Perceived Socioeconomic Status: A New Type of Identity That Influences Adolescents’ Self-Rated Health

A

K

twdl

Hi

1d

Original article

Perceived Socioeconomic Status: A New Type of Identity ThatInfluences Adolescents’ Self-Rated Health

Elizabeth Goodman, M.D.,a,* Bin Huang, Ph.D.,b Tara Schafer-Kalkhoff, M.A.,b

and Nancy E. Adler, Ph.D.caTufts-New England Medical Center and the Floating Hospital for Children, Tufts University School of Medicine, Boston, Massachusetts

bCincinnati Children’s Hospital Medical Center, Cincinnati, OhiocHealth Psychology Program, Department of Psychiatry, University of California–San Francisco, San Francisco, California

Manuscript received February 2, 2007; manuscript accepted May 16, 2007

bstract Purpose: The cognitive, social, and biological transitions of adolescence suggest that subjectiveperceptions of social position based on the socioeconomic hierarchy may undergo importantchanges during this period; yet how such perceptions develop is poorly understood, and no studieshave assessed whether changes in such perceptions influence adolescents’ health. This studydescribes adolescents’ subjective perceptions of familial socioeconomic status (SSS), how SSSchanges over time, and how age, race, and objective socioeconomic status (SES) indicators influenceSSS. In addition, the study determines whether SSS independently influences adolescents’ self-ratedhealth, an important predictor of morbidity and health service use.Methods: A total of 1179 non-Hispanic black and white baseline 7–12th graders from a Midwest-ern public school district completed a validated, teen-specific measure of SSS annually for 4consecutive years. A parent provided information on SES. Markov modeling assessed transitions inSSS over time.Results: SSS declined with age (p � .001) and stabilized among older teens. In addition to age,SES and race, but not gender, were significant correlates of SSS, but the relationships between thesefactors were complex. In cross-sectional and longitudinal analyses, black teens from families withlow parent education had higher SSS than white teens from similarly educated families, whereaswhite teens from highly educated families had higher SSS than black teens from highly educatedfamilies. Lower SSS and changes in SSS predicted poor self-rated health even when adjusting forrace and objective SES measures.Conclusion: Subjective evaluations of socioeconomic status predict adolescents’ global healthratings even when adjusting for the sociodemographic factors that shape them. © 2007 Society forAdolescent Medicine. All rights reserved.

Journal of Adolescent Health 41 (2007) 479–487

eywords: Socioeconomic status; Self-rated health; Health disparities

mppsgpth

Lower socioeconomic status (SES) is a powerful predic-or of poor health. The inverse association between SES andorse health outcomes has been documented throughout theeveloped world and across every age group. However, howower SES gets “under the skin” to affect physical and

*Address correspondence to: Elizabeth Goodman, M.D., The Floatingospital for Children at Tufts-New England Medical Center, 750 Wash-

ngton Street, Box 351, Boston, MA 02111.
iE-mail address: [email protected]
054-139X/07/$ – see front matter © 2007 Society for Adolescent Medicine. Alloi:10.1016/j.jadohealth.2007.05.020

ental health remains largely unexplained. Although com-elling structural and behavioral explanations have beenroffered, none fully explain why health disparities remaino prevalent today. The persistence of these disparities sug-ests that, in addition to behavioral and structural factors,sychological and physiological processes are also active inhe development and maintenance of social inequalities inealth [1–3].

In animal models used to study socioeconomic inequal-
ties in health, rank within the social hierarchy is considered
rights reserved.

aaisluispeasaed

ebstcthsf

pvbeTattueaTict

Tcip1foidipwo

saewsdTwddMbhwccslpubehcrhwsa

M

S

(saDdcAtlwespsawbi1

480 E. Goodman et al. / Journal of Adolescent Health 41 (2007) 479–487

nalogous to SES. Animal studies suggest that lower ranknd the stress related to social subordination affect themmune, adrenocortical, cardiovascular, and reproductiveystems, leading to increased morbidity and mortality inower status animals [1]. These studies, which have beensed to support models for how SES creates health dispar-ties among human beings [4], show both intra- and inter-pecies variation in the relationships between rank, the ex-erience of stress, and the physiological alterations that suchxperiences engender [1]. Traditional measures of SES suchs income, education, and occupation do not account foruch variation in personal experience of rank. Studiesmong human beings, therefore, assume that the personalxperience of SES is equivalent for each member of a groupefined by similar socioeconomic background.

Equality in the personal experience of objective socio-conomic status is assumed, in part, because how humaneings experience and understand their socioeconomic po-ition is not well understood. Prior work in this area tendedo focus on social class identification [5,6] rather than per-eived socioeconomic status, because the idea that subjec-ive perceptions of socioeconomic status influence humanealth is relatively new and measures to assess perceivedocioeconomic status have only been developed in the pastew years [7,8].

