socioeconomic status and health: a micro-level analysis of

Socioeconomic Status and Health: A Micro-level Analysis ofExposure and Vulnerability to Daily Stressors*

JOSEPH G. GRZYWACZWake Forest University School of Medicine

DAVID M. ALMEIDAPennsylvania State University

SHEVAUN D. NEUPERTBrandeis University

SUSAN L. ETTNERUniversity of California, Los Angeles

Journal of Health and Social Behavior 2004, Vol 45 (March): 1–16

This study examines the interconnections among education—as a proxy forsocioeconomic status—stress, and physical and mental health by specifying dif-ferential exposure and vulnerability models using data from The National Studyof Daily Experiences (N = 1,031). These daily diary data allowed assessment ofthe social distribution of a qualitatively different type of stressor than has pre-viously been examined in sociological stress research—daily stressors, or has-sles. Moreover, these data allowed a less biased assessment of stress exposureand a more micro-level examination of the connections between stress andhealth by socioeconomic status. Consistent with the broad literature describingsocioeconomic inequalities in physical and mental health, the results of thisstudy indicated that, on any given day, better-educated adults reported fewerphysical symptoms and less psychological distress. Although better educatedindividuals reported more daily stressors, stressors reported by those with lesseducation were more severe. Finally, neither exposure nor vulnerabilityexplained socioeconomic differentials in daily health, but the results clearlyindicate that the stressor-health association cannot be considered independentof socioeconomic status.

1

Stress exposure and vulnerability play a cen-tral role in the sociological study of socioeco-

nomic differentials in physical and mentalhealth (Aneshensel 1992; Baum, Garofalo, andYali 1999; Kessler 1979; Pearlin 1989).Individuals in lower socioeconomic groupsshoulder a disproportionate amount of bothacute and chronic stressful events (Turner andLloyd 1999; Turner, Wheaton, and Lloyd1995), and there is a well-established linkbetween higher levels of stress and decreasedphysical and mental health (Cohen and Herbert1996; Kelly, Hertzman, and Daniels 1997;Theorell 1982). Indeed, differential exposureand vulnerability to stress are among the pri-mary hypotheses offered by sociologists to

#1433—Jnl of Health and Social Behavior—Vol. 45 No. 1—45101-grzywacz

* This research was partially supported by a grantfrom the National Institute on Alcohol Abuse andAlcoholism (AA12744), awarded to Joseph G.Grzywacz and a grant from the National Institute onAging (AG19239), awarded to David M. Almeida.We are grateful to Nadine F. Marks of the Universityof Wisconsin for helpful comments on a preliminarydraft of this paper. Address correspondence to:Joseph G. Grzywacz, Department of Family andCommunity Medicine, Wake Forest UniversitySchool of Medicine, Medical Center Boulevard,Winston-Salem, NC 27157-1084.

explain the socioeconomic status-health asso-ciation (Evans, Barer, and Marmor 1994;House and Williams 2000; Wilkinson 1996).

The study of stress as a mediator of thesocioeconomic status and health relationshipremains encumbered, however. First, previousstudies have overlooked daily stressors, a formof stress that can and should be differentiatedfrom chronic or acute stressors because of theireffects on health (Wheaton 1994). Moreover,while the study of acute and chronic stress pro-vides a wide-angle view of general conditionsthat are socially structured, daily stressors cap-ture the day-to-day experiences that are contin-gent upon, and unfold within, a broader con-text shaped by socioeconomic status(Aneshensel 1992; Krieger, Williams, andMoss 1997). Next, previous research examin-ing socioeconomic status, stress, and healthsuffers from a type of ecological fallacy(Robinson 1950), because inferences havebeen made about within-person processes (anindividual’s daily stressful experiences under-mines her/his health) from studies usingbetween-person designs (Affleck, Zautra,Tennen, and Armeli 1999). Finally, the study ofstress as a mediator of the status-health rela-tionship frequently does not examine bothphysical and mental health, and a limited rangeof outcomes is a standing limitation in vulner-ability studies of stress and health (Aneshensel1992). In this study, we seek to attenuate theselimitations by examining stressor exposure andvulnerability models in daily data obtainedfrom the first nationally representative study ofdaily stressful experiences (National Study ofDaily Experiences).

THEORETICAL AND EMPIRICALBACKGROUND

Conceptual Underpinnings

The “life stress hypothesis,” including bothdifferential exposure and vulnerability, is oneof the major explanations invoked to explainhealth disparities such as those exemplified bysocioeconomic status (Baum et al. 1999;House and Williams 2000). The exposure com-ponent argues that lower status individuals aresubjected to quantitatively more physical, psy-chological, and social stressors than their high-er status counterparts (Dohrenwend 1973;Pearlin 1989), and this higher level of exposure

accounts for the increased incidence of mor-bidity and mortality among members of disad-vantaged groupings. The vulnerability compo-nent suggests that disadvantaged individualsare more vulnerable to the negative effects oflife stressors than more advantaged individualsbecause they have fewer or less effective cop-ing resources (Kessler 1979; Kessler andCleary 1980; Kohn 1972), or because theirstressors are qualitatively more potent.

Implicit in discussions of exposure and vul-nerability is acknowledgement that stressorscan take a variety of forms (Pearlin 1989;Wheaton 1994). The first form of stressor ischaracterized by specific or discrete lifeevents, and they are believed to stimulate animmediate and intense physiological responsethat returns to baseline with successful adapta-tion (for comprehensive discussion of differentphysiological stress response scenarios seeMcEwen 1998). Chronic or enduring experi-ences, such as unemployment or financialhardship (Catalano and Dooley 1983) charac-terize a second form of stressor, and a finaltype of stressor reflects frequent, perhapsminor “hassles” (Kanner, Coyne, Schaefer, andLazarus 1981). The physiological response tothese latter two forms of stressors can be char-acterized by heightened levels of physiologicalarousal that endure over time or frequent physio-logical spikes, both of which lead to extensivewear and tear over time (McEwen 1998).

The importance of daily hassles as a uniqueform of stress (Kanner et al. 1981; Wheaton1994) has not been fully appreciated in studiesof socioeconomic disparities in health. Thereis, for example, a long history of researchexamining life events and several reviews sug-gesting that socioeconomic disparities inhealth are attributed more to differential vul-nerability to life events than to differentialexposure (Aneshensel 1992; Kessler andCleary 1980; Kessler, Price, and Wortman1985; Thoits 1983). More recently, scholarshave demonstrated that lower status individu-als are exposed to more chronic stressors thantheir higher status counterparts (Turner et al.1995), and that the personal and socialresources involved in the stress process are lessavailable to lower status individuals (Pearlin1989; Turner and Lloyd 1999; Wheaton 1983).By contrast, a descriptive epidemiology ofdaily stressors and explicit examinations ofdifferential vulnerability to daily stressors by

2 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR


socioeconomic status are absent in the litera-ture (Eckenrode and Bolger 1995).

