Economic and social deprivation predicts impulsive choice in children

Impulsivity is an individual difference in decision-making that is a risk factor for a number of health concerns including addiction and obesity. Although impulsivity has a large heritable component, the health concerns associated with impulsivity are not uniformly distributed across society. For example, people from poorer backgrounds are more likely to be overweight, and be dependent on tobacco or alcohol. This suggests that the environmental component of impulsivity might be related to economic circumstances and the availability of resources. This paper provides evidence that children aged 4 to 12 from the most deprived areas in England show greater impulsivity in the form of delay discounting than do children from the least deprived areas. The data are discussed with reference to scarcity-based models of decision-making and to public health inequalities.

www.nature.com/scientificreports/ of gratification specifically refers to measures such as the Marshmallow Test developed by Walter Mischel. Delay discounting refers to psychophysical tests based on choices between monetary rewards that permit a discount parameter such as k to be derived. Some tests such as the one used in the English Longitudinal Study on Aging are based on choices between monetary rewards but do not permit k to be derived and are therefore referred to generically as a measure of time preference rather than discounting.
Delay discounting is sufficiently associated with impulsive behaviours that discount rates can be considered a measure individual differences in impulsivity 10 , and as a significant risk factor for developing health problems. For example, daily smokers have higher discount rates than either social or non-smokers 11 , problem gamblers have higher discount rates than non-gamblers 12 , people who are dependent on opiates have higher discount rates than people who don't use opiates 13 . Heavy drinkers discount the value of delayed rewards more steeply than light drinkers 14 , and people with larger Body Mass Indexes (BMI) have higher discount rates than people with healthy BMIs 15 . Discounting may even be a risk factor in the transition from recreational to problem gaming 16 , and is associated with symptoms of Conduct Disorder and ADHD 17 .
Individual differences in discounting emerge from 7 to 8 years old 9,18-20 and remains stable throughout life suggesting that there may be a large heritable component 17,21 . Indeed, estimates of heritability suggest that around 50% of the variance in delay discounting and other measure of impulsivity are attributable to genetic factors 17,[22][23][24][25] . This leaves around 50% of the variance attributable to environmental factors to be identified. Although it is already recognised that there are health inequalities across the UK relating to impulsive behaviour [26][27][28] , there is emerging evidence that the associated discount rates are not uniformly distributed across society and that these may be related to differences in economic circumstances. For example, Anokhin et al. 17 noted that discount rates in 14 year olds were associated with socio-economic status based on parental occupation. More recently, Tunney and James 29 reported that social-economic classification based on occupation predicted a measure of time preference similar to discounting in older adults. Why might a person's occupation, or their parent's occupation, predict how impulsive they are? Perhaps occupation is a proximal variable associated with a more fundamental driver of behavioural choice. One possibility is that a scarcity of resources or economic uncertainty leads the decision-maker to prefer immediate rewards when they become available, rather than waiting for larger and perhaps equally unpredictable rewards. In this respect, impulsive choice does not imply irrational choice. To test this hypothesis, Tunney and James 29 compared preferences for smaller-sooner rewards in older adults from areas in England that ranged from the most deprived to the most affluent using the English Index of Multiple Deprivation (IMD). This is the official measure of relative deprivation in England and ranks each small area in terms of income, employment, education, health, crime, housing, and environment. Tunney and James observed that people in the most deprived areas were more likely to prefer smaller sooner rewards than people from the least deprived areas, and people in technical or routine occupations tended to prefer smaller sooner rewards than people in professional or intermediate occupations. Of course, the direction of a causal connection between working in a poorly paid profession and living in a deprived area, and impulsive choice cannot be established in older adults. It could be that impulsive people drift into poorly paid positions and towns based on the choices that they make. The study that follows seeks to test the hypothesis that that relative deprivation is related to impulsivity by measuring delay discounting in children aged between 4 and 12 years, from a range of backgrounds from the relatively deprived to the relatively affluent. If confirmed this would lend support to the hypothesis that deprivation is a causal environmental influence on impulsivity.