To date, studies that use these measures indicate thaterceived socioeconomic status among adults is related to aariety of health-related outcomes and that the relationshipsetween perceived socioeconomic status and health persistsven after adjustment for objective SES measures [1,9–13].he relationship between perceived socioeconomic statusnd health appears to be less consistent among adolescentshan in adults [8,14,15]. This inconsistency suggests thathere may be developmental differences in subjective eval-ations of socioeconomic status that influence health. How-ver, because this field is a nascent one, little is knownbout how perceptions of SES develop or what shapes them.o date, no longitudinal study exists that could provide

nsight into the development of adolescents’ perceived so-ioeconomic status and how development of such percep-ions relates to health.

The purpose of the current study was to address this gap.he study has three aims. The first was to describe per-eived socioeconomic status in a racially and socioeconom-cally diverse cohort of youth and to assess transitions inerceived socioeconomic status over three successive-year periods. We hypothesized that prior perception ofamily socioeconomic status would be a major determinantf current perceptions. The second aim was to explore thentersection of race and objective SES in relation to theevelopment of perceived socioeconomic status. Because its associated with discrimination and disadvantage, we hy-othesized that minority race/ethnicity would be associatedith lower perceived socioeconomic status independent of
bjective SES measures and prior perceived socioeconomic s
tatus. Further, we hypothesized that those who were dis-dvantaged along multiple dimensions (i.e., minority race/thnicity, economically poor, and poorly educated parents)ould have significantly lower perceived socioeconomic

tatus than would be expected given the effect of eachomain in and of itself on perceived socioeconomic status.his hypothesis is based on the double jeopardy hypothesis,hich suggests that the total combined effect of beingisadvantaged along two status dimensions is greater thanisadvantage in either dimension alone [16] and the work ofcLeod and Owens, who noted that social evaluations may

e one of the mechanisms underlying the double-jeopardyypothesis [17]. The third aim of the study was to determinehether and how perceived socioeconomic status and

hanges in these perceptions over time influenced adoles-ents’ self-rated health. Such global health ratings synthe-ize multiple domains of health, including physical, psycho-ogical and social wellbeing [18,19]. They are powerfulredictors of mortality, future morbidity, and health servicese [20–22]. Among adults, poorer self-rated health haseen associated with lower SES and lower perceived socio-conomic status [7,12,23–26]. Although poor self-ratedealth has been associated with lower SES among adoles-ents [19,27,28], the relationship between SSS and self-ated health in adolescents has not been evaluated. Weypothesized that lower perceived socioeconomic statusould be associated with greater odds of having poor

elf-rated health in both cross sectional and longitudinalnalyses.

ethods

tudy sample and procedures

This study uses data from the Princeton School DistrictPSD) Study, a prospective study from a single Midwesternuburban public school system with one junior high schoolnd one senior high school [14]. The Princeton City Schoolistrict is a well-defined geographic area that includes sixiverse, independent, incorporated communities plus unin-orporated areas of Hamilton and Butler Counties (Ohio).pproval for the cohort study was received from the Insti-

utional Review Boards at the participating University andocal Children’s Hospital. Students in grades 7–12 whoere enrolled at the start of the 2001–2002 school year were

ligible for participation, which included completion of aurvey annually for 4 consecutive years. Surveys were com-leted privately in the school setting or through the mail iftudents had graduated and could not come to the school forstudy appointment [29]. Because the school populationas 47.3 % non-Hispanic white and 47.2% non-Hispaniclack, only non-Hispanic black and white students werencluded in these analyses. Of the 2116 eligible students,367 (64.6%) completed at least two consecutive yearly
urveys assessing SSS. 86.2% of these students (n � l179)

hcm(bqsetuwfmwcoiwpw.e

M

Sasettltgcemtp$

Ptspocttmafhfsfi

Dlg

Ssyma

A

cyPchipocMvwitfae

A

eoaSafoeirudaMlse

Tb

481E. Goodman et al. / Journal of Adolescent Health 41 (2007) 479–487

ad a parent provide information on SES. These 1179 teensomprise the study cohort. The cohort included 267 blackales (22.6%), 274 black females (23.2%), 321 white males

27.2%), and 317 white females (26.9%). Mean age ataseline was 15.1 � 1.6 years (range 12.2–19.3 years). Oneuarter (25.4%) had a sibling in the study, and 5.9% had twoiblings in the cohort. There were no age or gender differ-nces in participation. However, despite the fact that 46% ofhe study cohort was non-Hispanic black, these youth werender-represented relative to the target population, whichas 50% non-Hispanic black (p �.001). Mean length of

ollow up was 10.1 months between waves 1 and 2, 11.1onths between waves 2 and 3, and 11.6 months betweenaves 3 and 4. Most (78.1%) completed all 4 waves, 14.9%

ompleted three of the four waves, and 7.0% completednly the initial two waves. There were no gender differencesn the number of waves completed, although slightly morehites completed all four waves (83%) than blacks (81%,� .003). White students came from families with parentsith more education and greater household income (p �

0001). There were no gender or age differences in parentducation.

easures

ES. A parent provided information on parental educationnd total household income before taxes. Self and currentpouse/partner education was reported in nine ordered cat-gories ranging from never attended school to professionalraining beyond a 4-year college or university. For analyses,hese categories were condensed into four: high school oress (N � 265, 22.5%), some college or technical/vocationalraining beyond high school (N � 337, 28.6%), collegeraduation (N � 322, 27.3%), professional training beyondollege (N � 255, 21.6%). The highest level of parentalducation was used in analyses as prior work has shown thatother’s and father’s education function similarly in regard

o self-rated health [28]. Total household income was re-orted in nine ordered categories ranging from less than5,000 to $100,000 or more.

erceived socioeconomic status. The Subjective Social Sta-us (SSS) Scale, a validated, youth-specific measure, tapsubjective perceptions of the family’s SES by using a 10-oint self-anchoring scale [8]. The measure and its devel-pment have been described in detail elsewhere [8]. Basi-ally, subjects are shown a drawing of a ladder that is saido represent American society. Parallel language anchors athe top and bottom refer to traditional SES measures ofoney, schooling, and occupational prestige. Adolescents

re asked to chose the ladder rung that best represents theiramily’s place in American society. Higher scores indicateigher perceived family SES. Adolescents are asked to rankamily SES rather than their own individual SES, becauseocioeconomic status in childhood and adolescence is de-
ned by that of the family. e
emographic covariates. Date of birth (used in age calcu-ation), parent-identified race/ethnicity of the student, andender were available from school demographic data.

elf-rated health. A single five-point Likert scale item as-essed global perceptions of health: “In general, how wouldou rate your health?” Response categories were dichoto-ized into fair–poor and good–excellent categories for

nalysis, as has been done in prior reports [28].

nalytic strategy

Bivariate analyses used nonparametric techniques be-ause SSS was not normally distributed. Multivariable anal-ses were run using generalized estimating equations (SASroc GENMOD) to account for siblings in the cohort. Be-ause 14.1% (n � 166) were missing information on house-old income, multiple imputation was used to impute miss-ng income values for the multivariable modeling. Weerformed multiple imputation based on the method devel-ped by Rubin [30]. Reported results are based on theombined results of five separate imputations and SASIANALYZE software [31,32]. Data on other independent

ariables (age, gender, parent education, and race/ethnicity)ere complete for all subjects, and therefore did not require

mputation. Testing for all two-way and three-way interac-ions and examination for non-linear effects were per-ormed. Main effects were entered in testing for interactionsnd were retained in the final models. Results were consid-red significant at values of p � .05.

nalyses of SSS

First, descriptive analyses of baseline SSS, includingvaluation of the relationships of age, race, gender, andbjective SES to SSS were performed. Second, longitudinalnalyses assessed how these factors influence transitions inSS over time. In analyses of changes in SSS over time, welso needed to account for the influence of prior SSS onuture perceptions, as well as the possibility that the effectf certain factors on SSS may be different conditional on thearlier ranking. Failure to take into account the earlier rank-ng and its potential interactive effects could lead to incor-ect interpretations of the factors that influence SSS. Wesed Markov transitional models, which are specificallyesigned to address these issues of prior perceptions, toccount for the effect of the prior year’s ranking. Thearkov transitional model is described in the equation be-

ow, where Yi(t) is the current value of SSS at time t forubject i, and Yi(t-1) is the previous value of SSS at thearlier time point.

Yi(t 1)Xi(t 1) Yi(t 1) Xi(t 1)

Yi(t)

he transition in SSS from one point in time to another maye different from individual to individual, and can be mod-
led by the factors Xi(t) and their interaction with Yi(t).

Whttttwpis

A

waShswctus

R

Do

w�twsFh

cpmdf

C

saa..jbeispirt$ffilw3hpa

da

TB

S

W

W

W

W

S


hen the interaction is significant, it suggests Xi(t) mayave different effect conditional on the previous SSS forhis subject. Because of the complexity of the multiplewo-way interactions in the linear Markov models, cumula-ive logit models using SAS Proc GENMOD were also runreating SSS as an ordinal variable. Results of these modelsere then used to derive a set of figures to simplify inter-retation by graphically representing the probability that anndividual’s SSS would move up, move down, or stay theame over time.

nalyses of self-rated health

Logistic regression models were run for analyses inhich fair-poor self-rated health was the dependent vari-

ble. The initial models assessed the relationship of baselineSS and objective SES to baseline fair-poor self-ratedealth status adjusting for age, gender, and race/ethnicity. Aecond model then assessed if changes in SSS betweenaves 1 and 4 influenced self-rated health at wave 4. Be-

ause SSS and self-rated health are both subjective evalua-ions, this second model, which accounts for baseline eval-ations of both SSS and self-rated health, provides a moretringent test of the relationship of SSS to self-rated health.

