Article Text

Download PDFPDF

Student loans and racial disparities in self-reported sleep duration: evidence from a nationally representative sample of US young adults
  1. Katrina M Walsemann1,
  2. Jennifer A Ailshire2,
  3. Gilbert C Gee3
  1. 1Department of Health Promotion, Education, and Behavior, University of South Carolina, Arnold School of Public Health, Columbia, South Carolina, USA
  2. 2University of Southern California, Davis School of Gerontology, Los Angeles, California, USA
  3. 3Department of Community Health Sciences, University of California, Los Angeles, Fielding School of Public Health, Los Angeles, California, USA
  1. Correspondence to Dr Katrina M Walsemann, Department of Health Promotion, Education, and Behavior, University of South Carolina, Arnold School of Public Health, 915 Greene Street, Room 529, Columbia, SC 29208, USA; kwalsema{at}sc.edu

Abstract

Background Student loans are the second largest source of personal debt in the USA and may represent an important source of financial strain for many young adults. Little attention has been paid to whether debt is associated with sleep duration, an important health-promoting behaviour. We determine if student loans are associated with sleep duration. Since black young adults are more likely to have student debt and sleep less, we also consider whether this association varies by race.

Methods Data come from the US National Longitudinal Survey of Youth 1997. The main analytic sample includes 4714 respondents who were ever enrolled in college and who reported on sleep duration in 2010. Most respondents had completed their college education by 2010, when respondents were 25 to 31 years old. Multivariable linear regression models assessed the cross-sectional association between student loans accumulated over the course of college and sleep duration in 2010, as well as between student debt at age 25 and sleep duration in 2010.

Results Black young adults with greater amounts of student loans or more student debt reported shorter sleep duration, controlling for occupation, hours worked, household income, parental net worth, marital status, number of children in the household and other sociodemographic and health indicators. There was no association between student loans or debt with sleep for white or latino adults and other racial/ethnic groups.

Conclusions Student loans may contribute to racial inequities in sleep duration. Our findings also suggest that the student debt crisis may have important implications for individuals’ sleep, specifically and public health, more broadly.

  • SLEEP
  • INEQUALITIES
  • SOCIO-ECONOMIC

Statistics from Altmetric.com

Introduction

One of the largest sources of personal debt in the USA originates from student loans, amounting to over $1 trillion dollars.1 The USA has some of the least favourable terms for repayment or forgiveness of student loan debt (eg, shorter grace-period for repayment, stringent default rules and few options for loan forgiveness).2 As such, student debt may represent an important financial stressor for borrowers. Debt has been linked to greater stress and poorer mental health,3–6 but there has been little attention to how debt affects sleep, an important health-promoting behaviour.7–9

Financial strain and the stress, worry and anxiety that accompany it can disrupt sleep, which may result in shorter sleep duration.10–12 Prior research has found that individuals experiencing persistent financial strain and economic difficulties are more likely to lie awake at night, either due to difficulty falling asleep or because they wake in the night and cannot fall back to sleep, and thus, have shorter sleep.10 ,11 ,13 Although these studies demonstrate a link between financial strain and disrupted sleep, it is still unknown if debt is directly associated with sleep, in general or total sleep duration, specifically. We know, however, that psychosocial stressors strongly affect sleep,13–17 and that student debt may be an important psychosocial stressor that could ultimately impact the amount of sleep individuals experience.

Since student loans may be a major source of financial stress for many young adults,18 ,19 we hypothesise that borrowing greater amounts of student loans will be associated with shorter sleep duration. We further investigate differences by race/ethnicity. Compared to white individuals, black individuals report sleeping roughly 30–45 min less per night20 and are twice as likely to report sleeping five or fewer hours per night.21 Previous research finds that black individuals have a greater burden of daily hassles and stressors, including more financial strain, which can contribute to their shorter sleep duration.13 ,17 Further, black individuals generally have smaller economic22 ,23 and health returns23 ,24 to education compared with white individuals. For example, in 2008–2010, black male college graduates earned 28% less than their white counterparts.22 Black graduates may, therefore, experience greater financial strain as a result of student debt compared with white graduates, even at equivalent levels of student debt. Thus, we also hypothesise that the negative association between student loans and sleep duration will be stronger among black graduates than white graduates.

Methods

Data

We analysed data from the National Longitudinal Survey of Youth 1997 (NLSY97), a nationally representative sample of 8984 individuals born between 1980 and 1984.25 Baseline interviews were conducted in 1997, with annual follow-up interviews ongoing. Our analyses include data through 2010, the year when sleep was measured. The retention rate over the 13-year period was 80%. This study was considered exempt by the Institutional Review Board at the University of South Carolina and conforms to the principles of the Declaration of Helsinki.

