Download - Subjective age bias: A motivational and information processing approach

Transcript

International Journal of Behavioral Development2009, 33 (1), 22–31

http://www.sagepublications.com

© 2009 The International Society for theStudy of Behavioural Development

DOI: 10.1177/0165025408099487

The concept of subjective age, i.e., the idea of asking peoplehow old they feel, has intrigued scientists of different fields ofresearch, but in spite of the interest in the topic, many previousstudies have not linked this concept to a broader theoreticalperspective. Thus, the ability to make specific predictions andinterpret findings in a meaningful way has been compromised.

This article offers new insight into theoretical issues ofsubjective age in three regards: (1) an often used measurementmodel of subjective age, based on the four “ages of me”(Barak, 1987; Kastenbaum, Derbin, Sabatini, & Artt, 1972),is examined with regard to its underlying dimensional struc-ture and its fit with data; (2) two theoretical approaches thatexplain the subjective age bias are introduced and assessedempirically; and (3) to explore the relationships betweenvarious predictors and subjective age, a structural model ofsubjective age is developed and compared to empirical data.

Measurement of subjective ageKastenbaum et al. (1972) defined functional age as a multi -dimensional concept that includes four different domains of aperson’s subjective age. Based on this concept, Barak (1979,as cited in Barak, 1987) developed a measurement model,where respondents were asked to tick an age decade (twentiesto nineties) in response to four questions of the form:

1. feel-age: “I feel as if I were in my . . .”;2. look-age: “I look as if I were in my . . .”;3. do-age: “My activities are like those of people who are in

their . . .”;4. interest-age: “My interests are like those of people who are

in their . . .”.

Barak (1979, as cited in Barak, 1987) joined the four dimensions to a cognitive-age scale, which showed a very highinternal stability (alpha > 0.85). Subsequently, numerousstudies have used Kastenbaum et al.’s four aspects as a subjec-tive age measure (cf. Barak, Mathur, Lee, & Zhang, 2001;Barak, Mathur, Zhang, Lee, & Erondu, 2003; Barak & Rahtz,1999; Barak & Schiffman, 1981; Mathur, Barak, Zhang, &Lee, 2001; Mathur & Moschis, 2005; Montepare &Lachmann, 1989; Staats, 1996; Staats, Heaphey, Miller,Partlo, & Romine, 1993). The four aspects have been used, onthe one hand, to measure a person’s absolute subjective age (asin the questions above), and on the other, to measure compara-tive subjective age (e.g., “In comparison to people my age Ilook . . .” and the answer categories ranging from muchyounger to much older).

These four aspects of subjective age (feel-age, look-age, do-age, and interest-age) have often been referred to as the four“dimensions of functional age” (e.g., Barak et al., 2001; Barak& Stern, 1986). However, the high internal stability of the scalesuggests that the four items measure only one dimension.Montepare (1996c), on the other hand, found subjective ageto consist of three factors (psychological, physical, and socialage). In view of both Barak’s and Montepare’s research, oneobjective of this study was therefore to evaluate the measure-ment model of subjective age and to give an account of itsdimensional structure by means of structural equationmodeling.

Theoretical approaches to explain the subjective age biasThe concept of subjective age has been studied empiricallysince the 1950s (cf. a literature review by Barak & Stern,

Subjective age bias: A motivational andinformation processing approach

Ursina TeuscherDepartment of Cognitive Science, University of California, San Diego (UCSD), USA

There is broad empirical evidence, but still a lack of theoretical explanations, for the phenomenonthat most older people feel considerably younger than their real age. In this article, a measurementmodel of subjective age was assessed, and two independent theoretical approaches are proposed:(1) a motivational approach assuming that the age underestimation is a special form of self- enhancement; and (2) an information-processing approach assuming that elderly people look youngernowadays than formerly, which leads to an underestimation of their own age since people comparethemselves to an age-prototype which is no longer accurate. Data from 792 adults aged 58–70 supportthe motivational approach, without ruling out the information processing approach. Data from ayounger sample (42 students aged 20–35) provide evidence for the information processing approachadditionally. Further, a path model suggests that bodily aspects are the strongest predictor of subjective age.

Keywords: subjective age; age underestimation; self-enhancement; body image; time perspective

Correspondence should be sent to Ursina Tescher, Department ofCognitive Science, University of California, San Diego, 9500 GilmanDrive, La Jolla, CA 92093-0515; e-mail: [email protected]

This work was supported by the University of Fribourg, Switzerland.