esults

escription of SSS across the four wavesf data collection

Mean SSS (SD) across the four waves was as follows;ave 1, � � 6.65 (1.34); wave 2 � � 6.65 (1.38); wave 3� 6.66(1.36); and wave 4 � � 6.60 (1.35). Although

hese mean values are nearly identical, there was substantialithin-person variation in SSS (F � 3.78, p � .01) and a

ignificant linear trend across the four waves (linear contrast� 6.94, p � .009). Between waves 1 and 2, only 35.0%

able 1ivariate relationships among SSS, race, SES, and gender in the study

SS Race Parental education

Black White p �HS �HS, �college College

ave 1� 6.61 6.68 .10 6.32 6.45 6.70SD 1.42 1.28 1.53 1.26 1.37ave 2� 6.57 6.72 .03 6.35 6.42 6.70SD 1.47 1.29 1.49 1.36 1.41ave 3� 6.42 6.70 �.001 6.21 6.17 6.66SD 1.29 1.37 1.43 1.25 1.30ave 4� 6.37 6.79 �.001 6.19 6.28 6.60SD 1.34 1.32 1.41 1.39 1.28

Note: � � mean; HS � high school; Prof � professional; SES � objecD � standard deviation.

ad identical SSS at both waves and the within-person S

hange in SSS ranged from �7 to �8. Although the within-erson variation was statistically significant, most com-only, adolescents moved either up one rank (21.9%) or

own one rank (20.2%) if they changed their perceivedamilial socioeconomic position over the course of a year.

orrelates of SSS at each wave

Table 1 presents a description of the bivariate relation-hips between SSS, race, SES, and gender. At baseline,dolescents’ SSS was correlated with both parent educationnd household income (Spearman’s � for SSS-education �24, p � .001, Spearman’s � for SSS-income � .35, p �001). However, the correlations between SSS and the ob-ective SES measures was not as strong as the correlationetween the objective SES measures themselves (for parentducation–income, Spearman’s � � .60, p �.001). Thisndicates that disjuncture between SSS and SES exists forome. For example, 17% of students from families whosearents had high school or less education ranked their fam-lies in the upper quintile of the SSS scale. The high SSSating was not due to high household income, as 48% ofhose teens came from households with incomes less than25,000. Similarly, 6.7% of teens from highly educatedamilies had a baseline SSS in the bottom half despite theact that most (58%) came from families with householdncomes of $50,000 or higher. The teens with high SSS andow objective SES were largely (82%) black, whereas thoseith low SSS and high SES were mostly white (65% white,5% black). Overall, lower parent education, lower house-old income, and black race were associated with lowererceived social status. Gender was not associated with SSSt any wave.

Results of multivariable analyses to determine indepen-ent correlates of SSS are presented in Table 2. Older teensnd those from households with lower income had lower

Income Gender

degree p �$50K �$50K p Male Female p

�.001 6.24 6.95 �.001 6.70 6.60 .241.38 1.33 1.37 1.32

�.001 6.28 6.96 �.001 6.70 6.61 .321.41 1.34

�.001 6.08 6.94 �.001 6.66 6.50 .021.40 1.37 1.30 1.40

�.001 6.11 6.22 �.001 6.61 6.60 .941.34 1.36 1.28 1.41

ioeconomic status; SSS � subjective perception of socioeconomic status;

Prof

7.211.19

7.221.25

7.291.24

7.321.22

tive soc

SS. In addition, although the main effects of education and

rpis5lpadS1

L

twaadtpap

momwfSyHch

rpcftwFtwatp

Pt

trmfiLiulAtst

TLtS

PAPFBE

P

B

I

p

TCD

ABFP

R

I

0b

s


ace were not significant, there was a significant race byarent education interaction. This interaction was reflectedn the raw baseline data; white teens with parents with highchool education or less had lower SSS at baseline (� �.97, SD � 1.34) than black teens with parents at the sameevel of education (� � 6.56, SD � 1.62). However, asarent education increased, the reverse was found. Lookingt families with at least one parent with a professionalegree, black teens had lower baseline SSS (� � 7.00,D � 1.14) compared to white teens (� � 7.28, SD �.28).

ongitudinal changes in SSS

Correlations between yearly SSS rankings increased overhe three follow-up periods from .48 (95% CI � .44, .53) foraves 1 and 2 to .59 (95% CI .55, .63) for waves 2 and 3

nd to .64 (95% CI � .50, .67) for waves 3 and 4 (trendnalysis p � .001). Since there is no reason to suspectifferences in changes in parent SES in one versus anotherime period, this trend, which reflects stabilization in SSS,rovides evidence that developmental processes within andolescent influence perceptions of familial socioeconomicosition within American society.