The study of daily stressors, however, offersunique insight into the ordinary circumstancesthat may sustain and exacerbate social inequal-ities in health. First, it is clear that daily stres-sors have potent effects on physical and mentalhealth (Almeida and Kessler 1998; Almeida,Wethington, and Kessler 2002; Bolger,DeLongis, Kessler, and Schilling 1989;Delongis, Folkman, and Lazarus 1988; Lu1991), and some studies suggest that dailystressors have the strongest effects on healthsymptoms and health states (Jandorf,Deblinger, Neale, and Stone 1986; Weinberger,Hiner, and Tierney 1987). Next, although themagnitude of daily stressor effects is partiallyattributed to indirect effects from chronic oracute stressors, daily stressors do also exertadditive, independent effects on physical andmental health (Wheaton 1994); thus, the addi-tional and possible cumulative toll of dailyhassles can be substantial. Moreover, there isevidence that the effects of daily stressors onphysical and mental health are exacerbated bychronic stressors such as overcrowding, poorneighborhood quality (Caspi, Bolger, andEckenrode 1987; Lepore, Evans, and Palsane1991), or acute life events (Burks and Martin1985). In summary, studying daily stressorsprovides an important microlevel complementto wide angle studies of socioeconomic status,stress, and health because, as Wheaton (1994)contends, “daily stressors capture a level ofsocial reality that is untapped by other concep-tualizations of stress, and they offer insightinto the mundane realities of daily life” (p. 87)that characterize social disadvantage and maycontribute to inequalities in health.

Theoretical and Methodological Challenges

Although the study of daily stress is not new(for a review, see Eckenrode and Bolger 1995),there are several theoretical and methodologi-cal challenges to examining the role of dailystressors in socioeconomic health disparities.First, it is challenging to characterize the link-age between socioeconomic status and expo-sure to daily stressors. On one hand, theory andevidence suggests that lower status individualswould bear a disproportionate burden of dailyhassles than higher status individuals (Pearlin1989). For example, over 40 percent of the

variance in daily hassles can be attributed toprevious or ongoing stressors (Wheaton 1994);thus, lower status individuals should reportmore daily stressors because they have morestressful life events and more chronic stressors(Turner et al. 1995). On the other hand, stres-sors frequently arise from role-related transac-tions between individuals and their environ-ments (Aneshensel 1992; Turner and Wheaton1995), and, to the extent that higher status indi-viduals are engaged in more key roles or havehigher expectations from their roles, theyshould report more daily stressors than theirlower status counterparts. In short, to theextent that daily hassles can arise from social-ly structured hardships (e.g., poverty) as wellas status-based responsibilities or privileges(e.g., supervising a large staff), it remains anempirical question whether daily hardshipssystematically vary by socioeconomic status(Eckenrode and Bolger 1995).

Measurement is another challenge con-fronting the study of daily stressors and theirimplications in socioeconomic inequalities inhealth. Checklist formats, such as the DailyHassles Scale (Kanner et al. 1981), are mostcommonly used to study daily stressors(Eckenrode and Bolger 1995); unfortunately,stress measures that use checklist formats,daily or otherwise, tend to be problematic(Thoits 1983; Wethington, Brown, and Kessler1995). First, checklist measures are not sensi-tive to overall stressor severity unless they arespecifically adapted to probe the meaning anddetail of the stressor (Wethington et al. 1995).Foul weather, for example, may be an unpleas-ant nuisance to a business professional, but itcan be dangerous or reduce personal incomefor an outdoor laborer. Thus, the severity of thestressor is likely to be linked to social statusand health, even if the stressor itself is not.Another problem with checklist measures isthat they do not adequately cover the broaddomain of daily stressors, which could lead toartifactual biases in estimates of stressor expo-sure (Thoits 1983). Finally, unless employedwithin a repeated measures design, checklistmeasures rely upon retrospective reports ofstressful experiences (e.g., Daily Hassles Scalerequires consideration of the past month).Apart from potential for recall biases in self-reports (Schwarz 1999), the overlap of retro-spective reports of both stressors and health-related outcomes make causal inference diffi-cult, even when using a longitudinal design.

SOCIOECONOMIC STATUS AND HEALTH 3


This last issue raises a general limitation ofthe socioeconomic status, stress, and health lit-erature. Observational designs based on cross-sectional or longitudinal data collected fromsurveys are well suited for between-personcomparisons, such as differential exposure tochronic stressors by socioeconomic status;however, they are not well equipped to assesswithin-person associations between stress andhealth and whether these associations differ bya between-person factor such as socioeconom-ic status. Given ample evidence indicating thatpeople who are under stress suffer more healthproblems (Cohen and Herbert 1996), within-person associations between stress and healthare frequently inferred from between-persondesigns. However, between-person associa-tions can mask variation in within-person asso-ciations in terms of both magnitude and direc-tion (Kenny, Bolger, and Kashy 2002; Tennen,Affleck, Armeli, and Carney 2000), suggestingthat inferences from previous studies of somat-ic or psychiatric vulnerability to stressors bysocioeconomic status may be tenuous. In short,it is important to complement previous socio-logical stress and health research with designsthat allow stronger inferences of within-personassociations and the between-persons factorsthat may affect these associations.

Regarding the measurement challenge,Almeida and colleagues (2002) have recentlyarticulated a new approach for studying dailystressors. The Daily Inventory of StressfulExperiences (Almeida et al. 2002) consists ofa series of stem questions administeredthrough daily telephone interviews askingwhether stressors had occurred in broaddomains of life in the past 24 hours (e.g.,“since the last time we spoke, did anythinghappen at home that most people would con-sider stressful?”), along with a set of inter-viewer guidelines for probing affirmativeresponses. Individuals’ open-ended responsesto the questions and interviewer probes aretape recorded, transcribed, and coded for sev-eral characteristics including the nature of thestressor (e.g., interpersonal, risk to health andsafety, risk to future plans) as well as subjec-tive and objective appraisals of severity. Thisinvestigator-based approach facilitates distin-guishing between the stressful event (e.g., con-flict with spouse) and the affective response tothe stressor (e.g., crying or feeling sad). Thisapproach also does not superimpose an a pri-ori set of stressors; rather, it allows respon-

dents to evaluate and report their own experi-ences. Finally, the probes allow fine distinc-tions in the appraisal of the stressor.

In summary, previous research examiningthe “life stress” explanation for socioeconomicdisparities in health has overlooked daily stres-sors as a unique form of stress. A study ofexposure and vulnerability to daily stressors bysocioeconomic status provides an importantcomplement to studies of acute and chronicstressors because of the important effects dailyhassles have on symptoms and health states,and because it would examine a unique “levelof social reality” (Wheaton 1994:87) that maysustain or exacerbate health disparities.Although there are several challenges to such astudy, tools and data are available to beginexploring this relatively uncharted yet poten-tially fertile domain. Thus, the primary goalsof this study are to offer a micro-level perspec-tive on the role of stress for socioeconomic dis-parities in health by studying the distributionof daily stressors in the adult population, andto systematically test differential exposure andvulnerability hypotheses linking socioeconom-ic status and health.

METHODS

Sample

Data for the analyses are from the NationalStudy of Daily Experiences. Respondents were1,031 adults (562 women, 469 men), all ofwhom had previously participated in theNational Survey of Midlife Development inthe United States, a nationally representativetelephone-mail survey of 3,032 people, aged25–74 years, carried out in 1995–1996 underthe auspices of the John D. and Catherine T.MacArthur Foundation Network on SuccessfulMidlife. Respondents in the National Study ofDaily Experiences were randomly selectedfrom the National Survey of MidlifeDevelopment in the United States sample andreceived $20 for their participation in the pro-ject. Over the course of eight consecutiveevenings, respondents completed short tele-phone interviews about their daily experiences.Data collection spanned an entire year (March1996 to April 1997) and consisted of 40 sepa-rate “flights” of interviews, with each flightrepresenting the eight-day sequence of inter-views from approximately 38 respondents. The



initiation of interview flights was staggeredacross the day of the week to control for thepossible confounding between day of studyand day of week. Of the 1,242 National Surveyof Midlife Development in the United Statesrespondents we attempted to contact, 1,031agreed to participate, yielding a response rateof 83 percent. Respondents completed an aver-age of seven of the eight interviews, resultingin a total of 7,221 daily interviews.