The author is grateful to Benny Barak, Siegfried Macho, Dario Spini,and Christof Teuscher for their helpful comments and suggestions,and to Ben Motz for editing the language.

022-031 099487 Teuscher (Q8D):210 x 280mm 18/12/2008 14:16 Page 22

1986), so that there is now a considerable empirical descrip-tive basis for the phenomenon. Numerous studies have shownthat most older adults feel younger than their chronological age(e.g., Barnes, Farrell & Piotrowsky, 1989; Barnes-Farrell,Rumery, & Swody, 2002; Kaufman & Elder, 2002; Markides& Boldt, 1983; Montepare & Lachmann, 1989; Öberg &Tornstam, 2001; Staats et al., 1993), a general tendency thatcan also be referred to as subjective age bias, since it representsan overall inaccuracy of judgments if more than 50% of thepersons consider themselves younger than their cohort.However, in spite of the broad and highly consistent empiricalevidence, there is still a lack of theoretically founded psycho-logical explanations for the phenomenon that most elderlypeople feel considerably younger than their calendar age.

One promising theoretical approach might be the domain ofpositivity bias or positive illusions. There is extensive evidencethat most people tend to overestimate their own qualities andabilities (see e.g., Gramzow, Elliot, Asher, & McGregor, 2003;Kwan, Love, Ryff, & Essex, 2003; Mezulis, Abramson, Hyde,& Hankin, 2004; Sedikides, Gaertner, & Toguchi, 2003;Sedikides, Herbst, Hardin, & Dardis, 2002; Silvia & Duval,2001; Taylor, Lerner, Sherman, Sage, & McDowell, 2003).The tendency to view oneself as younger than one’s agecohorts could be seen as a special case of self-enhancement,since to feel, look and act young is generally considered to besomething positive. That explanation would assume thatpeople have a motivation to feel young.

Another approach offers a non-motivational explanation forthe subjective age bias in terms of information processing: itis possible that elderly people do really look younger today thanthey did a few decades ago, since the health state of elderlypeople is better today. Therefore, the inner prototype imagethat we have of a person in his or her sixties might no longercorrespond to the aspect of today’s “normal” people in theirsixties. If people have to decide whether they look younger thanother people of the same age, it is possible that they comparethemselves to that obsolete prototype.

The aim of the current study is thus to discover how appro-priate these two approaches are in explaining the phenomenonof the subjective age bias.

The assumption of a motivational – that is, self-enhancing –component leads to three specific predictions or hypotheses:

One difference concerns the attitude toward older age insociety. If the tendency to view oneself as younger was aspecial case of self-enhancement, we would expect theattitude towards older age to play an important role. Feelingespecially young can only be positive for a person’s well beingif “to be or feel old” is seen as something negative. We wouldtherefore expect a moderating effect of the attitude towardsold age on the correlation between a young subjective ageand a high life satisfaction. This correlation should be especi-ally high for people with a negative attitude towards old age.

A strong tendency to self-enhancement and positive illusionsabout oneself is highly related to other personality traits suchas optimism and self-efficacy belief, as well as to life satis-faction (Myers & Diener, 1995). If young subjective age wasa special case of self-enhancement, we would thereforeexpect a correlation between a young subjective age, on onehand, and high optimism, high self-efficacy belief, and highlife satisfaction, on the other.

Self-enhancement tendencies are usually stronger for domainsthat cannot easily be verified, that is, where no objective

feedback is available or where standards are ambiguous andsubjective (cf. Dunning, Meyerowitz, & Holzberg, 1989;Felson, 1981). We would therefore expect people to feelyounger for the general subjective age than for bodily aspectsof subjective age, assuming that for bodily aspects of subjec-tive age, such as fitness or attractiveness, there is usuallymore feedback available and standards are less ambiguousthan for the general subjective age, that is, for the generalquestion of how old or young a person feels.

The information processing approach, on the other hand,leads to another (independent) prediction:

If people make use of an obsolete age prototype to estimatetheir own subjective age, then they are supposed to under-estimate not only their own age, but also the age of others.

A path model of subjective ageA further goal of the present study was to analyze the relation-ship of several variables that relate to subjective age, thusproviding a path model of subjective age. The following groupsof variables were included in this study.