Table 3 presents results of the multivariable Markovodeling. The prior year’s SSS had the strongest influence

n the next year’s SSS both through direct effects andultiple interactive effects. Three two-way interactionsere found. The interaction of prior SSS with age provides

urther evidence for developmental influences. When priorSS was low, the effect of age remained negative, withounger teens transitioning to higher SSS than older teens.owever, when prior SSS was high, the effect of age be-

ame positive with older teens tending to transition to

able 2orrelates of baseline SSS in 1179 students from the Princeton Schoolistrict (PSD) Study

Parameter estimate SE p Value

ge (years) �.07 .02 .004lack �.05 .18 .80emale �.11 .07 .13arent education�HS .08 .20 .68�HS, �college �.18 .19 .33college �.11 .19 .56

ace � educationBlack � �HS .83 .24 .006Black � �HS,�college .16 .23 .50Black � college .25 .23 .28

ncome .0086 .0011 �.0001

Note: Race: 0 � non-Hispanic black, 1 � non-Hispanic white. Gender:� female, 1 � male. Income range is from 1–9. Professional degree

eyond college � reference category for parent education.HS � high school; SE � standard error; SSS � subjective perception of

ocioeconomic status.

igher SSS than younger teens. The interactions between s

ace, parent education, and prior SSS were complex. Therior SSS by education and education by race interactionsan together be interpreted to indicate that, for teens fromamilies with low parent education, a black teen transitionedo higher SSS than a white teen regardless of prior SSS. Thisas not the case for teens from highly educated families.or these teens, when prior SSS was low, a black teen

ransitioned to higher SSS than a white teen. However,hen prior SSS was high, the situation was reversed; black

dolescents transitioned to lower SSS than white teens. Ahree-way interaction between race, parent education, andrior SSS was not statistically significant.

robability of movement up, movement down, or stayinghe same

Because of the complexity of the interactions present inhe Markov models, we performed analyses to graphicallyepresent the probability that SSS would stay the same,ove down, or move up given different levels of prior SSS

or a 15 year old adolescent with annual household incomen the moderate range (approximately $58,000) (Figure 1).ooking across the three panels reveals that the vast major-

ty of those with prior SSS of 4 or less would probably movep. As prior SSS becomes more moderate (5–7), the like-ihood among the three possibilities is much more similar.t a high prior SSS, most teens would move down. Al-

hough these patterns may reflect regression to the mean toome extent, the differences in the likelihood of change foreens from different backgrounds suggests that other factors

able 3inear Markov model of SSS of transitions in SSS over Waves 2

hrough 4 in 1179 students from the Princeton School District (PSD)tudy

Parameter estimate SE p

rior SSS .32 .15 .03ge (years) �.11 .06 .050rior SSS � age .02 .008 .054emale �.05 .04 .22lack �.26 .10 .008ducation�HS 1.06 .32 0.001�HS,�college .85 .31 .007College .53 .33 .102

rior SSS � educationPrior SSS � �HS �.15 .04 �.001Prior SSS � �HS,�college �.13 .04 .005Prior SSS � college �.07 .05 .108

lack � educationBlack � �HS .40 .13 .002Black � �HS,�college .39 .12 0.002Black � college .23 .12 .06

ncome .004 .0006 �.001

Note: Professional degree beyond college � reference category forarent education.HS � high school; SE � standard error; SSS � subjective perception of

ocioeconomic status.

ieWfmHtml

R

st.

BI(aaslntt

dhn

F1Tl


nfluence SSS over time. For example, the significant parentducation by prior SSS interaction is apparent in this figure.hen prior SSS was low, those from high parent educated

amilies were more likely to move down and less likely toove up than youth from low parent educated families.owever, when prior SSS was high, the reverse was true:

hose from high parent educated families were less likely toove down and more likely to move up than youth from

ow parent educated families with similar prior SSS.

elationship of SSS to self-rated health

Self-rated health was correlated with both objective andubjective measures of socioeconomic position. However,he correlations were strongest for SSS (Spearman’s � SSS �

igure 1. The influence of race, parent education and prior subjective per5-year-old adolescent from household with moderate income will: (A) sthe y-axis represents probability for the current year’s SSS. Light gray �

ess parent. Dashed � white ethnicity; solid � black ethnicity.

23, � education � .18, � income � .14, p � .001 for all). h

aseline prevalence of fair-poor self-rated health was 7.1%.n bivariate analyses, lower SSS (p � .001), female genderp � .02), and lower parent education (p � .03) weressociated with baseline fair-poor self-rated health. Race,ge, and income were not associated with baseline fair-poorelf-rated health in bivariate analyses. Although the preva-ence of fair –poor self-rated health at follow up (6.7%) wasearly equal that at baseline, only one quarter (24.6%) ofhose with baseline fair-poor self-rated health maintainedhat same level of perceived health status 3 years later.