We constructed two analytical samples to correspond with our independent variables of student borrowing and student debt. Our main analytic sample (hereafter referred to as the full sample) consisted of any respondent who had ever enrolled in college by 2010 (n=5546). We excluded an additional 832 respondents who were not interviewed in 2010 or had missing data on sleep, resulting in a final analytic sample of 4714. Supplemental analysis indicated that non-response was more common among older respondents, immigrants and those raised in the wealthiest families and was less common among black individuals, hispanic individuals and transfer students.

Our second analytic sample (hereafter referred to as the subsample) consisted of any respondent who had ever enrolled in college by the year they reported on student debt (at or around age 25), resulting in an analytical sample of 4459. Although respondents were asked about student debt when they were age 25 and 30, only 18% of respondents turned 30 by 2010. Thus, we only include information about debt at age 25.

Measures

Dependent variable: sleep duration

Respondents were asked “On a typical weeknight, how many hours of sleep do you usually get?” Responses ranged from 0 to 24 h and were reported in 1 h increments. This yielded a normally distributed, continuous variable.

Independent variables: student loans and student debt

Student loans were measured by asking students how much they borrowed in government subsidised loans or other types of loans (excluding loans from relatives and friends) for each and every semester between 1997 and 2010. We then summed these amounts to arrive at the total amount of student loans borrowed during this period. We top-coded student loans at the 95th centile.

Student debt was measured by asking how much respondents owed on student loans at age 25. We top-coded student debt at the 95th centile. We also included a quadratic indicator of student debt, since the association between student debt and sleep duration was non-linear.

Covariates

We included demographic and socioeconomic indicators that might be associated with student loans/debt and sleep duration. All covariates were assessed in 2010 unless otherwise noted.

Enrollment history refers to the type of colleges respondents attended and was categorised as: (1) attended a 2-year college only; (2) attended both a 2-year and 4-year college; or (3) attended a 4-year college only. The latter two categories also include those who attended graduate school.

Last year enrolled in college uses enrolment data to adjust for variation in time since last year of enrolment. Values ranged from 1998 to 2010. For the subsample, this measure was created based on the year the respondent reported on student debt.

Educational attainment was measured using the highest degree attained, which was categorised as less than a Bachelor's degree versus a Bachelor's degree or higher. More refined categories of degree attainment yielded similar results.

Number of hours worked, on average, in the prior year was categorised as 0 h, 1–39 h, and 40 or more hours. Total number of jobs the respondent held was measured as a count from 0 to 6.

Occupational status was assessed based on the respondents’ current or most recent job using the 2002 Census Occupational Codes, and categorised as (1) professional or managerial; (2) sales, service or technician; (3) labour, production or farmer; (4) active military; and (5) not working.

Self-reported household income was categorised as 0–49th centile, 50th–74th centile and 75th centile and over.

Other financial debt (ie, credit cards, car loans, debt owed to businesses or to relatives) was self-reported by respondents. The variable excludes student debt. Inclusion of both a dichotomous indicator of any debt and a continuous measure of total debt in regression models showed that having any debt was related to sleep duration, but the amount of debt was not. Thus, we only include the dichotomous indicator in the analysis we present.

Marital status was categorised as (1) never married; (2) married; or (3) divorced, separated or widowed. Number of children under age 18 living in the household was included as a continuous indicator ranging from 0 to 9. We also include region of the USA (south, north, west or outside the USA), urbanicity (urban, rural or undetermined), self-reported race/ethnicity (non-Hispanic White, non-Hispanic Black, Latino and other), nativity (born a US citizen or not), gender, age and pregnancy status (pregnant vs not). For the subsample, regression models also include an indicator for the year they answered questions about student debt.

To partially control for selection into college, parents’ net worth reported by respondents’ parents in 1997 was calculated as the amount of debt subtracted from assets (home equity, bonds/CDs/stocks, life insurance policies, pension and retirement savings, car values, etc) and categorised as negative net worth (<$0), low net worth (∼5th to <25th centile), mid net worth (≥25th to <75th centile) and top net worth (≥75th centile).

To adjust for differences in adolescent health, we include an indicator of self-rated health in 1997 measured as 4=excellent, 3=very good, 2=good and 1=fair/poor.

Statistical analysis

We employed multivariable linear regression to examine the association between student loans/debt and sleep duration. Following recommendations by the NLS, we estimated weighted descriptive statistics and unweighted regression coefficients;25 however, weighted regression coefficient estimates were similar to those we present. To make interpretation of the regression coefficients for student loans and student debt more meaningful in our multivariable analyses, we divided by 5000. Thus, a one-unit change in student loans or student debt represents a change of $5000. For ease of presentation, sleep duration is presented in minutes in our multivariable regression models so that regression coefficients represent the number of minutes of sleep lost or gained.