Bodily aspects such as health state and satisfaction with one’s fitnessand attractiveness: Several lines of research suggest a strongconnection between bodily aspects and age identity.Montepare (1996b) suggested a process model wherebyvariations in subjective age are related to age markers, thatis, events that make age salient. One kind of marker wouldbe physical events such as health-related issues. Her researchshowed furthermore that overall body satisfaction wasrelated to subjective age (Montepare, 1996a). Catterall andMaclaran (2001) point out that subjective age reflects widerexpectations that we must maintain a youthful, fit body. Ayoung and healthy body is supposed to be a sign of a youngand healthy mind. The importance of the body is alsoreflected in concerns about health and loss of physicalstrength and agility, which have been identified as the mainconcern of older adults in a major study of the US popu-lation (Moschis, Lee, Mathur, & Strautman, 2000). It cantherefore be expected that bodily aspects are good predic-tors of a young subjective age. This connection has beenconfirmed in previous studies of the respondents’ healthstate (Baum & Boxley, 1983; George, Mutran, & Penny-baker, 1980; Gwinner & Stephens, 2001; Markides & Boldt,1983; Markides & Ray, 1988; Mathur & Moschis, 2005;Mutran & George, 1982; Sirgy, Mentzer, Rahtz, & Meadow,1991) as well as of attractiveness (Guiot, 2001; Montepare,1996a). The variable of bodily fitness was recently includedin this work.

Life satisfaction and personality aspects such as self-efficacy beliefand optimism: Myers and Diener (1995) showed that positiv-ity bias is stronger for people with high life satisfaction, self-efficacy belief and optimism. It was thus hypothesized thatlife satisfaction, self-efficacy belief, and optimism correlatealso with a young subjective age. The connection of subjec-tive age with self-efficacy belief and satisfaction has alreadybeen shown in previous literature (cf. Barak & Gould, 1985;Baum & Boxley, 1983; George et al., 1980; Mutran &George, 1982), even though there is also a contrary findingby Montepare and Lachmann (1989), showing higher satis-faction in older adults to be correlated with an older subjec-tive age. The variable of optimism was recently included in

INTERNATIONAL JOURNAL OF BEHAVIORAL DEVELOPMENT, 2009, 33 (1), 22–31 23

022-031 099487 Teuscher (Q8D):210 x 280mm 18/12/2008 14:16 Page 23

the present study. The attitude towards old age is a furtherinteresting variable, since there are two possible ways ofinfluence. As Gwinner and Stephens (2001) point out, wemight on first intuition suppose that a negative attitudetowards old age will lead to a young subjective age, becausepeople who view old age negatively would try to dissociatethemselves from being old. Gwinner and Stephens ascribetheir own surprisingly contrary finding to their highlyeducated sample, but there is another possible explanationthat would speak for a real correlation between an oldsubjective age and negative views of old age: a positiveattitude towards aging as well as feeling young might bothbe correlated to life satisfaction, and therefore also to eachother. In that case, however, the correlation is expected tovanish if it is controlled for satisfaction.A young subjective age can be assumed to predict the follow-

ing behavioral tendencies, which were recently included in thisstudy:

Learning new things: According to the socio-emotional selectiv-ity theory (cf. Carstensen, Isaacowitz, & Charles, 1999;Fung, Carstensen, & Lutz, 1999; Lang & Carstensen,2002), the limited future time perspective that comes withaging leads to an increased importance of emotionallysignificant goals, since these are immediately rewarding, anddecreased importance of knowledge-oriented goals whichwould be important for a further future. If subjectively oldpeople perceive their future to be more limited than subjec-tively young people, we would therefore expect them to haveless knowledge-oriented goals, that is, to be less interestedin learning new things.

Engaging in “youth-preserving” activities: There are a lot ofactivities that are supposed to delay the process of aging (orat least the process of looking or feeling older) such asexercise, healthy food, cosmetic products, memory training,and so forth. If at least some of these activities are useful,or entail effort justification, engaging in them should berelated to a young subjective age. In the same way, peoplefor whom feeling or looking young is important can beassumed to engage more in these activities.

The above discussed considerations are integrated in a hypo-thetical path model, as shown in Figure 1. (The arrows inFigure 1 do not stand for causality but for the hypothesizedpredictive value of one variable for another.)

Thus, the study presented in this article focuses on threemajor aspects of subjective age: (1) a measurement modelbased on the functional “ages of me” model of Kastenbaum

et al. (1972) will be assessed, on the one hand, regarding itsapplicability as a scale and, on the other, regarding its under-lying dimensional structure. The so far unquestioned assump-tion that the model consist of four dimensions will thus bechallenged. (2) The two aforementioned theoretical approacheswill be evaluated with regard to their appropriateness toexplain the subjective age bias; the competing predictionsdeduced from the two approaches being tested empirically.(3) The above-developed predictive model of subjective agewill be compared to empirical data by means of structuralequation modeling.