Table 4 presents results of multivariable modeling toetermine independent predictors of fair–poor perceivedealth status. At baseline, only gender and SSS were sig-ificant independent correlates of fair–poor self-rated

of familial socioeconomic status (SSS) on the probability that SSS of aame, (B) move down, or (C) move up. Prior SSS is noted on the x-axis.with a professional degree; black � parent with high school education or

ceptionsay the sparent

ealth. Female gender and lower SSS were associated with

isaaSpcefc

D

siafcbsiSsbltnhSpod

a

etbqitonf

wststseemSimhamofpmseatfi

TRc

BBFP

IBCBT

� sub


ncreased odds of fair–poor self-rated health. Neither mea-ure of objective socioeconomic status was significantlyssociated with fair–poor self-rated health when SSS wasccounted for. In longitudinal analyses, both lower baselineSS and decrease in SSS were associated with subsequentoor self-rated health, even after adjusting for baseline per-eived health status and other covariates. Lower parentducation was also associated with fair–poor health status atollow-up, which suggests that the influence of parent edu-ation on health perceptions may strengthen with age.

iscussion

In this study, adolescents’ perceptions of their family’socioeconomic status was shaped by multiple social factors,ncluding age, race, and objective SES. As youth developed,

complex pattern of relationships emerged between theseactors that influenced how perceptions of family SEShanged over time. These data suggest that internalizedeliefs of family socioeconomic position could be under-tood as a type of identity akin to gender and racial/ethnicdentity. Further, the robust relationship between perceivedES and self-rated health we demonstrate here suggest thatocioeconomic status identity, which is easily assessed, maye a more sensitive predictor of health status among ado-escents than objective measures of SES. Data suggest thathis is the case with adults, for whom perceived socioeco-omic status has also been shown to be a better predictor ofealth status and change in status over time than objectiveES measures [13]. However, further research assessing theerformance of perceived SES in relation to specific healthutcomes among adolescents still needs to be performed toetermine if this is the case.

Although subjective perceptions of family SES were

able 4elationship of demographic characteristics, objective socioeconomic statross-sectional and longitudinal analyses over 3 years in the Princeton Sc

Baseline (Wave 1) Fair–Poor SRH

Parameter estimate SE

aseline age (years) .114 .086lack �.106 .300emale .890 .392arent education�HS .771 .544�HS, �college .232 .509College .484 .447

ncome .001 .005aseline SSS �.238 .113hange in SSSaseline fair–poor SRHime to follow up (months)

Note: Income range is from 1–9. Professional degree beyond college �HS � high school; SE � standard error; SRH � self-rated health; SSS

ffected by objective SES measures, the influence of these l

xternal measures on internalized perceptions was not iden-ical for all youth, nor was there a simple linear relationshipetween SES and perceived SES. These findings call intouestion the assumption underlying many studies of socialnequalities in health—namely, that each person’s evalua-ion of socioeconomic status mirrors the external measuresf SES associated with that individual. This assumptioneeds to be re-examined and approached with more nuanceor adolescents.

The idea that external markers of socioeconomic statusould be internalized and create the subjective within per-

on experience of socioeconomic position is part of a longradition in research on self-concept (the “looking glasself”) [33]. Based on this tradition, we had hypothesizedhat characteristics associated with social disadvantage,uch as minority race/ethnicity, would decrease SSS andven enhance the influence of external SES. As opposed tonhancing the influence of SES, our data suggest that raceay be one factor that could create disjuncture betweenES and SSS. Multiple strands of evidence support this

dea. Those whose SSS exceeded their objective SES wereostly black, whereas those whose objective SES was

igher than their perceived status were largely white. Inddition, race by education interactions were present in allodels. Possible explanations for these findings are that

bjective SES may be a less powerful a determinant of SSSor black youth or that the upper SES black youth arearticularly sensitive to the effects of discrimination andarginalization they experience because of their minority

tatus. Higher parental education may bring with it greaterxposure to discrimination and/or awareness of racismmong higher SSS black adolescents, which did not affectheir white counterparts. Whatever the explanation, thesendings highlight the need for careful theoretical and ana-

subjective perception of socioeconomic status to self-rated health instrict (PSD) Study

Follow-up (Wave 4) Fair–Poor SRH

Value Parameter estimate SE p Value

.16 �.063 .093 .473

.72 .36 �.29 .22

.001 .649 .288 .024

.15 1.919 .708 .007

.65 1.604 .679 .018

.28 1.157 .668 .083

.88 .003 .004 .488

.036 �.34 .12 .007�.26 .11 .0241.507 .357 �.001.032 .064 �.001

nce category for parent education.jective socioeconomic status.

us, andhool Di

p

�

refere

ytic treatment of race and socioeconomic status [34]. The

ipcp

ptsrqbeopaftpclsssawsdlscnth

ttrdsudwcihjwwcctsahfl

nhcais

satdtnmgnshmiiS

A

gddctt

R


ntersections between these two factors which have social,olitical, historical, and psychological dimensions may beritical to understanding how society influences health andsychological well-being.

