Delay Discounting and the Income-Food Insecurity-Obesity Paradox in Mothers

Food insecurity, defined as unpredictable access to food that may not meet a person's nutritional needs, is paradoxically associated with higher BMI (kg/m2) and obesity. Research has shown delay discounting, a behavioral economic measure of the preference for immediate rather than delayed rewards, is related to higher BMI, and moderates the relationship between income and food insecurity. Based on this research, we used regression models to test whether delay discounting, consideration of future consequences, and perceived stress were atemporal mediators of the food insecurity-BMI relation in 313 mothers, controlling for demographic variables. A secondary aim was to replicate the finding that delay discounting moderates the relationship between low income and high food insecurity. Results showed that low income was associated with higher food insecurity, and higher food insecurity was associated with higher BMI. Delay discounting was the only variable that was indirectly related to both paths of the food-insecurity-BMI relation. Delay discounting accounted for 22.2% of the variance in the low-income-food insecurity-obesity relation, and the total model accounted for 38.0% of the variance. The relation between low income and food insecurity was moderated by delay discounting. These data suggest that delay discounting is a potential mediator of the relationship between food insecurity and high BMI, which suggests reducing discounting in the future could be a novel target to reduce food insecurity and help people with food insecurity to reduce their excess body weight. Trial Registration. This trial is registered with NCT02873715.


Introduction
Food insecurity due to lack of regular access to food that meets their nutritional needs [1,2] afects many lowincome families [3].Data from the Panel Study of Income Dynamics suggest that the percentage of families with children who experience food insecurity has increased from 8.8% in 1990-2003 to 10.9% in 2015-2019.Data from this study suggest that the experience of food insecurity may change over time, as only 33.1% of low-income families report high food security at both time points, suggesting there are protective factors that reduce the likelihood that a low-income family consistently experiences food insecurity [3].
Tere is a strong relation between food insecurity and elevated body mass index (BMI � kg/m 2 ) and obesity (BMI ≥30), which has led to the food insecurity-obesity paradox [4,5].Given that people with food insecurity experience unpredictable access to food, may lack access to healthy food in their environment, and may miss meals and go hungry [4,6,7], it is not expected that they would have excess body weight and obesity.Tis unexpected relation is the basis for the food-insecurity obesity paradox.Te relation between food insecurity and excess body weight is stronger for women than men [4,8].Many children in low-income families also experience food insecurity [3], and the study of mothers may be particularly interesting as they may be more sensitive to the efects of food insecurity on body weight [4,8], and they may model behaviors related to weight gain that can infuence eating behaviors of their children [9].
Shah and colleagues [10] theorize that resource scarcity that is characteristic of food insecurity creates a mindset in which individuals allocate attention to behaviors that reinforce poverty (e.g., borrowing money) while neglecting others (e.g., failing to enroll in assistance programs).Economic scarcity can lead to a person focusing on immediate needs, including putting food on the table that night or for the week, rather than considering longer-term goals such as saving to pay of debts, purchasing a home, or saving for children's college [11][12][13].Low income is related to higher discounting of the future [14,15], and a narrow temporal window focuses on immediate demands and immediate gratifcation as an adaptive response to pressing needs and tending to trade-ofs that must be considered in meeting those needs [16,17].Delay discounting (DD), which measures a person's temporal orientation towards either small, but immediate rewards, or large, delayed rewards [18,19], is associated with income [14,15], obesity [20][21][22], and food insecurity [11][12][13]23].Given that DD is associated with low income [14,15], high food insecurity [11][12][13]23], obesity [20][21][22], and increases during scarcity conditions [11-15, 23, 24], DD may provide insight into the relation between low income and food insecurity as well as the food insecurity-obesity paradox [4,5].Tis suggests that having a wide temporal window and not discounting the future may be a protective factor for people with low income not experiencing food insecurity [11].
While DD assesses a person's style of decision-making that tends to focus on smaller, immediate rewards versus larger, later rewards, there are other approaches to studying how people make decisions that focus on the balance between immediate and long-term needs.Tis is the basis for prospective thinking or a person's temporal orientation towards the future rather than the present.A widely used measure of temporal orientation is the Consideration of Future Consequences Scale (CFCS) [25][26][27], which assesses how much people consider future consequences of their behavior when making a decision about present behavior.It is relevant to consider whether another measure of a person's temporal orientation could help explain the relationship between food insecurity and BMI.
People with low income and food insecurity often experience stress [5,23,28].Stress is associated with obesity [23,29], and it can increase food reward [30] and can lead to emotional eating to regulate negative afect [31], both of which can lead to positive energy balance, weight gain, and obesity.Stress also impacts cognitive function, including reducing working memory capacity [32], increasing attentional bias for food [33], and increasing DD [23], which can shift decision-making towards immediate gratifcation and characteristics of people who experience food insecurity [11][12][13]23].
Te primary goal of this study is to assess whether DD, CFCS, and stress can help understand the relationship between food insecurity and higher BMI in mothers.A secondary goal is replicating the moderating efect of DD on the low-income-high food insecurity relationship that we observed in a previous study in which DD and fnancial planning moderated the relationship between low income and food insecurity [11].In that study, people with low income who engaged in some fnancial planning and did not discount the future showed levels of food insecurity equivalent to people with higher income levels.People with low income who engaged in fnancial planning with a narrow temporal window, as defned by high levels of DD, experienced high levels of food insecurity, equivalent to not engaging in fnancial planning.Tese data suggest that not discounting the future for people with low income was an important style of decision-making that is associated with lower food insecurity.