Method

ParticipantsA questionnaire was sent to a random sample of 2000 persons1

aged 58–70 years (M = 63.65, SD = 3.77) who live in Thun(a town in Switzerland with around 40,000 inhabitants). Thisrelatively restricted age range was chosen because the theoreti-cal approaches that were tested in this study (especially themotivational approach) are specific for subjective age bias inolder adults, and are not necessarily generalizable to youngerpersons’ age identification. The random sample of addresseswas drawn by the municipal authority of Thun. Some 802questionnaires were completed and sent back, 10 of whichwere excluded from the analysis because more than a quarterof the values were missing. Thus the analysis included n =792 persons; 386 men and 406 women. This corresponds toa rate of return of 39.6%, which can be considered high,given that the persons addressed had not been previouslyasked if they wanted to participate and there was no rewardor reminder letter afterwards (cf. Newby, Watson, & Woodliff,2003).

For the age ratings of other people, 42 students of theUniversity of Fribourg participated in the study. The age ofthese participants ranged from 20–35 years (M = 24.68, SD =4.12).

24 TEUSCHER / SUBJECTIVE AGE BIAS

1 Since some of the effects were hypothesized to be rather small, a poweranalysis suggested a sample size of around 500 persons (Cohen, 1988). Thereturn rate of a questionnaire survey without incentives can in general not beexpected to exceed 25%, therefore the questionnaire was sent to a primarysample of 2000 persons.

Figure 1. Predictive path model of subjective age.

022-031 099487 Teuscher (Q8D):210 x 280mm 18/12/2008 14:16 Page 24

QuestionnaireThe original language of the questionnaire was German. Therelevant items or scales of the questionnaire will be describedin the following, the items having been translated into Englishby the author.

1. Subjective age measures. The study used nine items todetermine the participants’ comparative subjective age ingeneral as well as regarding specific domains. The domainscovered bodily, cognitive and social aspects of how old onefeels. One item measured the subjective age in general with thequestion: “If you compare yourself to people your age, howold do you feel in general?” There were five possible answersranging from “much older” to “much younger”.

Eight items then covered specific bodily, social and cogni-tive domains of subjective age:

If you compare yourself to people your age and think aboutdifferent life domains, how old do you feel, regarding:

your bodily fitness?your activities?your relations to other people?your interests?your mental abilities?your esteem by society?your esteem by your family and friends?your attractiveness?

For every item, there were again the same five possibleanswers ranging from “much older” to “much younger”. Thenine items had a Cronbach’s alpha of .86; all corrected item-total correlations were above .5.

Four of these overall nine items corresponded to Kasten-baum et al.’s (1972) four aspects of subjective age (“ages ofme”): the general feeling (= feel-age), the activities (= activity-age), the interests (= interest-age), and the attractiveness(= look-age). Additionally, the participants were asked how oldthey felt in years, how old other persons would estimate themand how old they would prefer to be.

2. Health measures. There were two items regarding the stateof health of the participants, one on the general subjective stateof health, asking: “How would you characterize your generalhealth state?”, with five possible answers from “very good” to“very bad”. Another item referred to behavioral consequencesof the state of health, asking: “How many times have you beenso ill in the past 12 months, that you could not accomplishimportant everyday tasks?”, with five possible answers:“never”, “once”, “twice to three times”, “more than threetimes”, and “chronically”.

The two items were joined to a scale, averaging the midpointvalues of responses to these two items (Cronbach’s alpha =0.71).

3. Psychographic measures. There were four psychographicmeasures:

Self-efficacy: The measurement of perceived self-efficacy reliedon five items of the German Version of Jerusalem andSchwarzer’s (1986) General Self-efficacy Scale. The fiveitems had a Cronbach’s alpha of .79; all corrected item-totalcorrelations were above .5.

Optimism: The measurement of optimism relied on five items

on Scheier and Carver’s (1985) Optimism Scale (Germantranslation by Wieland-Eckelmann and Carver (1990)). Thefive items had a Cronbach’s alpha of .64; all corrected item-total correlations were above .3.

Satisfaction: There were ten items asking the respondents howsatisfied they were with different life domains. The fivepossible answers ranged from “very satisfied” to “verydissatisfied”. The value for a respondent’s overall satisfactionwas computed by averaging the midpoint values of responsesto all 10 items. Reliability assessment showed the scale to beinternally stable (Cronbach’s alpha = .86; all corrected item-total correlations > .4).