The National Study of Daily Experiencessubsample and the National Survey of MidlifeDevelopment in the United States sample fromwhich it was drawn had very similar distribu-tions for age, marital status, and parenting sta-tus. The National Study of Daily Experiencessample had a slightly greater percentage ofwomen (54.5% versus 51.5%), was better edu-cated (60.8% of the National Survey of MidlifeDevelopment in the United States sample hadat least 13 years of education versus 62.3% ofthe National Study of Daily Experiences sub-sample), and had a smaller percentage ofminority respondents than the National Surveyof Midlife Development in the United Statessample. Of the National Study of DailyExperiences sample, 90.3 percent were white,5.9 percent African American ,and 3.8 percentall other races (cf. a National Survey ofMidlife Development in the United Statessample that was 87.8% white, 6.8% AfricanAmerican, and 4.4% all other races).Respondents for the present analysis were onaverage 47 years old. 38 percent of the house-holds reported having at least one child under18 years old in the household. The averagefamily income was between $50,000 and$55,000. Men were slightly older than women,had similar levels of education, and were morelikely to be married at the time of the study(77% of the women versus 85% of the menwere married).

Sampling weights correcting for selectionprobabilities and non-response allow the origi-nal National Survey of Midlife Developmentin the United States sample to match the com-position of the U.S. population on age, sex,race, and education based upon the October1995 Current Population Survey. A new sam-pling weight for the National Study of DailyExperiences sample was computed by dividingthe unweighted National Study of DailyExperiences sample by the National Study ofDaily Experiences sample weighted using theNational Survey of Midlife Development in

the United States sampling weight, and thenmultiplying this proportion by the originalMIDUS sampling weight (i.e., unweighted N /weighted N * National Survey of MidlifeDevelopment in the United States S weight).Because the variables used in constructing thesample weights were controlled in all analysesand the pattern of results from analyses withand without the sampling weights were similar,unweighted results are presented (Winship andRadbill 1994).

Measures

Socioeconomic status was operationalizedas a series of dichotomous indicators of educa-tional attainment representing less than highschool education (reference category); highschool or some college; and college graduate.This strategy was chosen because it capturesthe well-established gradient of socioeconom-ic disadvantage (Adler et al. 1994; Marmot,Ryff, Bumpass, Shipley, and Marks 1997;Marmot et al. 1998), and it captures the prima-ry educational benchmarks that provide thefoundation for subsequent stratificationprocesses by occupation and earnings (Marksand Shinberg 1998). Moreover, educationalattainment has been the primary proxy forsocioeconomic status used in previous studies,thereby allowing comparability with otherstudies; it is less prone to exhibiting missingdata values; it is relatively stable across the lifecourse after early adulthood; it is more compa-rable across men and women than occupation;and it is more comparable across single andmarried persons than income. Most important-ly, education is less prone to endogeneity biasfrom reverse causality (e.g., health affectingthe socioeconomic status measure) than mea-sures such as income and occupation.

Daily psychological distress was opera-tionalized using an inventory of ten emotionsexpanded from the psychological distress scaledesigned for the National Survey of MidlifeDevelopment survey (Mroczek and Kolarz1998) and queried during each telephone inter-view. This scale was developed from the fol-lowing well-known and valid instruments: TheAffect Balance Scale (Bradburn 1969), theUniversity of Michigan’s CompositeInternational Diagnostic Interview (Kessler etal. 1994), the Manifest Anxiety Scale (Taylor1953), and the Center for Epidemiological



Studies Depression Scale (Radloff 1977).Respondents were asked how much of the timethey felt: worthless; hopeless; nervous; restlessor fidgety; that everything was an effort; and“so sad that nothing could cheer you up.”Response categories for the index items were 1= none of the time, 2 = a little of the time, 3 =some of the time, 4 = most of the time, and 5 =all of the time. Scores across the ten itemswere summed (α = .89).

Daily physical symptoms were measuredusing a shortened version of Larsen andKasimatis’s (1991) physical symptom check-list. Items that overlapped with the psycholog-ical distress scale (e.g., “urge to cry”) wereomitted. Our five-item scale assessed five con-stellations of symptoms: aches/pain(headaches, backaches, and muscle soreness),gastrointestinal symptoms (poor appetite, nau-sea/upset stomach, constipation/diarrhea),chest pain or dizziness (symptoms often asso-ciated with cardiovascular functioning), flusymptoms (upper respiratory symptoms, sorethroat, runny nose, fever, chills) and a catego-ry for “other” physical symptoms or discom-forts. Open-ended responses to the “other”physical symptoms question were subsequent-ly coded and placed into an existing category,deleted if the symptom was psychological(e.g., felt anxious), or left in a miscellaneouscategory if no other category existed. Each daythe respondents indicated how frequently theyexperienced each symptom over the past 24hours on a five-point scale where 1 = none ofthe time, 2 = a little of the time, 3 = some ofthe time, 4 = most of the time, and 5 = all ofthe time. Scores across the five items weresummed (α = .71).

Daily stressors were assessed with the DailyInventory of Stressful Experiences (Almeida etal. 2002). The Daily Inventory of StressfulExperiences is a semi-structured instrumentcontaining seven “stem” questions for identify-ing whether stressful events occurred in variouslife domains, as well as a series of questions forprobing affirmative responses (see theAppendix for a detailed description of the“stem” questions and examples of “probes”).Almeida and colleagues’ (2002) analyses high-light several descriptive features of DailyInventory of Stressful Experiences measuresthat are highly relevant to the current study.First, respondents reported experiencing atleast one stressor on 37.8 percent of the inter-view days, and multiple stressors were reported

on 10 percent of interview days. Next, the mostcommon form of daily stress for women andmen was interpersonal stressors, followed bywork stressors for men and network stressorsfor women. Finally, although subjective andobjective appraisals of stressors are based onthe same experience, the association betweenthese measures was modest (r = .36). Thus, theDaily Inventory of Stressful Experiences pro-duces estimates of daily stressors with amplevariation, and objective and subjective charac-terizations of stressor severity appear to be rel-atively independent of each other.

For each daily interview, individuals whoresponded affirmatively to any of the stemquestions received a value of 1 on an indicatorvariable of any stress; they were coded 0 oth-erwise. Respondents’ narrative responses toinvestigator probes provided objective infor-mation on the content of the stressful experi-ences as well as the meaning of the stressor forthe respondent. Objective severity, similar toBrown and Harris’s (1978) ratings of short-term contextual threat, was assigned by trainedcoders based upon the degree of disruptivenessand unpleasantness associated with the stres-sor. Coders’ scores ranged from a minor ortrivial annoyance (coded 1) to a severely dis-ruptive event (coded 4). Inter-rater reliability(kappa) on the objective severity measure was.75. Subjective severity reflects respondents’assessments of each stressful event on a four-point scale ranging from “not at all stressful”to “very stressful.” Four mutually exclusivecategorical variables reflecting stressor severi-ty were then constructed by first dichotomiz-ing each of the severity measures as high ver-sus low (i.e., greater than or equal to one stan-dard deviation above the sample mean codedas one); then categories were created reflectinglow subjective/low objective severity, low sub-jective/high objective severity, high subjec-tive/low objective severity, and high subjec-tive/high objective severity. The cutoff of themean plus one standard deviation was chosenbecause it is commonly used to represent“high” values (Aiken and West 1990; Jaccard,Turrisi, and Wan 1990) and it provided a moreconservative measure of severe stressors thanalternatives such as a mean split.