Methods
2.1.Subjects and Design.Tis study was approved by the University at Bufalo Institutional Review Board and was conducted online in November 2018.Te study was a secondary analysis of data collected for a cross-sectional study of parents recruited from ResearchMatch.org, a crowdsourcing platform.Te parent study was designed to study the relationship between stress and parenting.No publications were generated from the analysis of these relations.A recruitment message was sent to randomly selected individuals who met our study target (e.g., adults 18-55).Email addresses of those interested in participating were exported from ResearchMatch.org, and potential participants were sent an e-mail with the link to a Qualtrics questionnaire.An IRB-approved consent was presented to the participant upon opening the link, and consent was obtained by the participant indicating (e.g., clicking) "I agree to the terms described above."569 people consented and were eligible to participate (e.g., an adult 18-55 with at least one child between 2 and 15), and of those, 369 participants provided full, complete, and valid responses to all of the study questionnaires, including the DD task [34], and 313 of those were mothers, the focus of this analysis.

Measures.
Participants completed the MacArthur network socio-demographic questionnaire (e.g., age, gender, marital status, employment status, education, income, race, and perceived social status) [35] as well as the short form, 6-item food insecurity questionnaire [36] that asks questions regarding experience of food insecurity in the last 12 months.Cutofs for low food security and very low food security are 2 and 5, respectively.Parents self-reported their height and weight, which was used to calculate body mass index (BMI � kg/m 2 ).Overweight and obesity are defned as BMI ≥25 and BMI >30, respectively [35].Previous research has validated the use of height and weight self-reports for observational studies [37].Income was converted to percent over poverty based on household size using 2018 poverty data [38].
Participants completed an adjusting amount DD task [39].During this task, participants chose between receiving a larger amount of money at a future time point (1 day, 2 Journal of Obesity 7 days, 30 days, 182 days, and 365 days) and a smaller amount of money now (e.g., "Would you rather have $50 now or $100 in one week?).Te amount of money ofered "now" adjusted based on participant's prior response.Delayed time choices were presented in a random order.Depending on the participant's response, the immediately available amount was adjusted up (if delayed choice selected) or down (if immediate choice selected).Each time delay included fve trials with the amount adjusting half that of the previous adjustment [39].Patterns of responding were checked to see if systematic responding was observed using two rules: (1) if the subsequent indiference point was 20% less than the preceding indiference point of the larger, later reward and (2) if the last indiference point was not less than the frst indiference point by at least a magnitude equal to 10% of the larger, later reward [34].
For each participant, indiference points, or the amount of money ofered now that was just as appealing to them as $100 at a future time point, were calculated for each delay.Te ordinal area under the curve (AUC ord ) was used as the measure of DD.AUC ord ranges from 0 (maximum discounting of the delayed reward) to 1 (no discounting); higher values of AUC ord indicate lower levels of discounting.AUC ord was calculated using the ordinal values for each of the fve delays.Tis measure was chosen over the traditional AUC measure because it does not overweight the contribution of distal delays [40].
In addition, in a sensitivity analysis, we calculated discount rates using a hyperbolic discounting model [41], where V is the discounted value, A is the reward amount, D is the delay, and k is a free parameter that indexes the rate of discounting.Higher values of k indicate more rapid devaluation of the delayed reward and greater impulsivity.
Te CFCS is a 12-item scale that measures how much a person considers the long-term consequences of their current behavior [25][26][27].Research has shown signifcant relations between DD and CFCS [42].Te consideration of future consequences is related to obesity [43,44], such that the more you consider the future, the lower your BMI, but we could not identify any research on consideration of future consequences and food insecurity.CFCS asks how much one considers immediate (e.g., "I only act to satisfy immediate concerns, fguring the future will take care of itself") and distal (e.g., I am willing to sacrifce my immediate happiness or well-being in order to achieve future outcomes") outcomes [27].In the present study, immediate items [4][5][6][9][10][11][12] were reverse scored and averaged with the future items [1-3, 7, 8] for a total possible score between 1 and 5. Sample scores on the CFCS ranged from 2 to 5 (M = 3.8, SD = 0.61).Both the bi-and unidimensional scales are psychometrically valid and reliable [26], with a Cronbach's alpha of 0.82 [25].Recent research has found the CFCS future factor to be redundant [26] and that the unidimensional model is more parsimonious than two discrete factors [25].
Te Perceived Stress Scale (PSS) is a 10-item survey with questions that indicate how stressed an individual perceives themselves to be and how well equipped they are to deal with that stress (e.g., "In the last month, how often have you felt nervous and "stressed"?", "In the last month, how often have you felt that you were unable to control the important things in your life?") [45].Answers range from "0" (never) to "4" (very often), positively stated items (e.g., 4, 5, 7, 8) are reverse-scored, and responses are summed.In the present study, PSS scores ranged from 0.00 to 38.00 (M � 16.23, SD � 7.41).Te PSS has good test-retest reliability and criterion validity for measuring psychological stress, with Cronbach's alpha greater than 0.70 [46].