Method
This study was conducted as part of the University of Nottingham Summer Scientist week in 2018. This is an annual event in which children from the Nottingham community take part in a wide range of psychological studies during the school holiday period.
Participants. Informed consent to take part in this research was obtained from the parents or legal guardians of each participant. One-hundred and fifty-six children took part in this study. Their average age was 8.395 years (SD = 2.067). The youngest was 4.15 years old, and the oldest was 12.13 years old. Sixty-nine were female and 87 were male. The parents or legal guardians completed demographic information including post-codes that were used to derive the Index of Multiple Deprivation for their address, and also about ethnicity and the languages spoken in the home. The majority (143) indicated that English was the only language spoken at home, 2 Arabic and English, 2 Arabic, 3 English and Urdu, 1 French and English, 1 Japanese and German, 2 Mandarin, 2 did not respond to this question. All of the children were students in the British education system and were proficient in English. Procedure. There were five-time preference questions from which the parameter k was derived. On each trial the participants were given a choice between £10 after an interval or an immediate outcome that titrated upwards from £0.50 to £9.50 in 50p increments. These choices were displayed on an smaller-sooner card (Fig. 1A) and a www.nature.com/scientificreports/ larger later card (Fig. 1B) that showed the value of the choice in coins. The intervals were "in one-week" (coded as 7-days), "in two weeks" (14-days), "when school starts (30-days)", "£10 at Christmas" (180-days), or £10 next summer" (365-days). For example, the child would first be asked "would you prefer £10 in one-week, or £0.50 now?" followed by "would you prefer £10 in one-week, or £1.00 now?". The larger later card was titrated from the largest to the smallest value. Each trial would stop when the participant switched their preference from the larger later outcome to the smaller sooner outcome. The participants completed the British Picture Vocabulary Scale III (BPVS-3) 32 . The BPVS-3 is used to test receptive vocabulary in children ages from 3 to 18 years that provides an age adjusted standard score with a mean of 100. They also completed the Autism Quotient Children's Version (AQ-Child) 33 . The minimum AQ-Child score (0) indicates no autistic traits; the maximum score (150) suggests full endorsement on all autistic items. Finally, participants completed the Strengths and Weaknesses of Attention Deficit Hyperactivity Disorder Symptoms and Normal Behaviour (SWAN) 34 . This measure has two subscales for Inattention and Hyperactivity/ Impulsivity. Each test was age appropriate and the participants confirmed that they were able to comprehend instructions for each of the tests.

Ethics statement.
Results. Figure 2 shows the average subjective discounted value for each delay period. A repeated measures ANOVA indicated a reliable main effect of delay (F 4,516 = 32.822, MS e = 1.592 p < 0.001, η 2 p = 0.203), and a reliable linear effect of delay (F 1,129 = 67.756, MS e = 2.970 p < 0.001, η 2 p = 0.344) indicating that the participants showed a robust discounting effect.
Individual non-linear regressions were performed for each participant to derive the discount function (k) using Mazur's (1987) hyperbolic discount function (1). Where the subjective discounted value (v) is estimated from the preferred smaller sooner outcome (v) for the five delay intervals (D). Because k is typically skewed it was transformed to their base-10 logarithms (log-k) for analyses. www.nature.com/scientificreports/ Table 1 shows the average discount rates (log-k), BPVS scores, AQ scores, and SWAN scores for males and females. There were reliable differences between males and females in AQ and SWAN Scores, but not in discount rates or BPVS.
The indices of multiple deprivation (IMD) were computed from the demographic information provided by parents using the publicly available resource provided by the Ministry of Housing, Communities and Local Government (http:// imd-by-postc ode. opend ataco mmuni ties. org/). This provides decile ranks from 1 = most deprived to 10 = least deprived. These were collapsed into quintiles (from 1 to 5). The average subjective discounted value for delay period for each IMD decile are shown in Table 2. Figure 3 shows the implied discount curves for each IMD quintile. A one-way ANOVA on the log-transformed discount rates with IMD quintile as a between-subject factor was significant (F 1,149 = 3.109, MS e = 1.929, p = 0.017, η 2 p = 0.080) indicating that there are reliable differences in impulsive choice made by children from different economic and social backgrounds. Planned contrasts showed that participants from the most deprived postcodes (IMD 1) show steeper discounting than participants in the least deprived postcodes (IMD 5, p < 0.003, and IMD 4, p = 0.007), but not the intermediate postcodes (IMD 3, p = 0.212; IMD 2, p = 0.161).
Finally, to test the hypothesis that relative deprivation is a casual factor in the development of impulsivity the IMD quintiles, age, gender, AQ score, SWANN score, and standardised BPVS-3 scores were entered into a linear regression as predictors of the log-transformed discount function. The model was significant (R 2 = 0.190, se = 1.131, F 6,135 = 5.037, MS e = 1.718, p < 0.001). Table 3 shows the regression coefficients. Only age and the Index of Multiple Deprivation were reliable predictors of discount rates. Because younger children tend to discount more than older children it is possible that the effect of deprivation might be due to differences in the distribution of ages across the IMD quintiles. The average age of participants in each quintile are shown in Table 3 and  www.nature.com/scientificreports/ suggest that this is not the case. A one-way ANOVA on age by quintile eliminated this possibility F 4,155 < 1.0). Deprivation is a reliable and independent predictor of discount rates in children.  Figure 3. Showing the implied discount curves for each index of multiple deprivation quintile from 1 = most deprived to 5 = least deprived.