ANALYSES

The method used to examine the association



between physical symptoms, psychologicaldistress, stressor exposure, and socioeconomicstatus within individuals over time was basedon a multilevel model, also commonly referredto as a hierarchical linear model (Bryk andRaudenbush 1992). In this multilevel model, alag-analysis was used, with prior day physicalsymptoms predicting current day physicalsymptoms, and prior psychological distresspredicting the level of psychological distressreported on the current day. By controlling forprior-day values for physical symptoms anddistress when predicting the current day val-ues, the specification is equivalent to (butmore flexible than) a change score model.

Stressor exposure was alternately defined as(1) whether the respondent experienced anystressor, and (2) whether the respondent expe-rienced any stressors in the following severitycategories: low subjective/low objective, lowsubjective/high objective, high subjective/lowobjective, high subjective/high objective.Respondents experiencing multiple stressorson the same day were categorically assignedbased upon the average severity of all stres-sors. In both sets of analyses, persons experi-encing no stressors were the comparisongroup.

The simple form of an hierarchical linearmodel can be conceived of as two separatemodels, one a within-person model (Level 1)and the other a between-person model (Level2). A distinctive feature of hierarchical linearmodel is that the intercepts and slopes areallowed to vary across persons (Lee and Bryk1989), allowing estimates of between-personmodels of within-person variability. To exam-ine the temporal links between daily psycho-logical distress and stressors, we fit a within-person model essentially equivalent to 1,031regressions assessing daily covariation ofstressors and distress. The unit of observationfor each of these regressions is the person-day,so the sample size for each of these regressionsis N = 8. Using a simple example in whichhealth depends on a single explanatory vari-able—stressors—the model can be expressedas:

Level 1: HEALTHit = a0i +

a1iSTRESSORt + eit,(1)

where HEALTHit is the reported health out-come (i.e., physical symptoms or psychologi-

cal distress) of person i on day t, STRESSORindicates whether person i experienced a stres-sor on day t, a0i is the intercept indicating per-son i’s average level of health when no stressorwas reported, a1i is the slope indicating theassociation between stressor exposure andhealth for person i, and eit is the random com-ponent or error associated with distress of per-son i on day t . To estimate average effects forthe entire sample, the intercepts and slopes ofthe Level 1 within-person model become theoutcomes for the Level 2 between-personequations as follows.

Level 2: a0i = B0 + di, (2)

a1i = B1 + gi (3)

The sample size for each of the Level 2 regres-sions is N = 1,031. Equation 2 shows that per-son i’s average health score across the diarydays (a0i) is a function of the intercept for theentire sample (B0)—the grand mean of thesample—and a random component or error(di). Likewise, equation 3 shows that person i’sslope between distress and health (a1i) is afunction of the grand mean of the entire sam-ple (B1), and a random component or error (gi).As discussed earlier in this paragraph, thisbasic model was extended to include prior dayphysical symptoms or negative affect ascovariates for their respective outcomes toattenuate the possibility of reverse causality,whereby previous days poor health (physical ormental) contributed to both experiencing astressor and health problems on any given day.

Hierarchical linear modeling provides theflexibility to allow the intercepts and slopes tovary across persons by stable individual char-acteristics (e.g., socioeconomic status). Forexample, to examine socioeconomic status dif-ferences in the daily covariation of distress andstressor exposure, one can formulate the fol-lowing model:

Level 1: DISTRESSit = a0i +

a1iSTRESSOR + eit (4)

Level 2: a0i = B0 + B1(SES) + di, (5)

a1i = B2 + B3(SES) + gi (6)

Equations 5 and 6 model socioeconomic status



differences in Level 1 intercepts and slopes. Ofparticular note is equation 6 because it consid-ers the differential vulnerability hypothesis bytesting whether the stressor-distress slopes (a1i)vary according to socioeconomic status.

In these analyses, a model where the slope isconstrained to be equal across subjects (forexample, a model where the strength of theassociation between distress and stressor expo-sure is the same across all participants) is com-pared to one where the slopes are allowed tovary across individuals (in this example, amodel where the association is not the sameacross individuals with differing socioeconom-ic status). The models are compared by takingthe difference between the obtained model fits(i.e., –2 ln(likelihood)) and testing its signifi-cance with the degrees of freedom equal to thedifference in the number of parameters of thetwo models (df = 2, in this example) (Bryk andRaudenbush 1992). If the models are not sig-nificantly different, the model constraining theslopes to be equal is chosen for reasons ofparsimony.

RESULTS

Table 1 presents means and standard devia-tions for all variables of interest across educa-tion levels. There is a clear inverse gradient indaily physical symptoms and daily psycholog-ical distress such that college graduates report-ed better health than those with a high schooldegree or some college, and the latter group

was in turn in better health than those with lessthan a high school degree. More conservativeanalyses (i.e., p ≤ .01), however, suggested nodifferences in physical symptoms or psycho-logical distress between individuals with lessthan a high school degree and those with ahigh school degree or some college.

In terms of overall exposure to stressors,respondents with less than a high schooldegree reported experiencing stressors on 30percent of the study days, while those with ahigh school degree and/or some college andthose with a college degree reported experi-encing stressors on 38 percent and 44 percentof the study days, respectively. Although bettereducated individuals reported stressors on alarger percentage of days, the stressors thatwere experienced were objectively less severe,on average, for college graduates and thosewith a high school degree than for those withless than a high school degree. College gradu-ates experienced stressors that were also sub-jectively less severe than stressors experiencedby individuals with less than a high schooldegree.

Hierarchical linear modeling estimates ofthe effects of education and stressors on dailyphysical symptoms and daily psychologicaldistress are presented in Tables 2 and 3. Model1 in each table is the baseline model, showingthe effect of education on daily physical healthand psychological distress adjusting for theeffects of age, gender, race, and previous day’ssymptoms (physical or psychological, depend-ing upon outcome). Because these models con-



TABLE 1. Means and Standard Deviations for Physical Symptoms, Psychological Distress, andStressor Characteristics by Educational Attainment

Less than High School High School or Some College College Graduate(n = 78) (n = 642) (n = 311) F

Physical symptoms 2.39 1.77a 1.43a,b 9.87***(3.21) (1.79) (1.44)

Psychological distress 3.51 1.91a 1.30a,b 21.19***(5.79) (2.77) (1.83)

Frequency of stressors 0.30 0.38a 0.44a,b 9.40***(0.34) (0.27) (0.23)

Severity of stressor—Subjectivec 2.91 2.76 2.64a 5.13**

(0.89) (0.69) (0.58)—Objectivec 2.26 1.81a 1.75a 16.72***

(0.90) (0.69) (0.56)

* p ≤ .05; ** p ≤ .01; *** p ≤ .001Note: Weighted estimates from the National Study of Daily Experiences.a different ( p ≤ .05) from individuals with less than a high school degree.b different from individuals with a high school degree or some college.c values are based on reported stressors only.

trol for symptoms on previous days, theyshould be interpreted as models of healthchange as opposed to models of levels ofhealth. Additionally, it is important to recog-nize that this model specification will likelyattenuate associations between education andhealth because part of the exogenous effect ofsocioeconomic status on current symptomswill be indirect, through the endogenous effectof prior symptoms.

The remaining models systematically assessdifferent aspects of the differential exposureand vulnerability hypotheses. Model 2 in eachtable adds an indicator covariate assessing ifany stressors were reported on the given day.