Attitude towards aging: The respondents’ attitude towards agingand older age was measured in two different ways. First, threedirect questions asked the respondents how positive ornegative they judged life when one is around 75 years old, orolder than 80; and if they would like to reach 100 years oldprovided that they remained healthy. The midpoint values ofresponses were averaged for all three items, resulting in asatisfyingly reliable scale (Cronbach’s alpha = .71; allcorrected item-total correlations > .4). Then, as an indirectmeasure, the respondents were asked to rate the fit of theterm “old age” with a list of 11 positive or negative adjec-tives. There were five possible answers (“suits very well” to“does not suit at all”) to rate every adjective. The midpointratings of adjectives were averaged and joined to a scale(Cronbach’s alpha = .87; all corrected item-total correla-tions > .46). The measure provides information as to howpositively or negatively associated a respondent’s image ofolder age is.

4. Behavioral measures. There were two behavioral measures:

Learning new things: Respondents were asked what they hadlearnt in the past few years. They were given five answersuggestions (e.g., a foreign language, new skills for a hobby,etc.) and were asked to tick every appropriate answer.Furthermore, they were given free space to list additionalthings that they had learnt. The crosses and free answerswere counted.

Engaging in “youth-preserving” activities: Respondents wereasked what they did to feel or appear youthful. They weregiven eight answer suggestions (e.g., taking exercise, dyinghair, training memory skills, etc.) and were asked to tickevery appropriate answer. Furthermore, they were given freespace to list additional things. The crosses and free answerswere counted.

Picture material for the age ratingsThirty-five colored pictures cut from a Swiss weekly journalserved as target pictures. The pictures had all appearedbetween June and September 2003, and were collected follow-ing the criterion that the real age of the presented person beindicated in the accompanying journal article (the age rangedfrom 30–82 years; M = 55.07, SD = 14.60).

ProcedureQuestionnaire survey. The questionnaire was posted on April29, 2002, to the 2000 randomly drawn citizens of Thun, witha prepaid addressed response envelope and an accompanyingletter asking them to send back the completed questionnaire.

INTERNATIONAL JOURNAL OF BEHAVIORAL DEVELOPMENT, 2009, 33 (1), 22–31 25

022-031 099487 Teuscher (Q8D):210 x 280mm 18/12/2008 14:16 Page 25

Age ratings. Each of the 42 students was given all the targetpictures one by one and was asked to estimate the age of eachperson pictured. The subjects were additionally asked to tickthe targets that seemed in any way familiar to them. Ratings ofpictures that appeared familiar to any of the subjects wereexcluded from the analysis.2

Results

Structural measurement model of subjective ageOne of the objectives of this work was to evaluate the measure-ment model based on Kastenbaum et al.’s (1972) functionalmodel of subjective age, which has often been used, especiallyin marketing psychology (cf. Barak et al., 2001; Barak et al.,2003; Barak & Rahtz, 1999; Mathur et al., 2001; Staats, 1996;Staats et al., 1993), but so far had not been compared withother measures of subjective age.

Table 1 shows the means and item-total statistics of thesubjective age scale comprising all nine items that wereacquired in this study. That is, the scale included not only theitems based on Kastenbaum et al.’s “ages of me” (items 1, 3,4, and 9: “feel-age”, “activity-age”, “interest-age”, and “look-age”) but also five additional items concerning other aspectsof subjective age. The scale proved to be internally stable withCronbach’s alpha = .86; all corrected item-total correlationswere greater than .5.

When only those items that were based on Kastenbaum etal.’s (1972) “ages of me” (items 1, 3, 4, and 9) were joined toa scale, Cronbach’s alpha decreased to .74; all four correcteditem-total correlations being greater than .4. The brevity of thisreduced scale is certainly an advantage regarding the usabilityof the scale – given that it is still reliable – since the economyof a scale is an important criterion for most questionnairestudies.

Figure 2 shows a structural equation model that representsKastenbaum et al.’s four aspects of subjective age. Themodeling was realized with AMOS (Version 5) for SPSS, thatis, full information maximum likelihood (FIML) estimationwas used as a built-in imputation procedure to deal withmissing values.

The goodness of fit of this model proved to be high; !2(2)= 1.223, p = .542, TLI = 1.00, CFI = 1.00, RMSEA = .00.In particular, the chi-square was astonishingly low (and thep-value accordingly high), considering that these criteria areusually not used since they seem to be unrealistically severe,especially for large sample sizes, as the chi-square value issample-sensitive (Hu & Bentler, 1998). It appears thereforesafe to say that the model based on Kastenbaum’s “ages of me”fit the empirical data exceptionally well.