Analytic Plan.
Preliminary analyses used zero-order correlations to explore relations among variables.To understand the income-food insecurity relation, analysis of variance (ANOVA) was used to explore the efect of income expressed as a three-category classifcation of low income (N � 53; <200% of poverty), middle income (N � 117; 200-399% over poverty), and higher income (N � 153; >400% over poverty) using 2018 standards for poverty [38] and to explore the interaction between income and DD on food insecurity.Te main analytic approach was to use atemporal mediation for cross-sectional data to understand whether DD, CFCS, or perceived stress could be pathways between food insecurity and BMI, controlling for age, years of education, and percent above poverty.Atemporal mediation uses the analytic approach developed to identify mediators of change, but these methods can be used in cross-sectional research to identify relations among variables and to get a signal and potential true mediators that can be tested in future prospective research [47].Te working model that was tested is shown in Figure 1, with DD, CFCS, and stress tested for their indirect infuence on food insecurity and BMI.Te indirect efects and confdence intervals of DD or CFCS mediating the efect of food insecurity on BMI were frst estimated in a separate analysis using Hayes process macro 4.0 [48] which constructed estimates from 10,000 bootstrap resamples.Te indirect efects are presented in three ways.First, the indirect efect is signifcant if the bootstrapped 95% confdence interval does not contain zero.Second, the magnitude of the indirect efect refects the estimated change in the dependent variable through the efects of the mediator variable per unit change of the predictor variable [49].Tird, the proportion of the total efect explained by the indirect efect will be quantifed by the efect ratio (indirect efect divided by the total efect) [50,51].In addition, we assessed the distribution of food insecurity ratings and the percentage of food insecurity and high food insecurity as a function of low, middle, or high-income groups.Tis was accomplished using ANOVA for the food insecurity ratings and chi-square for the percentages.We also present the intercorrelations among the variables.Data were analyzed using SYSTAT 11 [52], SAS 9.4 [53], and the Hayes process macro 4.0 [48].