Model 3 for each of the outcomes adds dummyvariables reflecting both objective and subjec-tive stressor severity. Models 4 and 5 for eachof the outcomes add interaction terms of thesemeasures with education to assess vulnerabili-ty to stressors and stressor severity. (Note: sen-sitivity analyses using a household incomecovariate in addition to educational attainmentyielded a pattern of results identical to thosemodels without the income covariate. Tostreamline already complex models, resultsfrom the models without household income arereported below.)

Table 2 describes the analyses predictingwithin-person change in daily physical symp-



TABLE 2. Hierarchical Linear Modeling Estimates and Standard Errors Predicting Daily PhysicalSymptoms

Model 1 Model 2 Model 3 Model 4 Model 5

Less than High School (reference) .— .— .— .— .—High School or Some College –.36** –.36** –.35** –.25 –.23

(.14) (.14) (.13) (.17) (.15)College Graduate –.54*** –.57*** –.55*** –.44** –.41**

(.15) (.14) (.14) (.15) (.15)Previous Physical Symptoms .38*** .38*** .38*** .38*** .39***

(.01) (.01) (.01) (.01) (.01)Any Stressors .35*** .75***

(.05) (.19)Stressor Severity—No Stressor (reference) .— .——Low Subjective/Low Objective .15 .75*

(.08) (.38)—Low Subjective/High Objective .28** –.10

(.11) (.45)—High Subjective/Low Objective .46*** 1.18***

(.08) (.35)—High Subjective/High Objective .45*** .82***

(.06) (.24)HS/College * Any stress –.41

(.20)*College grad * Any Stress –.44

(.21)*HS/College * Low Sub/Low Obj. –.57

(.39)College Grad. * Low Sub/Low Obj. –.72

(.40)HS/College * Low Sub/High Obj. .19

(.47)College Grad. * Low Sub/High Obj. .68

(.48)HS/College * High Sub/Low Obj. –.78*

(.36)College Grad. * High Sub/Low Obj. –.75*

(.37)HS/College * High Sub/High Obj. –.36

(.26)College Grad * High Sub/High Obj. –.45

(.27)Variance Explained 71.1% 72.5% 73.3% 72.9% 73.7%

* p ≤ .05; ** p ≤ .01; *** p ≤ .001Note: Unweighted estimates from the National Study of Daily Experiences. Models adjust for the effects of age, gen-der, and race.

toms. Independent of the expected associationbetween physical symptoms on the previousday and current symptoms, respondents with ahigh school degree and/or college degree wereless likely to have an increase in physicalsymptoms than those without a high schooldegree (see model 1). Model 2 shows a posi-tive association between daily stressors andcurrent daily symptoms, indicating that ondays when people reported any stressors, theyalso reported more physical symptoms com-pared to days when no stressors were reported.Exposure to any stressors did not mediate theeducation-symptom association, which is con-sistent with the evidence from Table 1 indicat-

ing that better-educated individuals reportedstressors more, rather than less, frequently.Model 3 shows that individuals report morephysical health symptoms when they experi-ence stressors that are either objectively orsubjectively severe, compared with experienc-ing no stressors. These results further suggestthat subjective severity has a stronger associa-tion with physical health than does objectiveseverity. Indeed, the magnitude of the estimatefor stressors characterized as high subjec-tive/high objective was comparable to thosecharacterized as high subjective/low objective.Again however, the education-symptom rela-tionship was not mediated by the inclusion of



TABLE 3. Hierarchical Linear Modeling Estimates and Standard Errors Predicting Daily NegativeAffect.

Model 1 Model 2 Model 3 Model 4 Model 5

Less than High School (reference) .— .— .— .— .—High School or Some College –.97*** –.99** –.93*** –.52** –.47*

(.20) (.19) (.19) (.21) (.20)College Graduate –1.15*** –1.24** –1.15*** –.61*** –.57**

(.21) (.21) (.20) (.22) (.21)Previous Physical Symptoms .32*** .31*** .32*** .31*** .31***

(.01) (.01) (.01) (.01) (.01)Any Stressors 1.01*** 2.71***

(.07) (.27)Stressor Severity—No Stressor (reference) .— .——Low Subjective/Low Objective .20 –.12

(.11) (.54)—Low Subjective/High Objective .26 .98

(.15) (.63)—High Subjective/Low Objective 1.08*** 4.11***

(.12) (.49)—High Subjective/High Objective 1.72*** 3.61***

(.09) (.34)HS/College * Any stress –1.66***

(.29)College grad * Any Stress –2.06***

(.30)HS/College * Low Sub/Low Obj. .41

(.56)College Grad. * Low Sub/Low Obj. .20

(.57)HS/College * Low Sub/High Obj. –.79

(.66)College Grad. * Low Sub/High Obj. –.78

(.68)HS/College * High Sub/Low Obj. –3.17***

(.51)College Grad. * High Sub/Low Obj. –3.30***

(.53)HS/College * High Sub/High Obj. –1.83***

(.36)College Grad * High Sub/High Obj. –2.40***

(.38)Variance Explained 71.3% 74.2% 77.2% 74.8% 77.5%

* p ≤ .05; ** p ≤ .01; *** p ≤ .001Note: Unweighted estimates from the National Study of Daily Experiences. Models adjust for the effects of age, gen-der, and race.

different risk appraisals of stressors byeducation.

The final two models reported in Table 2assess the mediating role of stressor vulnera-bility in the socioeconomic status-health asso-ciation by examining between-person differ-ences in the effects of stressor characteristicson physical symptoms. Model 4 suggests thatexperiencing a stressor is associated with anincrease in average physical symptoms for allindividuals; however, the increase is larger forthose with less than a high school degree thanfor those with more education (.75, .34, and.31, respectively, for individuals with less thanhigh school, high school and some college, andcollege graduates). Model 4 also suggests thatpart of the health differences between thosewith less than a high school degree and thosewith a high school degree and some college areexplained by greater vulnerability to stressorsby those with less education. Model 5 furthersuggests that individuals with some collegeand college graduates are particularly less vul-nerable to stressors characterized by high sub-jective/low objective severity.

The first three models of Table 3 describeresults from analyses predicting within-personchanges in daily psychological distress from adifferential exposure perspective. Model 1demonstrates that education influences currentday psychological distress independent of pre-vious day distress as well as age, gender, andrace. Model 2 shows that when a stressor isreported there is a greater average increase indistress in contrast to days when a stressor isnot reported. A comparison of estimates frommodel 2 with those of model 1 shows thatalthough the differences in magnitude aresmall and insignificant, the direction of changeis consistent with the hypothesis that thegreater exposure to daily stressors among col-lege graduates (demonstrated in Table 1) sup-presses educational differences in distress (i.e.,the effect of being a college graduate gets larg-er after controlling for whether any stressorswere experienced). This estimate is slightlyreduced once the stressor severity is includedin the model (see model 3), as one wouldexpect, given that college graduates experienceless severe stressors. Although experiencingstressors that are not subjectively severe doesnot appear to undermine mental health, stres-sors with high subjective severity are associat-ed with increased negative affect, and thiseffect is accentuated when subjective

appraisals cor respond with objectiveappraisals. As with the physical health out-come, the overall pattern of results from thefirst three models of negative affect provide noevidence that differential exposure to dailystress explains educational differences in men-tal health.