Findings related to the development of self-esteem mayrovide a good model for how to think about the intersec-ions of race and objective SES in the development ofocioeconomic status identity among adolescents. In theesearch on development of self-esteem, investigators haveuestioned the validity of the looking glass self-conceptecause many studies showed that blacks have higher self-steem than whites, despite the fact that blacks are membersf a disadvantaged and marginalized group [35]. Threeotential mechanisms have been proffered to explain thesepparently paradoxical findings: 1) attribution of negativeeedback to prejudice toward one’s social group, 2) selec-ive devaluation of factors on which one’s group fareoorly, and 3) use of more in-group versus out-group socialomparisons (social identity theory). These ideas—particu-arly the last—are relevant and applicable to the currenttudy [36]. Research on segregation has shown that blackchool children have higher self-esteem if they attend aegregated school, probably because in-group comparisonsre more likely than from those in desegregated schoolsho may compare themselves to whites [37]. Although the

chool district these teens were drawn from is an integratedistrict, the high proportion of black students may create aarge enough cohort for in-group comparisons among blacktudents to occur in the district schools. Although socialomparisons and social identity theory [38] suggest expla-ations for our findings, further research will be required toest if these constructs account for the patterns we describeere.

There are some important limitations to note. Becausehe study had slight over-representation of white students,he possibility that selection bias may have influenced theesults cannot be ruled out. In addition, measures of variousomains of social status and health status were based oningle items. However, these single items have been widelysed in studies throughout the developed world to assesseterminants of health. Only parent-identified race/ethnicityas available, and perceived social status is a relatively new

onstruct. However, the perceived SES measure was a val-dated, adolescent-specific instrument and the self-ratedealth measure has been extensively used. In addition, ob-ective SES was measured at baseline only, not yearly, asas SSS. Although it is unlikely that parental educationould change much in these 3 years, there may have been

hanges to household income which may have influencedhanges in SSS. There were no major economic changes inhe community during the period of data collection, soystematic change in either direction is unlikely. In addition,lthough there may have been changes at the individualousehold level, it is not clear how cognizant teens are of
uctuations in household income from year to year. As
oted in the methods section, both the SSS and self-ratedealth scales were self-report measures, which could in-rease the correlation due to report bias. However, thenalyses, which found an association between SSS, changesn SSS, and perceived health status controlled for baselineelf-rated health, thereby factoring out the joint variance.

In conclusion, these findings indicate that socioeconomictatus identity, which can be thought of as the subjectivessessment of one’s socioeconomic position, is formedhrough a complex interweaving of factors related to socialisadvantage and developmental changes. The data suggesthat this identity reflects external markers of SES but mayot be totally congruent with such measures. Some groupsay rely more heavily on relative comparisons within the

roup, rather than comparisons against a broader societalorm. In addition, these findings suggest that perceivedocioeconomic status may be a more sensitive predictor ofealth and health changes than the traditional objectiveeasures of socioeconomic position. Whether differences

n perceived status could explain why some teens are resil-ent to the well-documented detrimental effects of lowerES on health remains to be determined.

cknowledgments

This work was supported by National Institutes of Healthrants HD41527 and DK59183, the William T. Grant Foun-ation and the John D. and Catherine T. MacArthur Foun-ation. The authors thank Hirokazu Yoshikawa, Ph.D., foromments on an earlier draft of this manuscript, as well ashe students, parents, teachers, administration and staff ofhe Princeton City School district and the PSD study staff.

eferences

[1] Sapolsky RM. Social status and health in human beings and otheranimals. Annu Rev Anthropol 2004;33:393–418.

[2] Marmot M, Wilkinson RG. Psychosocial and material pathways inthe relation between income and health: a response to Lynch et al. BrMed J 2001;322:1233–6.

[3] Adler NE, Newman K. Socioeconomic disparities in health: pathwaysand policies. Inequality in education, income, and occupation exac-erbates the gaps between the health “haves” and “have-nots.” HealthAff (Millwood) 2002;21:60–76.

[4] McEwen BS. From molecules to mind: stress, individual differences,and the social environment. Ann NY Acad Sci 2001;935:42–9.

[5] Centers R. Social class identifications of American youth. J Personal1950;18:290–302.

[6] Simmons R, Rosenberg M. Functions of children’s perceptions of thestratification system. Am Sociol Rev 1971;3:235–49.

[7] Adler NE, Epel ES, Castellazzo G, et al. Relationship of subjectiveand objective social status with psychological and physiological func-tioning: preliminary data in healthy white women. Health Psychol2000;19:586–592.

[8] Goodman E, Adler NE, Kawachi I, et al. Adolescents’ perceptions ofsocial status: development and evaluation of a new indicator. Pedi-
atrics 2001;108:e31.

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[


[9] Adler N, Snibbe AC. The role of psychosocial processes in explainingthe gradient between socioeconomic status and health. Curr DirectPsychol Sci 2003;12:119–23.

10] Ostrove JM, Adler NE, Kuppermann M, et al. Objective and subjec-tive assessments of socioeconomic status and their relationship toself-rated health in an ethnically diverse sample of pregnant women[in process citation]. Health Psychol 2000;19:613–8.