Results
Mothers had an average food security score of 1.02 (2.00) (mean ± SD), with 21% reporting food insecurity and 12.9% reporting high food insecurity.Tese numbers are greater than the 11.8% of the U.S. population reporting low food security and 4.3% of the U.S. population reporting very low food security in 2018 [54].Te percent of families receiving government assistance was 14.8%, almost all who received Supplemental Nutrition Assistance Program (SNAP) benefts (43/46), which is similar to the 11.3% of families in the United States who receive SNAP benefts [55].Te average mother had a BMI of 28.0 (7.2), with 57.8% overweight or obese (BMI ≥25) and 27.4% obese (BMI ≥30), compared to population averages in 2018 of 73.1% overweight or obese and 42.5% obese [56].Te average mother was married (77.3%), 32.7 (7.1) years of age, with 1.9 (0.9) children (see Table 1).
Te indirect efects of food insecurity on BMI through DD, CFCS, and perceived stress are shown in Table 3 and Figure 1.Te only variable that was related to both food insecurity (p < 0.01) and BMI (p < 0.001) was DD.When parallel paths were included, CFCS was related to BMI, and stress was related to food insecurity (p's ≤ 0.001), but CFCS was not related to food insecurity (p � 0.09), and stress was not related to BMI (p � 0.75).Te estimated indirect efect of DD on the relation between food insecurity and BMI was 0.133; 95% CI 0.206 to 0.287, the indirect efect of CFCS was 0.075, 95% CI −0.026 to 0.054, and the indirect efect of stress was 0.019, 95% CI −0.106 to 0.067.Te efect ratios for the parallel efects of DD, CFCS, and stress were 22.2%, 12.5%, and 3.2% of the total observed efect of food insecurity-BMI relation on BMI, respectively (see Table 3), which accounted for 38.0 percent of the variance in the food insecurity.Te pattern of results did not difer in a sensitivity analysis using log k as the DD measure.Te Indirect effect (food insecurity (x) effect on BMI (y) via Delay Discounting (mediator M1) = 0.0133 % of total effect due to M1 = 22.2%; CFC (M2) and PSS (M3) did not mediate x effect on y.

