioritizing immediacy over quality time or social time over health concerns) decisions. As such, public health policy that considers consequences of immediacy that may result in worse future outcomes, framed in a way that underscores gain with potential consequences, could be a useful way to aid in optimizing decision choices. For instance, in the context of monetary decision making, programs that offer regular, continuous payout options may offset choosing less financially advantageous options. Regarding the desire for immediate social gains, the provision of multiple choices (e.g., regular phone calls and video chats, book clubs, going for walks, volunteer work, and joining groups of interests, such as a gardening club) may provide a stable reinforcement schedule and be done in a manner that helps maintain social distance when applicable. Finally, it has been noted that those with higher future self-continuity display reduced discounting of future rewards on decision-making tasks (Ersner-Hershfield et al., 2009). These findings suggest that helping individuals maintain a stable self-identity despite life changes and role transitions may lead to improved decision-making processes.
Decision Making in the Context of COVID-19
The striking differences in discounting rates between monetary and social outcomes should also be considered within the context of COVID-19. Contextual decision making has become more stressful since the COVID-19 pandemic, as individuals contend with high inflation and personal and public safety risks in social situations. Although “shelter-in-place” and social distancing procedures attenuated disease transmission, these policies also increased the risk of social isolation and loneliness among individuals, particularly older adults. Namely, adults aged 65 and older had a higher risk of contracting a COVID-19 infection as compared to other age groups, due to a higher likelihood of preexisting health conditions and weaker immune systems (Wu, 2020). Furthermore, social distancing procedures were particularly consequential within this cohort given the common reliance on family and community members (Hwang et al., 2020; Wu, 2020). Thus, it may be that social restrictions executed during quarantine and shelter-in-place protocols disproportionally affected older adults’ desire for immediate social outcomes. In this respect, the findings may partially reflect the perceived tradeoff between the frequency of social interactions with both personal and public safety, which became an important aspect of social decision making during the pandemic.
Socioaffective Factors
Our results suggested socioeconomic and socioaffective characteristics were not associated with temporal discounting behavior across conditions. These findings were unexpected given prior findings that lower income may be associated with steeper temporal discounting (for review, see Fiorenzato & Cona, 2022; Haushofer & Fehr, 2014). The hypothetical nature of our study may have contributed to our null findings. Specifically, hypothetical rewards may not have been viewed as motivating or enticing as real rewards or hypothetical losses may not have been felt as aversively as real losses. For example, in our monetary loss paradigm, it would be unlikely for someone to willingly choose a larger amount to pay back. However, in a real-life situation, such as making a credit card payment now or later, decision-making behavior might look drastically different due to practical limitations. In addition, our null socioaffective findings likely reflect the relatively low levels of loneliness and social isolation within the current sample. Therefore, future research may consider more thoroughly assessing loneliness and isolation in one’s life as it relates to recent losses (e.g., death of family members or friends, health ailments that limit social interaction, geographical barriers, etc.) given their theoretical links to decision making.
Strengths and Limitations
This study provides an important first step towards better understanding differences in social as compared to monetary decision making across gain and loss paradigms in middle-to-older aged adults; however, it is not without limitations. First, while our analyses and use of standardized values provide a psychometrically justified method for comparing different variables (Lange, 2011, pp. 2368–2371), it is important to note that these condition comparisons assume some form of equivalence between comparator conditions. Like most research that utilizes comparative conditions (for review, see Odum et al., 2020) this assumption is statistically driven and thus it cannot truly be ascertained that social and monetary conditions are equally weighted. This study also used an abbreviated decision-making task to lessen fatigue, increase task engagement, and reduce participant burden. Restricting the delay periods to four levels may have meant that we did not observe the same normative delay effect or steep discounting arc seen in other studies. This may have led to our relatively high AUC values. Although this is possible, prior research has demonstrated abbreviated paradigms can still produce valid and sensitive discounting indices (Yi et al., 2010) and that shortened testing often increases the validity of findings (Kost & de Rosa, 2018). Additionally, it is important to note that this study does not provide a direct comparison of age-related differences in decision-making behavior as younger adults were not investigated within this study. Finally, our assessment of sociability factors was limited to select questions. Although evidence supports the clinical utility of single-item screener use for the detection of loneliness in older adults (Kotwal et al., 2022), future research may wish to explore these relationships more comprehensively in a sample population adequately powered to evaluate interactions amongst these variables.
Summary
Results from this study yielded large effect sizes for decision-making preferences indicating that middle-to-older adults discounted gains more strongly than losses, and discounted social outcomes more steeply than monetary outcomes. These results may be informative from a public policy standpoint, with an eye toward framing messages in a way that details consequences of immediate as compared to delayed gains in financial decision making and providing opportunities for more frequent social incentives.
Funding
This work was supported by University of Maine start-up funds provided to the Principal Investigator (R.K.M.).
Conflict of Interest
None.
Data Availability
We report how we determined our sample size, and describe all data exclusions, manipulations, and all measures in the study in accordance with APA journal article reporting standards. Given the nature of this research and Institutional Review Board ethical considerations, the data cannot be made publicly available because participants did not give permission to publicly share their data. The data and analytic code will be made available upon request by contacting the first author directly ([email protected]). The study design, hypotheses, and analytic plan were not preregistered.
Statement of Ethics
This study protocol was reviewed and approved by the University of Maine Institutional Review Board, approval number 2018-09-01.