11] Hu P, Adler NE, Goldman N, et al. Relationship between subjectivesocial status and measures of health in older Taiwanese persons. J AmGeriatr Soc 2005;53:483–8.

12] Singh-Manoux A, Adler NE, Marmot MG. Subjective social status:its determinants and its association with measures of ill-health in theWhitehall II study. Soc Sci Med 2003;56:1321–33.

13] Singh-Manoux A, Marmot MG, Adler NE. Does subjective socialstatus predict health and change in health status better than objectivestatus? Psychosom Med 2005;67:855–61.

14] Goodman E, Adler NE, Daniels SR, et al. Impact of objective andsubjective social status on obesity in a biracial cohort of adolescents.Obes Res 2003;11:1018–26.

15] Chen E, Paterson LQ. Neighborhood, family, and subjective socio-economic status: how do they relate to adolescent health? HealthPsychol 2006;25:704–14.

16] Beale F. Double-jeopardy: to be black and female. In: Cade T, ed.The Black Woman. New York, NY: Signet; 1970. pp 90–110.

17] McLeod JD, Owens TJ. Psychological well-being in the early lifecourse: variations by socioeconomic status, gender, and race/ethnicity. Soc Psychol Q 2004;67:257–278.

18] Benyamini Y, Idler EL, Leventhal H, et al. Positive affect andfunction as influences on self-assessments of health: expanding ourview beyond illness and disability. J Gerontol B Psychol Sci Soc Sci2000;55:107–16.

19] Wade TJ, Pevalin DJ, Vingilis E. Revisiting student self-rated phys-ical health. J Adolesc 2000;23:785–91.

20] Idler EL, Benyamini Y. Self-rated health and mortality: a reviewof twenty-seven community studies. J Health Soc Behav 1997;38:21–37.

21] Shadbolt B. Some correlates of self-rated health for Australianwomen. Am J Public Health 1997;87:951–6.

22] Wolinsky FD, Stump TE, Johnson RJ. Hospital utilization profilesamong older adults over time: consistency and volume among survi-vors and decedents. J Gerontol B Psychol Sci Soc Sci 1995;50:S88–
100.
23] Bobak M, Pikhart H, Hertzman C, et al. Socioeconomic factors,perceived control and self-reported health in Russia. A cross-sectionalsurvey. Soc Sci Med 1998;47:269–79.

24] Zimmer Z, Natividad J, Lin HS, et al. A cross-national examinationof the determinants of self-assessed health. J Health Soc Behav2000;41:465–81.

25] Blakely TA, Lochner K, Kawachi I. Metropolitan area income in-equality and self-rated health—a multi-level study. Soc Sci Med2002;54:65–77.

26] Lopez R. Income inequality and self-rated health in US metropolitanareas: a multi-level analysis. Soc Sci Med 2004;59:2409–19.

27] Torsheim T, Currie C, Boyce W, et al. Material deprivation andself-rated health: a multilevel study of adolescents from 22 Europeanand North American countries. Soc Sci Med 2004;59:1–12.

28] Goodman E. The role of socioeconomic status gradients in explainingdifferences in US adolescents’ health. Am J Public Health 1999;89:1522–8.

29] Cline A, Schafer-Kalkhoff T, Strickland E, et al. Recruitment strat-egies for the Princeton (Ohio) city school district epidemiologicstudy. J Sch Health 2005;57:189–91.

30] Rubin DB. Multiple imputation after 18� years. J Am Stat Assoc1996;91:473–89.

31] Little RJA, Rubin DB. Statistical Analysis with Missing Data. 2nd

edition. Hoboken, NJ: John Wiley & Sons; 2002.32] SAS Institute I. SAS/STAT User’s Guide, release 6.03 Edition. Cary,

NC: SAS Institute, Inc, 1988.33] Cooley CH. Human Nature and the Social Order. New York, NY:

Scribner’s; 1902.34] Williams DR. Race/ethnicity and socioeconomic status: measurement

and methodological issues. Int J Health Serv 1996;26:483–505.35] Crocker J, Major B. Social stigma and self esteem: the self protective

properties of stigma. Psychol Rev 1989;96:608–30.36] Branscombe NR, Schmitt MT, Harvey RD. Perceiving pervasive

discrimination among African Americans: implications for groupidentification and wellbeing. J Pers Soc Psychol 1999;77:135– 49.

37] Rosenberg M, Simmons R. Black and White Self Esteem: The UrbanSchoolchild. Washington, DC: American Sociological Association;1972.

38] Tajfel H, Turner JC. The social identity theory of inter-group behav-ior. In: Worchel S, Austin LW, eds. Psychology of Intergroup Rela-
tions. Chicago, IL: Nelson-Hall; 1986.

Documents

Perceived Socioeconomic Status: A New Type of Identity That Influences Adolescents’ Self-Rated Health