Discussion
We found that people who discounted the future more had lower income, greater food insecurity, and higher BMI, consistent with previous research [20,21].Te atemporal mediation model showed that DD was a potential mediator of the food insecurity and BMI relation when considered in parallel with a more general measure of prospective thinking and with perceived stress.Delay discounting has been reliably related to food insecurity [11][12][13]23] and to obesity [20][21][22].Having low income may shape people to allocate resources to solve immediate problems, such as paying rent, utilities, and getting food on the table.Saving and thinking about the future are luxuries that many low-income people do not experience.Paying attention to immediate needs and not focusing on an uncertain future are adaptive responses to economic scarcity [16,17].Disregarding immediate needs could be catastrophic for someone in a lower income bracket.In this regard, discounting the future should be considered a learned, adaptive response to their current environmental context, and not an inappropriate Journal of Obesity response for people with low income.In fact, research suggests that food insecurity is related to cognitive fexibility [57].Te narrow temporal window observed for people with food insecurity may be due in part to the unpredictable nature of access to food and meals.Basic preclinical research in diferent animal species has shown that animals that are provided unpredictable access to food gain more weight and fnd food to be more reinforcing than animals that are provided the same amount of food, but with predictable access [58,59].Unpredictable and irregular access to food is reliably related to food insecurity [4,7,59], and people with lower income may have more variability in access to food and the types of food consumed, and a strong desire to consume that food when it is available, shaping people to seek the immediate gratifcation associated with eating food when hungry [4,7,59].Also, low income is related to a stronger reinforcing value of food [60], and unpredictable access to food makes food more reinforcing [58,59].Reinforcer pathology theory [61][62][63] integrates high food reinforcement and a short temporal window associated with immediate gratifcation that may help explain some of the psychological processes behind the food insecurity-obesity paradox.
Te current study suggests that DD, a behavioral task designed to measure the relative strengths of immediate versus delay rewards, is more strongly related to the foodinsecurity BMI relation than a more general measure of prospective thinking.We think this may be the case for two reasons.First, CFCS is a self-report measure of how people report they think about the future, as opposed to a direct behavioral measurement of this construct, and while these constructs are related, the correlation is modest.Second, the CFCS focuses on how someone might act to improve their future, but it does not compare the value of immediate versus future decisions, which is central to DD.It may be quite diferent to think about what you would do in the future if you have access to a very powerful, immediate reinforcer.Tird, CFCS focuses on decision-making in regard to general future consequences, while the DD task involves monetary decision-making.
While stress is known to be associated with food insecurity [5,23,28], stress was not shown to be indirectly associated with the relation between food insecurity and BMI.Tis may be due to several reasons.Te analytic model we used compared the parallel infuence of DD, CFCS, and stress as indirect infuence on food insecurity and BMI.DD may have a stronger indirect infuence when stress is high.It is also possible that the infuence of food insecurity on BMI is mediated by a pathway that includes stress-modifying DD which then modifes BMI, as research has shown that stress modifes decision-making [23,32,33].A moderated mediator model could test if the mediating efect of discounting on BMI depends on how much stress someone is experiencing.For these reasons, we think future studies should continue to measure stress in addition to DD as part of more complex and prospective models to understand the relationship between food insecurity and BMI.
Tese data suggest that DD may be a novel target to reduce obesity in those with food insecurity.DD is associated with a short temporal window, and the desire for immediate gratifcation, rather than taking into account possibilities associated with inhibiting the immediate desires for food [18,19].Interventions to modify DD, such as episodic future thinking [64] or fnancial planning [65], aim to teach people to use prospective thinking to consider how immediate behaviors infuence long-term, rather than immediate, goals.While eating a favorite food now may be very satisfying, it will not move you towards the goal of losing weight or improving your health.
As a secondary aim, we replicated the fnding that income is related to food insecurity and that DD moderates the relation between low income and high food insecurity [11].Participants with low income who did not discount the future had lower levels of food insecurity in comparison to low-income participants who did discount the future.Given that food insecurity is more common in people with low income, and low income is related to DD, future research should study a more complex model that relates income, food insecurity, DD, and BMI.Prospective research is needed to test how these variables are related.Journal of Obesity Journal of Obesity 4.1.Limitations.Despite the fact that the study used a large sample and replicates previous studies showing that DD is related to both food insecurity and obesity, and DD may help explain the food insecurity-obesity paradox, there are limitations to the study.First, the study only included women, as food insecurity has a greater impact on obesity in women than men [4,8], but research is needed on large samples of men with food insecurity.Te sample was relatively high income, and predominantly white, but the percentage of families with food insecurity was representative of the percentage of families in the United States who experience food insecurity.However, given that rates of food insecurity are greater in African American than white women [66], and the reason for this relationship is poorly understood [66], it is possible that a diferent pattern of results would be observed if a larger percentage of African American mothers was studied.
While the atemporal mediational model suggested an infuence of DD on the relation between food insecurity and BMI, the study was cross-sectional.We believe that cross-sectional studies can provide ideas for future prospective analyses of mediators, but prospective research testing whether DD mediates the relation between food insecurity and weight gain is needed.One hypothesis for a prospective study might be that low income is associated with food insecurity only when people have a narrow temporal window for decision-making and discount the future, and that discounting of the future also mediates the relation between food insecurity and weight gain.Since research has shown that low income is associated with greater food reinforcement, unpredictable access to food can increase the reinforcing value of food, including food reinforcement plus DD using reinforcer pathology theory in future research may provide greater explanatory power than DD alone.Another potential limitation is that only discounting of money was measured, and discounting of food may be a more sensitive measure for people with food insecurity, as shown by Rodriguez and colleagues [12].However, there may be a challenge in interpreting discounting for food when studying people with obesity who know they should consume less food.Asking them whether they want a small amount of food now or a larger amount later is difcult to interpret, as a person with obesity may know they should eat less now rather than more later, and respond as if they were seeking immediate gratifcation, rather than just eating less.Finally, BMI was based on self-reported height and weight, and while self-report has been shown to be a valid measure of BMI in a large sample of U.S. adults [37], people do tend to overreport their height and underreport their weight, and stronger food-insecurity obesity relations may have been observed if height and weight were measured objectively.BMI also has limitations in measuring excess body fat [67], which may be particularly relevant for people with food insecurity who demonstrate substrate oxidation patterns consistent with relative carbohydrate oxidation and storage of fat [68].

Conclusions
Steeper discounting of the future is associated with greater food insecurity and higher BMI, and DD adds to our understanding of factors that infuence the relationship between food insecurity and body weight.

Figure 1 :
Figure 1: Display of the parallel infuences of delay discounting, consideration of future consequences, and perceived stress on the food insecurity-BMI relation.

Figure 2 :
Figure 2: Food insecurity ratings (Mean ± SEM) (a), percent of participants with food insecurity (≥2; (b)), and percent of people with high food insecurity (≥5; (c)) in relation to income expressed as percent over poverty index adjusted for household size.

Table 3 :
Models the indirect efect and efect ratios of delay discounting and food choice on the relation between food insecurity and BMI, controlling for education, age, stress, income, house size, and exercise.