We imputed data using the mi impute command with chained equations specification in Stata13. This process produced 25 data sets. All analyses were replicated across the 25 data sets and combined based on Rubin's rules26 using mi estimate in Stata13.

Results

Table 1 presents sample characteristics separately for the full sample and subsample. Reported characteristics were generally similar across the two samples. Half of the respondents had a college loan. The average loan amount was $10 176. At age 25, 36.2% reported student debt, with an average student debt amount of $17 765. In 2010, respondents slept an average of 6.8 h. Approximately 70% of the sample was white, 52.8% were female, and 95.7% were born a US citizen. Most respondents attended a 4-year college at some point in their college career (69.3%), and 44.3% attained at least a Bachelor's degree. Additional sample characteristics are reported in table 1.

Table 1

Characteristics of NLSY97 respondents for full sample (n=4714) and subsample (n=4459), weighted estimates

Table 2 shows differences in sleep duration, student loans and student debt by race/ethnicity. White respondents reported higher average sleep duration than black respondents, latino respondents or individuals of other race/ethnicity. Black respondents were more likely to take out student loans (55.9%) compared with white respondents (50.4%), whereas latino respondents were less likely to take out student loans (42.6%). However, the amount of student loans was highest among white individuals ($10 434), followed by black individuals ($9335) and latino individuals ($8998). Similarly, black individuals were more likely (41%) and latino individuals less likely (30.3%) to report student debt at age 25 than white individuals (36.3%). No racial differences were found in the amount of student debt reported at age 25.

Table 2

Selected bivariate associations for the full sample (n=4714) and the subsample (n=4459), weighted estimates

Table 3 presents the regression estimates, first for the full sample and then for the subsample. Model 1 examines the association between race/ethnicity and sleep duration after adjustment for demographic and socioeconomic variables. Black inidividuals, latino individuals and those from other racial/ethnic groups slept 21.5, 7.5 and 22.1 fewer minutes, respectively, than white individuals. Lower socioeconomic resources, including less parental wealth and having less than a Bachelor's degree, were associated with less sleep.

Table 3

Unweighted regression coefficients from OLS regression models predicting sleep duration (min) in 2010, NLSY97†

Model 2 adjusts for student loans and shows that student loans is associated with fewer minutes of sleep after controlling for socioeconomic and sociodemographic factors; for every $5000 acquired in student loans, respondents slept 1.9 fewer minutes. Model 3 includes interactions between race/ethnicity and student loans. For every $5000 in student loans, black respondents lost an additional 5.4 min of sleep compared with white respondents. Thus, black individuals slept 15.5 fewer minutes than white individuals if they did not borrow student loans and 42.3 fewer minutes than white individuals if they borrowed around $25 000 in student loans (see figure 1).

Figure 1

Predicted sleep duration (hours) for white individuals and black individuals by amount of student loans borrowed, National Longitudinal Survey of Youth 1997 (n=4714).

A parallel set of models is reported for the subsample. Similar race/ethnic disparities in sleep duration were found in model 1. Student debt reported at 25 shows no association with sleep duration (model 2). Inclusion of interactions between race/ethnicity and both student debt and a quadratic indicator of student debt shows that student debt was significantly associated with sleep duration only among black individuals (b=−3.8; b=0.4, respectively). As shown in figure 2, black individuals slept 19 fewer minutes than white individuals if they held no student debt at age 25 and 32 fewer minutes than white individuals if they held $35 000 in student debt at age 25.

Figure 2

Predicted sleep duration (hours) for white individuals and black individuals by amount of student debt at age 25, National Longitudinal Survey of Youth 1997 (n=4459). Notes: All covariates centred at their grand mean. Year of last enrolment and year respondents reported on debt were centred at 2010.

Sensitivity analysis

We included both a dichotomous and a continuous variable for student loans and student debt in our models. The dichotomous indicators for student loans and student debt were non-significant (results available on request); thus, for parsimony we excluded them from our analyses. We also considered if the association between student loans/debt and sleep duration diminished as a function of time since last college enrolment. Interactions between student loans/debt and time since last enrolment were not statistically significant.

Discussion

Inadequate sleep is widespread in the US population,27 associated with a number of adverse health outcomes,7–9 and a growing public health concern.28 Here, we discuss an important factor that can reduce sleep among young adults—student loans—that has been largely overlooked in research on the social determinants of sleep duration and the health impact of financial stress. In recent decades, the cost of attending college in the USA has increased dramatically29 while the proportion of tuition covered by US federal grant aid has simultaneously declined.30 This has resulted in an increasing burden of student debt among US young adults.1 Surprisingly, few studies have examined if this potential source of financial strain is associated with health,5 ,19 and to our knowledge none have studied the impact of student debt on sleep duration.