Adding more items, as well as exchanging one or more ofthe four “ages of me” items with any of the remaining fiveitems, resulted in a dramatic decrease of the goodness of fit,all modified models showing !2(2) > 11 (p < .01), andRMSEA > .1. Thus, Kastenbaum et al.’s four aspects appearto provide a very good basis for a measurement model ofsubjective age.

This high model fit suggests that the structure of Kasten-baum et al.’s (1972) four aspects of subjective age is one-dimensional. A factor analysis including those four itemsshowed indeed that they all loaded on the same factor.However, including all nine items in a factor analysis yieldedtwo – though highly oblique – factors, as can be seen in Table 2and Figure 3.

Since Kastenbaum et al.’s four aspects were appropriatemeasures of subjective age, all of the following computationsin this article are based on that four-item scale instead of theexpanded scale consisting of nine items, in order to allow

26 TEUSCHER / SUBJECTIVE AGE BIAS

Table 2Principal component analysis with different aspects of subjectiveage: Rotated component matrix (Oblimin with Kaisernormalization, absolute values less than .3 suppressed)

Component

1 2

General feeling .626Fitness .944Activities .730Relations to others .827Interests .801Mental abilities .743Esteem by society .644Esteem by family and friends .570Attractiveness .798

2 Familiar pictures might be rated younger than unfamiliar pictures, sincefamiliarity often leads to more positive ratings in general. For the questionwhether the age underestimation can be observed independent of any positivitybias, it was therefore important to exclude the familiar pictures.

Table 1Reliability analysis: Means and item-total statistics for themeasurement of subjective age

Corrected AlphaItem: Mean item-total if item(domain of subjective age) score correlation deleted

General feeling 3.73 .63 .84Fitness 3.51 .57 .85Activities 3.65 .62 .84Relations to others 3.56 .54 .85Interests 3.66 .54 .85Mental abilities 3.53 .56 .84Esteem by society 3.46 .60 .84Esteem by family and friends 3.63 .59 .84Attractiveness 3.49 .62 .84

Figure 2. Structural measurement model of comparative subjectiveage, based on Kastenbaum et al.’s (1972) “ages of me”.

022-031 099487 Teuscher (Q8D):210 x 280mm 18/12/2008 14:16 Page 26

better comparability of the present results to the results ofother studies.

The tendency of people to underestimate their age:Descriptive resultsThe participants’ average real age was 63.65 years (SD = 3.77),but they rated their subjective age or feel-age on average 5.57years younger, that is 57.94 years old (SD = 6.18). They esti-mated the age that others would give them as 57.78 years(SD = 5.26), i.e. 5.72 years younger than their real mean age.When asked how old they would like to be, participantsspecified their ideal age on average as 55.07 years old (SD =12.49), that is, 8.38 years younger than their real mean age.

When participants were asked to compare themselves toother people, 67.7% stated that they felt younger or very muchyounger, 29.7% that they felt about equally old, and only 1.9%that they felt older or very much older than most other peopleof their age cohort (see Figure 4). These percentages suggestthat a considerable number of the participants did not possessan accurate view concerning their own and other people’ssubjective age, since it is not possible that more than 50% feelyounger than most other people feel. Assuming that the samplewas tolerably representative, the distribution should be moreor less symmetrical if the ratings were accurate, which wasclearly not the case here.

Assessment of two theoretical approaches explainingsubjective age biasBased on the self-enhancement approach, a question of thisstudy was if the subjective age bias implies a motivationalcomponent. Three hypotheses dealing with this question weretested here:

Hypothesis 1: Attitude towards aging and its moderatinginfluence on subjective age correlates. Hypothesis 1 concernsthe moderating influence of the attitude toward old age,

suggesting that if there is motivation involved in the subjec-tive age bias, the correlation between a young subjectiveage and a high life satisfaction should be stronger forpeople with a negative attitude towards old age.

Figure 5 shows that this was indeed the case: for bothmeasures of attitude towards old age (the direct and the associ-ative question), a moderating effect on the interrelationshipbetween subjective age and satisfaction could be shown. Peoplewho felt younger were in general more satisfied with differentaspects of their life – the correlation was in any case consider-able – but this connection was especially strong for people witha negative view of old age. The difference between the corre-lations was highly significant for both measures; for the directquestion: z(275, 475) = 3.15, p (one-tailed) < 0.001, as wellas for the associative ratings: z(359, 399) = 4.71, p (one-tailed)< 0.001.