The final models reported in Table 3 assessstressor vulnerability as a mediator of the edu-cation-distress association by considering ifthe slopes for the stressor characteristics varyby education. Model 4 suggests that nearly halfof the educational differences in daily negativeaffect may be attributed to differential vulner-ability whereby better-educated people are lessreactive to any stressors in contrast to thosewith less than a high school degree. Estimatesfrom model 6 further suggest that individualswith less than a high school degree are partic-ularly more vulnerable to those stressors with ahigh level of subjective severity.

DISCUSSION

Four main patterns of results emerged fromthis micro-level examination of the intercon-nections between socioeconomic status, stress,and physical and mental health. First, higherstatus individuals, using education as proxy,reported better physical and mental healthacross the eight days of the interview, and theyhad more day-to-day improvements (or alter-natively, smaller decrements) in physicalsymptoms and psychological distress thanlower status individuals. Second, exposure todaily stressors was status-related, but higher-status individuals reported more, rather thanfewer, stressful situations than lower-statusindividuals. However, of the stressors that werereported, lower-status individuals’ stressorswere more severe. Next, the results of thisstudy provide strong evidence that daily stres-sors contribute to decrements in individuals’physical and mental health, particularly whenthe stressor is subjectively severe. Finally, incontrast to individuals with less education, bet-ter-educated individuals’ physical and mentalhealth were influenced less by daily stressors(i.e., they were less vulnerable). These resultscomplement and extend previous research inseveral ways, and they compel new ways ofthinking about and researching the stressprocess and its role in socioeconomic inequali-ties in health.



The descriptive epidemiology of daily stres-sors in this study extends the literature in sev-eral ways. First, most studies approach thecontext of socioeconomic disadvantage from abroad perspective by studying acute or chronicstressors; this is the first study using national-ly representative data that examines daily stres-sors in the context of socioeconomic status. AsWheaton (1994) argues, daily stressors offerunique insight into the daily lives that areshaped by status hierarchies. Next, the investi-gator-based approach to measuring stressallowed consideration of both the incidence ofstressors as well as the meaning of these stres-sors. Wethington and colleagues (1995) haveargued that the distinction between a stressorevent and its meaning may be pivotal to under-standing the stress-health relationship. Thiscontention was partially borne out in ouranalyses; that is, subjective appraisals of stres-sor severity were more strongly and consistent-ly associated with physical and mental healththan objective assessments of severity were.Finally, our results suggest that socioeconomicdifferentials in acute or chronic stressors donot manifest themselves in daily stressors.

The distribution of daily stressors in thisstudy vis-à-vis previous studies of acute lifeevents and chronic stressors raises severalimportant issues for future empirical and theo-retical development, because of its implica-tions for evaluating exposure and vulnerabilityto daily stressors. In terms of exposure, giventhat nearly half of the variation in daily stres-sors can be attributed to previous acute orongoing chronic stressors (Wheaton 1994), ourresults are compelling because they suggestthe possibility of an intervening factor thatmitigates the effects of chronic stressors ondaily life. It is possible that cumulative socioe-conomic disadvantage may desensitize lowerstatus individuals to specific daily events thatare reflective of hardship, and that these indi-viduals’ reports of daily stressors reflect inde-pendent, non-normative experiences. It is alsopossible, however, that the gendered nature ofdaily stressors (Almeida, Wethington, andKessler 2002) as well as variation by gender orrace in the types of chronic stressors producedby socioeconomic disadvantage (Krieger et al.1997) may mask systematic variation in expo-sure to daily stressors: Our control variablesfor gender and race may not have adequatelycaptured these complex associations. Ofcourse, this pattern of findings could also be

an artifact of measurement and the possibilitythat respondents of lower socioeconomic statusare less reflective and articulate than higherstatus respondents about all of the stressorsthey may experience in their lives. The funda-mental issue is that comprehensive assessmentof exposure to daily stressors requires consid-eration of the substantive and methodologicalfactors affecting reports of daily stressors.

The unique and shared features of dailystressors and enduring conditions also haveimplications for stressor vulnerability. Morespecifically, enduring stressors may depletephysical or social resources for coping withnew stressors, thereby making individualsmore susceptible to the deleterious healtheffects of stressors, particularly severe stres-sors. The moderating effects of chronic stresssuch as overcrowding and poor neighborhoodquality on the daily stressor-health associationhas been borne out in previous research (Caspiet al. 1987; Lepore et al. 1991); however, addi-tional research applying the “double jeopardy”of chronic and daily stress to socioeconomicinequalities in health needs to be undertaken.Of course, the interplay of chronic and dailystressors also has implications for other expla-nations of differential vulnerability. For exam-ple, perhaps lower status individuals are morevulnerable to subjectively severe stressorsbecause exposure to chronic stressors hasdepleted already disadvantaged socialresources (Turner and Marino 1994) or exacer-bate differences in personal resources such asmastery and self-esteem (Lachman andWeaver 1998; Turner et al. 1995; Wheaton1983). Clearly, much more theoretical devel-opment and empirical work are needed to fullyunderstand the cumulative toll of daily andother forms of stress.

Finally, the design and execution of thisstudy provide strong evidence of the basicbuilding blocks for implicating stress insocioeconomic disparities in health. Results ofanalyses examining within-person covariationof stress and health (i.e., Level 1 hierarchicallinear model) clearly indicated that experienc-ing daily stressors, particularly those that aresubjectively severe, promote declines in physi-cal and mental health. In addition, results fromthe between-person analyses (i.e., Level 2 hier-archical linear model) strongly suggest thatexperiencing subjectively severe stressors pro-mote negative changes in daily health more forthose with less than a high school degree than



for those with a high school degree or collegeeducation. Although neither differential expo-sure nor differential vulnerability to dailystress completely explained socioeconomicdifferences in physical and mental health, it isclear that the stressor-health relationship can-not be considered independent of socioeco-nomic status.

Although this study provides an importantand unique perspective of the interconnectionsamong socioeconomic status, stress, andhealth, it is important to recognize its limita-tions. Perhaps the most significant limitationof this study is that data were only collectedover an eight-day period and may not have ade-quately captured overall stress exposure.However, even though individuals’ overallstress exposure may not have been fully mea-sured, the overall design and execution of theNational Study of Daily Experiences shouldhave effectively captured the experiences ofdifferent socioeconomic groups. That is, therandom assignment of start days for the dailyinterviews, the large sample and the corre-spondingly large number of interview days,and the “flight” methodology whereby individ-uals were interviewed throughout the yearshould minimize the possibility that the patternof results of this study are an artifact of the rel-atively short duration of data collection. Next,although sample means and standard devia-tions are frequently used in research, our deci-sion criteria for classifying stressors as “highseverity” is relatively arbitrary, and the use ofmultiple categorical measures increases thepossibility of Type I error because of multiplecomparisons. Thus, interpretation should focuson the overall pattern of findings rather thanindividually significant effects. Finally, it is

worthwhile to restate that reporting biases byeducation in both symptoms and daily hasslesmay contribute to the overall pattern of results.On the other hand, potential reporting biasshould be minimized by the statistical controlsfor prior-day symptoms.

Limitations notwithstanding, the resultsfrom this study extend previous research andemphasize the complexity of the interconnec-tions among socioeconomic status, stress, andhealth, and they provide guidance for futuretheory development and research. The resultsof this study strongly suggest that daily stres-sors are linked to physical and mental morbid-ity, and that lower status individuals are morevulnerable to these stressors. The distributionof stressors in this study, in contrast to others,pushes future research to further explore theconnections among acute, chronic, and dailystressors because socioeconomic differences inhealth likely reflect incremental and synergis-tic effects of different stressors across thestress universe. The results of this study alsohighlight the possibility that the meaning ofstressors, in terms of severity, may be moreimportant than stressors per se in explainingsocioeconomic inequalities and health, andthey lead to additional questions. For instance,which resources allow better educated individ-uals to better handle severe daily stressors?Similarly how can resources depleted bychronic stressors be rejuvenated to promote thecapacity to cope with the stressors of dailylife? Answers to these types of questions and amultidimensional, multilevel handling ofstress are required for a comprehensive under-standing of the role of stress in socioeconomicinequalities in health.