In this study, we found race differences in student borrowing, self-reported sleep duration and their joint association. Specifically, black individuals were more likely to acquire student loans and to have student debt at age 25 compared with white individuals. Among those with student loans, black individuals borrowed about $1000 less in student loans by the 2010 interview, but by age 25, reported equivalent amounts of student debt. Latino young adults were the least likely to acquire student loans or have student debt, and also carried the lowest burden of student debt. Further, black individuals slept the least, followed by latino individuals and then white individuals.

Interestingly, student loans and student debt were associated with less sleep among black individuals, but not among white individuals or latino individuals, suggesting that black individuals lose more sleep over their student borrowing compared with white individuals. One possible reason for this finding relates to economic returns from education. Prior research shows that compared with white individuals, black individuals have less earning potential for the same amount of education22 and also acquire less wealth at equivalent levels of education.31 ,32 Earnings and wealth may reduce the strain associated with having student debt, which may explain why white individuals are not as disadvantaged by student debt with respect to sleep duration as are black individuals. Further, financial strain from student borrowing may be worse for black individuals because they are also exposed to more psychosocial stressors than white individuals, including discrimination and harassment, two important stressors that have been associated with sleep duration among black individuals.13 ,17 Our study could not directly evaluate the extent to which worry, perceptions of financial strain, or other psychosocial stressors explained the stronger association between student borrowing and sleep duration among black individuals, but future research could evaluate these potential mechanisms.

Although latino individuals also experience lower returns to earnings,22 we did not see an association between sleep duration and either student loans or student debt. This could be explained by lower student debt among latino individuals, but in our sample latino individuals held similar amounts of student debt as black individuals. Although puzzling, other studies have also noted that, despite generally similar socioeconomic levels, latino individuals appear to have better health compared with black individuals;33 however, research suggests that this finding is not uniformly found across latino ethnic subgroups.34 We were unable to disaggregate our results by ethnicity; thus, our findings may be masking important heterogeneity.

In addition to those already noted, several limitations warrant mention. First, although self-reported sleep duration is commonly used in population-based studies of sleep,21 ,35 the NLSY97 only collects information on self-reported sleep duration on a typical weeknight, which may not accurately represent total sleep duration, since sleep duration can vary substantially between weekday and weekend nights.36 Self-reported sleep duration can also imprecisely estimate actual time spent sleeping, but comparisons of self-reported sleep duration to sleep measured in a laboratory setting show only minimal differences between self-reports and lab-based measurements.37 Thus, we may be underestimating average sleep duration.

Second, our measures of student loans and student debt may be prone to underestimation or recall bias.38 For example, because student debt was measured around age 25, student debt acquired after age 25 was not captured in our measure. Even so, regardless of whether we used measures of student loans or student debt, we found similar associations with sleep duration, bolstering confidence in our findings. Third, although we control for a range of sociodemographic and socioeconomic variables, it is possible that other omitted variables related to the acquisition of student loans or student debt and sleep duration might confound the results. Finally, our sample is generalisable to US young adults who were born between 1980 and 1984. More recent cohorts bear an even greater burden of student loan debt30 and we would therefore expect stronger associations among younger cohorts.

Our results provide preliminary evidence that greater acquisition of student loans is negatively associated with sleep duration among black young adults, but not among white or latino young adults. Prior research finds, however, that student borrowing is associated with other health indicators, regardless of race/ethnicity.5 ,18 ,19 Our study suggests that a more nuanced view of the association between student borrowing and health is needed and that research on the impact of student borrowing on health should focus on those who are at greater risk of both poor health and student borrowing.

What is already known on this subject

  • US students are increasingly financing their higher education via student loans. These loans have some of the least favourable payback terms globally.

  • Financial strain is associated with shorter sleep duration.

What this study adds

  • Student loans are associated with shorter sleep duration among black US young adults, independent of current socioeconomic and sociodemographic factors.

  • Student loans may be an important source of financial strain that reduces sleep and warrants additional investigation.

References

View Abstract

Footnotes

  • Contributors KMW conceptualised the study and led the data management, analysis and writing. JAA assisted with the conceptualisation of the study, data analysis and writing. GCG assisted with the conceptualisation of the study and writing. All authors contributed to each version of the paper and approved the final version.

  • Funding This project was supported by funds from the University of South Carolina’s Office of the Provost’s Social Sciences Grant Programme (11540 A116). The funding source had no role in the study design or submission of the manuscript.

  • Competing interests None declared.

  • Ethics approval This study was considered exempt by The Institutional Review Board at the University of South Carolina and has conformed to the principles embodied in the Declaration of Helsinki.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement The data used in this study are publically available at http://www.bls.gov/nls/nlsy97.htm

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.