INTERNATIONAL JOURNAL OF BEHAVIORAL DEVELOPMENT, 2009, 33 (1), 22–31 27

Figure 3. Component plot in rotated space showing different aspects of comparative subjective age (extraction method: principal componentanalysis; rotation method: Varimax).

Figure 4. Answer distribution of comparative subjective age ratings.

022-031 099487 Teuscher (Q8D):210 x 280mm 18/12/2008 14:16 Page 27

Hypothesis 2: Correlation of subjective age with optimism andself-efficacy belief. Hypothesis 2 concerned the correlationof a younger subjective age with optimism, self-efficacybelief, and satisfaction. If the general age underestimationis a special form of self-enhancement, younger subjectiveage is supposed to correlate with these variables.

The findings indeed showed such correlations: youngsubjective age correlates with high self-efficacy belief: r(772) =.305**, p < .001; with high optimism: r(772) = .303**, p <.001; as well as with high life satisfaction: r(773) = .483**,p < .001. All these correlations remained significant ifcontrolled for health state

Hypothesis 3: Bodily versus general subjective age. Hypoth-esis 3 suggests that if subjective age is a special case ofself-enhancement, the underestimation should be morepronounced for the general subjective age than for bodilyaspects.

The results confirmed this hypothesis: when the participantswere asked about their general subjective age (i.e. the question:“If you compare yourself to people your age, how old do youfeel in general?”), their subjective age was younger (M = 3.72,SD = .59) than when they were asked about bodily aspects ofsubjective age, i.e.:

• the subjective age regarding attractiveness (M = 3.48,SD =.79), t(765) = 9.14; p (one-tailed) < 0.001; d = 0.35;

• the subjective age regarding fitness (M = 3.50, SD =.80),t(780) = 8.54; p (one-tailed) < 0.001; d = 0.31.

Figure 6 shows the average ratings of all subjective agedomains that were measured here, illustrating the differencebetween the general subjective age and the bodily aspects(fitness and attractiveness). This result also speaks for the self-enhancement explanation, because people estimated them-selves as younger for domains where there are fewer standardsand less feedback – from the mirror, for example.

The tendency to underestimate other people’s ageIs this motivation for self-enhancement enough to explain thesubjective age bias? If so, we could assume that subjects makeaccurate age ratings when they are asked to rate other people’s

age, since underestimating other people’s age cannot beassumed to be helpful for a subject’s own self-esteem.However, the findings of this study suggest that evencompletely unknown faces are generally rated too young. Onaverage and across all the target pictures, the subjects in thisstudy underestimated the (unfamiliar) target persons by 4.68years. The difference between the real age of the target persons(they were on average 55.06 years old) and the estimated ageacross all subjects and all target pictures (M = 50.38 years,SD = 2.09) was highly significant in a one-sample t-test, t(42)= 14.66, p < .001.

A predictive structural equation model of subjective ageTo test the predictive model of subjective age, structuralequation modeling (SEM) was used. Figure 7 shows the pathmodel, which is based on the hypothetical model as shown inFigure 1.

The chi-square fit index was significant, !2(112) = 642.185,p < .001, indicating that the model differed significantly fromthe empirical data. However, since the chi-square statistic issensitive to sample sizes, it has been found to over-rejectsubstantially true population models in settings with largesample sizes (cf. Backhaus, 1996; Hu & Bentler, 1998). Thus,the root mean square error of approximation (RMSEA) wasadditionally examined, being especially appropriate forcomplex models, and showed acceptable, though not verygood model fit: RMSEA = .077.

Discussion

Measurement of subjective ageThe findings presented in this article support the classicmeasurement model of subjective age (Barak, 1987) based onthe functional model by Kastenbaum et al. (1972) that consistsof four aspects, the so-called “ages of me”. The evaluation ofthis measurement model suggests that the model can berecommended for further use, since it has both exceptionallygood fit indices as well as economic advantages, consisting ofonly four items. The model is one-dimensional though,

28 TEUSCHER / SUBJECTIVE AGE BIAS

Figure 5. Correlation between younger subjective age and higherlife satisfaction showing a moderating effect of the attitude toward oldage.

Figure 6. Answer distribution for different domains of subjective age.