REFERENCES

Adler, Nancy E. et al. 1994. “Socioeconomic Statusand Health: The Challenge of the Gradient,”American Psychologist 49:15–24.

Affleck, Glenn, Alex Zautra, Howard Tennen, andStephen Armeli. 1999. “Multilevel Daily ProcessDesigns for Consulting and Clinical Psychology:A Preface for the Perplexed.” Journal ofConsulting and Clinical Psychology 67:746–54.

Aiken, Leona S. and Stephen G. West. 1991.Multiple Regression: Testing and InterpretingInteractions. Newbury Park, CA: SagePublications, Inc.

Almeida, David M. and Ronald C. Kessler. 1998.“Everyday Stressors and Gender Differences inDaily Stress.” Journal of Personality and SocialPsychology 75:670–80.

Almeida, David M., Elaine Wethington, and RonaldC. Kessler. 2002. “The Daily Inventory ofStressful Events: an Interview-based Approachfor Measuring Daily Stressors.” Assessment9:41–55.

Aneshensel, Carol S. 1992. “Social Stress: Theoryand Research.” Annual Review of Sociology18:15–38.

Baum, Andrew, J. P. Garofalo, and Ann M. Yali.1999. “Socioeconomic Status and ChronicStress. Does Stress Account for SES Effects onHealth?” Annals of the New York Academy ofSciences 896:131–44.

Bolger, Niall, Anita DeLongis, Ronald C. Kessler,and Elizabeth A. Schilling. 1989. “Effects ofDaily Stress on Negative Mood.” Journal ofPersonality and Social Psychology 57:808–18.

Bradburn, N. 1969. The Structure of PsychologicalWell-being. Chicago, IL: Aldine.

Brown, George, W and T. O. Harris. 1978. SocialOrigins of Depression: A Study of DepressiveDisorder in Women. New York: Free Press.

Bryk, Anthony S. and Stephen W. Raudenbush.1992. Hierarchical Linear Models: Applicationsand Data Analysis. Newbury Park, CA: Sage.

Burks, Nancy and Barclay Martin. 1985. “EverydayProblems and Life Change Events: OngoingVersus Acute Sources of Stress.” Journal ofHuman Stress 11:27–35.

Caspi, Avshalom, Niall Bolger, and JohnEckenrode. 1987. “Linking Person and Contextin the Daily Stress Process.” Journal ofPersonality and Social Psychology 52:184–95.

Catalano, Ralph and David Dooley. 1983. “HealthEffects of Economic Instability: A Test ofEconomic Stress Hypothesis.” Journal of Healthand Social Behavior 24:46–60.

Cohen, Sheldon and Tracy B. Herbert. 1996.“Health Psychology: Psychological Factors andPhysical Disease From the Perspective ofHuman Psychoneuroimmunology.” AnnualReview of Psychology 47:113–42.

Delongis, Anita, Susan Folkman, and Richard S.Lazarus. 1988. “The Impact of Daily Stress onHealth and Mood: Psychological and SocialResources As Mediators.” Journal of Personalityand Social Psychology 54:486–95.

Dohrenwend, Barbara S. 1973. “Social Status andStressful Life Events.” Journal of Personalityand Social Psychology 28:225–35.

Eckenrode, John and Niall Bolger. 1995. “Daily andWithin-Day Event Measurement.” Pp. 80–101 inMeasuring Stress: A Guide for Health and Social



APPENDIX A. Daily Inventory of Stressful Experiences

Stem Question

Did you have an argument or disagreement with anyonesince this time yesterday?

Since (this time/we spoke) yesterday, did anything hap-pen that you could have argued about but you decidedto let pass in order to avoid a disagreement?

Since (this time/we spoke) yesterday, did anything hap-pen at work or school (other than what you’ve alreadymentioned) that most people would consider stressful?

Since (this time/we spoke) yesterday, did anything hap-pen at home (other than what you’ve already men-tioned) that most people would consider stressful?

Many people experience discrimination on the basis ofsuch things as race, sex, or age. Did anything like thishappen to you since (this time/we spoke) yesterday?

Since (this time/we spoke) yesterday, did anything hap-pen to a close friend or relative (other than whatyou’ve already mentioned) that turned out to bestressful for you?

Did anything else happen to you since (this time/wespoke) yesterday that most people would considerstressful?

Examples of Probes for Description

Think of the most stressful disagreement or argument youhad since (this time/we spoke) yesterday. Who was thatwith?

What happened and why did you decide not to get into anargument about it?

How does this affect your job?

Have you had any problems with this in the past?

Think of the most stressful incident of this sort. What wasthe basis for the discrimination you experienced—yourrace, sex, age, or something else?

Think of the most stressful incident of this sort. Who didthis happen to?

What happened and what about it would most people con-sider stressful?

Scientists, edited by Sheldon Cohen, Ronald C.Kessler, and Lynn U. Gordon. New York: OxfordUniversity Press.

Evans, Robert G., Morris L. Barer, and Theodore R.Marmor, eds. 1994. Why Are Some PeopleHealthy and Others Not?: The Determinants ofHealth of Populations. New York: Walter deGruyter.

House, James S. and David R. Williams. 2000.“Understanding and Reducing Socioeconomicand Racial/Ethnic Disparities in Health.” Pp.81–124 in Promoting Health: InterventionsStrategies From Social and BehavioralResearch, edited by Brian D. Smedley and S. L.Syme. Washington, DC: National AcademyPress.

Jaccard, James, Robert Turrisi, and Choi K. Wan.1990. Interaction Effects in Multiple Regression.Newbury Park, CA: Sage.

Jandorf, Lina, Esther Deblinger, John M. Neale, andArthur A. Stone. 1986. “Daily Versus Major LifeEvents As Predictors of Symptom Frequency: AReplication Study.” Journal of GeneralPsychology 113:205–18.

Kanner, Allen D., James C. Coyne, CatherineSchaefer, and Richard S. Lazarus. 1981.“Comparison of Two Modes of StressMeasurement: Daily Hassles and Uplifts VersusMajor Life Events.” Journal of BehavioralMedicine 4:1–39.

Kelly, Shona, Clyde Hertzman, and Mark Daniels.1997. “Searching for the Biological Pathwaysbetween Stress and Health.” Annual Review ofPublic Health 18:437–62.

Kenny, David A., Nial Bolger, and Deborah A.Kashy. 2002. “Traditional Methods forEstimating Multilevel Models.” Pp. 1–24 inModeling Intraindividual Variability withRepeated Measures Data: Methods andApplications, edited by Debbie S. Moskowitzand Scott L. Hershberger. Mahway, NJ:Lawrence Erlbaum.

Kessler, Ronald C. 1979. “A Strategy for StudyingDifferential Vulnerability to the PsychologicalConsequences of Stress.” Journal of Health andSocial Behavior 20:100–108.

Kessler, Ronald C. and Paul D. Cleary. 1980.“Social Class and Psychological Distress.”American Sociological Review 45:463–78.