Discussion
Relative deprivation predicts delay discounting in children aged 4 to 12 years. The data confirm a related report of a link between Indices of Multiple Deprivation and time preferences based on the English Longitudinal Study of Aging 29 . This adds to the growing evidence that economic uncertainty is an environmental factor in an individual difference in decision-making that is associated with a range of individual and social issues. Previous research has shown a relationship between impulsive behaviours such as gambling and both social status based on occupation and relative deprivation based on IMD 35 . For example, unhealthy diets resulting in high BMI and obesity in both children 36 and adults 37 . Both delay discounting 11,38 and relative deprivation 39 are key risk factors in the severity of tobacco dependence. Similarly, excessive alcohol consumption is also associated with relative deprivation 40 and delay discounting 14 . One previous study explored if time preferences play a mediating role between relative deprivation and smoking and BMI 41 . Time preference in that study was measured using a psychometric scale Consideration of Future Consequences Scale (CFCS) that is only partially related discount rates 42 . Although this study did not find a clear relationship between time preferences and smoking that suggests that it is not as a robust measure of time preferences as discount rates. Nonetheless there is considerable evidence that discount rates as a stable individual difference may be a risk factor in for impulsive behaviours and poor health. The evidence presented here indicates that at least some of the individual difference in the impulsivity is associated with relative deprivation.
The relationship between childhood obesity and relative deprivation is particularly interesting when impulsivity is viewed through the lens of ecological models of foraging. Obesity in nursery age children is associated with both relative deprivation and food insecurity 43 . Evidence from the Bradford Cohort study shows that families living with relative deprivation tend to consume fewer fruit and vegetables and more high sugar drinks than more affluent families 44 . The UK Millennium Cohort study of 7262 children aged 11, showed that two thirds of children living in the most deprived areas in England were overweight or obese, compared to one fifth in the least deprived areas 45 . The link between food insecurity and relative deprivation is thought to be food consumption cycling in which food, and particularly energy dense food, is overconsumed in periods of abundance in anticipation of periods of scarcity 46,47 . This mechanism is synonymous if not identical to the Thrifty Phenotype Hypothesis 48 and the Insurance Hypothesis 49 .
Early experience of relative deprivation, economic uncertainty, or food scarcity could result in a shift in choice parameters to generally prefer smaller sooner rather than larger later rewards. If this shift becomes stable, then it may result in overconsumption and unhealthy behaviours and a risk of developing dependencies and addictions throughout life 29 . Since 2010 the number of families in poverty in the United Kingdom has increased. Relative poverty in the UK has essentially unchanged in from 2007 to 2020 at around 22% of households. However, relative child poverty has continued to increase to around 700,000 50 . Between 2020 and 2021, the UKs largest network of food banks distributed 2.5 million emergency food parcels to people in crisis, a 33% increase the number of parcels and a 53% increase in the volume of parcels on the previous year 51 . Of these nearly 1 million went to children. Many of these were due to problems with the benefits system including delays, insufficient welfare payments and cuts to welfare payments. The Trussell Trust reported that in 2020 72% of families who received a food parcel had a family member with poor mental health. If it is the case that relative deprivation in childhood is causally related to impulsivity related health inequalities in adult life then current welfare policies may be creating future health inequalities and potentially a future public crisis in mental health.

Limitations
This study used hypothetical rather than real monetary rewards. Real monetary rewards provide more compelling findings than hypothetical rewards in any decision-making task. However, this is challenging in delay discounting where the rewards could be delayed for months or years, and in children for whom the perception of monetary value or time might differ from adults. A recent systematic review of delay discounting in children reported 25 studies, of which only 8 involved real rewards, of which only 5 were monetary rewards 18 . Nonetheless there is a general consensus that delay discounting, even with hypothetical rewards, produces meaningful and reliable results. The cross-sectional nature of this study, both in terms of age and demographics, precludes any firm conclusions being drawn on the causal relationship between deprivation and impulsivity.

Data availability
All data generated or analysed during this study are included in this published article and its supplementary information file (Supplementary Table S1). www.nature.com/scientificreports/