Kessler, Ronald C., Katherine A. McGonagle,Shanyang Zhao, Christopher B. Nelson, MichaelHughes, Suzann Eshleman, Hans-UlrichWittchen, and Kenneth S. Kendler. 1994.“Lifetime and 12–Month Prevalence of DSM-III-R Psychiatric Disorders in the United States:Results From the National Comorbidity Survey.”Archives of General Psychiatry 51:8–19.

Kessler, Ronald C., Richard H. Price, and CamilleB. Wortman. 1985. “Social Factors inPsychopathology: Stress, Social Support, and

Coping Processes.” Annual Review ofPsychology 36:531–72.

Kohn, Melvin L. 1972. “Class, Family, andSchizophrenia.” Social Forces 50:295–302.

Krieger, N., D. R. Williams, and N. E. Moss. 1997.“Measuring Social Class in US Public HealthResearch: Concepts, Methodologies, andGuidelines.” Annual Review of Public Health18:341–78.

Lachman, Margie E. and Suzanne L. Weaver. 1998.“The Sense of Control As a Moderator of SocialClass Differences in Health and Well-Being.”Journal of Personality and Social Psychology74:763–73.

Larsen, Randy J. and Margaret Kasimatis. 1991.“Day-to-Day Physical Symptoms: IndividualDifferences in the Occurrence, Duration, andEmotional Concomitants of Minor DailyIllnesses.” Journal of Personality 59:387–423.

Lee, Valerie E. and Anthony S. Bryk. 1989. “AMultilevel Model of the Social Distribution ofHigh School Achievement.” Sociology ofEducation 62:172–92.

Lepore, Stephen J., Gary W. Evans, and M. N.Palsane. 1991. “Social Hassles andPsychological Health in the Context of ChronicCrowding.” Journal of Health and SocialBehavior 32:357–67.

Lu, Luo. 1991. “Daily Hassles and Mental Health:A Longitudinal Study.” British Journal ofPsychology 82:441–47.

Marks, Nadine F. and Diane S. Shinberg. 1998.“Socioeconomic Status Differences in HormoneTherapy.” American Journal of Epidemiology148:581–93.

Marmot, Michael G., Rebecca Fuhrer, Susan L.Ettner, Nadine F. Marks, Larry L. Bumpass, andCarol D. Ryff. 1998. “Contribution ofPsychosocial Factors to SocioeconomicDifferences in Health.” The Milbank Quarterly76:403–40.

Marmot, Michael G, Carol D. Ryff, Larry L.Bumpass, Martin Shipley, and Nadine F. Marks.1997. “Social Inequalities in Health: NextQuestions and Converging Evidence.” SocialScience and Medicine 44:901–10.

McEwen, Bruce S. 1998. “Protective and DamagingEffects of Stress Mediators.” New EnglandJournal of Medicine 338:171–79.

Mroczek, Daniel K. and Christian M. Kolarz. 1998.“The Effect of Age on Positive and NegativeAffect: A Developmental Perspective onHappiness.” Journal of Personality and SocialPsychology 75:1333–49.

Pearlin, Leonard I. 1989. “The Sociological Studyof Stress.” Journal of Health and SocialBehavior 30:241–56.

Radloff, Lenore S. 1977. “The CES-D Scale: ASelf-Report Depression Scale for Research inthe General Population.” Applied PsychologicalMeasurement 1:385–401.



Joseph G. Grzywacz is an Assistant Professor of Family and Community Medicine with Wake ForestUniversity School of Medicine. His research focuses on issues surrounding the individual, psychosocial,and contextual factors related to health and health behaviors. Primary areas of interest include the physicaland mental health effects of socioeconomic status, employment adequacy, and the integration of work andfamily.

David M.Almeida is an Associate Professor of Human Development and Family Studies with PennsylvaniaState University. He is the Principle Investigator for the National Study of Daily Experiences, and hisresearch interests center on the general question of how daily experiences within the family and other socialcontexts, such as work and leisure, influence individual health and well-being.

Shevaun D. Neupert is a Postdoctoral fellow in the Department of Psychology with Brandeis University.Her research focuses on the daily well-being of older adults with respect to stressors and cognition.

Susan L. Ettner is Associate Professor in the Division of General Internal Medicine and Health ServicesResearch in the University of California at Los Angeles Department of Medicine and in the Department ofHealth Services in the UCLA School of Public Health. Her research interests include reciprocity in the rela-tionship between health and labor market outcomes, mental health and substance abuse services, insurancemarkets and managed care, chronic disability, and post-acute and long-term care.



Robinson, W. S. 1950. “Ecological Correlations andthe Behavior of Individuals.” AmericanSociological Review 15:351–57.

Schwarz, Norbert. 1999. “Self-Reports: How theQuestions Shape the Answers.” AmericanPsychologist 54:93–105.

Taylor, Janet A. 1953. “A Personality Scale ofManifest Anxiety.” Journal of Abnormal andSocial Psychology 48:285–90.

Tennen, Howard, Glenn Affleck, Stephen Armeli,and Margaret A. Carney. 2000. “A Daily ProcessApproach to Coping: Linking Theory, Research,and Practice.” American Psychologist55:626–36.

Theorell, Tores G. M. 1982. “Review of Researchon Life Events and Cardiovascular Illness.”Advance Cardiology 29:140–47.

Thoits, Peggy A. 1983. “Dimensions of Life EventsThat Influence Psychological Distress: AnEvaluation and Synthesis of the Literature.” Pp.33–103 in Psychological Stress: Trends inTheory and Research, edited by Howard B.Kaplan. New York: Academic Press.

Turner, R. J. and Donald A. Lloyd. 1999. “TheStress Process and the Social Distribution ofDepression.” Journal of Health and SocialBehavior 40:374–404.

Turner, R. J. and Franco Marino. 1994. “SocialSupport and Social Structure: A DescriptiveEpidemiology.” Journal of Health and SocialBehavior 35:193–212.

Turner, R. J. and Blair Wheaton. 1995. “ChecklistMeasurement of Stressful Life Events.” Pp.29–58 in Measuring Stress: A Guide for Health

and Social Scientists, edited by Sheldon Cohen,Ronald C. Kessler, and Lynn U. Gordon. NewYork: Oxford University Press.

Turner, R. J., Blair Wheaton, and Donald A. Lloyd.1995. “The Epidemiology of Social Stress.”American Sociological Review 60:104–25.

Weinberger, Morris, Sharon L. Hiner, and WilliamM. Tierney. 1987. “In Support of Hassles As aMeasure of Stress in Predicting HealthOutcomes.” Journal of Behavioral Medicine10:19–31.

Wethington, Elaine, George W. Brown, and RonaldC. Kessler. 1995. “Interview Measurement ofStressful Life Events.” Pp. 59–79 in MeasuringStress: A Guide for Health and Social Scientists,edited by Sheldon Cohen, Ronald C. Kessler,and Lynn U. Gordon. New York: OxfordUniversity Press.

Wheaton, Blair. 1983. “Stress, Personal CopingResources, and Psychiatric Symptoms: AnInvestigation of Interactive Models.” Journal ofHealth and Social Behavior 24:208–29.

———. 1994. “Sampling the Stress Universe.” Pp.77–114 in Stress and Mental Health:Contemporary Issues and Prospects for theFuture, edited by William R. Avison and Ian H.Gotlib. New York: Plenum Press.

Wilkinson, Richard G. 1996. Unhealthy Societies:The Afflictions of Inequality. New York:Routledge.

Winship, Christopher and Larry Radbill. 1994.“Sampling Weights and Regression Analysis.”Sociological Methods and Research 23:230–63.

socioeconomic status and health: a micro-level analysis of

Documents