022-031 099487 Teuscher (Q8D):210 x 280mm 18/12/2008 14:16 Page 28

therefore it does not seem appropriate to refer to Kasten-baum’s four aspects of subjective age as a multidimensionalmodel, as was done in previous literature (e.g., Barak et al.,2001; Barak & Stern, 1986). If the construct of subjective agewas measured with more items (which resulted in higherreliability but worse fit), it seemed to consist not only of onebut of two factors, but these factors were not independent, andthus cannot be interpreted as different dimensions. However,it might be expedient to further evaluate the two-factorialstructure of the concept. A possible interpretation of the twofactors is that Factor 1 covers the cognitive and social aspects,whereas Factor 2 covers the rather body-oriented aspects ofsubjective age. This interpretation is consistent withMontepare’s (1996a) finding that women’s physical subjectiveages seem to have a stronger relationship with their bodyesteem than did their psychological or social subjective ages.Even though Montepare (1996c) extracted not two but threefactors of subjective age (her study involved more and partlydifferent items to measure subjective age), both studies showa distinction between body-oriented and other aspects ofsubjective age in their factor analyses.

As a multiple item scale that relies on other-group referentinterdependent age measurement, the present scale possessesmore measurement equivalency across different cultures, andthus a higher external validity, than individuated subjective agescales (Van Auken, Barry, & Bagozzi, 2006; Barak, 2009).

Theoretical approaches to explain the subjective agebiasConsistent with previous studies on subjective age of olderadults, participants of the present study felt in general consid-erably younger than their calendar age. However, the resultsfor the absolute subjective age showed that in this study thediscrepancies between real age and feel-age as well as ideal agewere much less pronounced than they were in former studieswith other samples: for instance, Öberg and Tornstam (2001)found that Swedish people in that age group felt around 15years younger than their real age (as opposed to 5.6 years inthe present study), and the ideal age was even around 20 years

younger than the real age (as opposed to 8.3 years in thepresent study). These considerable differences between thestudies suggest the strong influence of intercultural differenceson the subjective age. (However, as long as there is no compari-son study with the exact same method of measuring subjectiveage, it cannot be excluded that the differences were measure-ment artifacts.) To learn more about the reasons for theseintercultural differences, it would be interesting to compare theself-enhancing tendencies, as well as to analyze the images ofaging and the attitude towards higher age of these differentcultures.

The standard deviation was higher for the participants’subjective age than for their real age, and still higher for theirideal age. That is, the respondents gave less homogenousanswers when asked how old they felt or how old they wouldprefer to be than when asked about their real age. This findingcorresponds to the results of a previous study by Kaufman andElder (2002), suggesting that the question about the ideal ageis the only question that allows the subjects to give any answer,without concern about having to be realistic.

The results suggest further that the phenomenon of thesubjective age bias implies two components: motivation aswell as information processing. Three findings speak for a self-enhancement – that is, motivational – component: (1) thecorrelation between young subjective age and high life satis-faction was stronger for people with a negative attitudetowards old age;3 (2) consistent with findings about strongself-enhancement tendencies (cf. Dunning et al., 1989;Felson, 1981; Myers & Diener, 1995), a strong age-underes-timation correlated with high optimism, self-efficacy belief,and satisfaction; and (3) the age underestimation was morepronounced for the general subjective age than for specificbodily aspects.

INTERNATIONAL JOURNAL OF BEHAVIORAL DEVELOPMENT, 2009, 33 (1), 22–31 29

3 However, the results showed also that a negative attitude towards old agecould not entirely explain the difference: even for people with a very positive viewof old age, there was a significant connection between feeling young and a highlife satisfaction. In other words, the attitude towards old age had a moderating,not a mediating effect on that connection.

Figure 7. Predictive structural equation model of subjective age.

022-031 099487 Teuscher (Q8D):210 x 280mm 18/12/2008 14:16 Page 29

The finding that subjects also underestimated the calendarage of entirely unknown target persons cannot be explained byself-enhancement tendencies but suggests an additional information-processing influence. This finding supports thehypothesis that most people may have an inner prototype of aperson in a certain age that no longer corresponds to the looksof the “average” person in that age. However, the present studydoes not allow any comparison between these ratings and thedata of the older adults’ subjective age, since they stem fromdifferent samples. It would therefore be important to replicatethese findings with one sample in order to evaluate whether therespondents’ tendency to underestimate the ages of others isrelated to their own tendency to report feeling younger thantheir chronological age. Also, a direct measure of the ageprototype would be important to consolidate the presentinterpretation. Another interesting approach in this regardwould be to investigate the influence of age schematicity (cf.Montepare & Clements, 2001), on the one hand on people’sjudgments about their own subjective age, but also on theirestimations of other people’s real ages.

For future studies on subjective age, it seems promising toinclude both approaches (motivational and informationprocessing) in the theoretical frameworks. Especially in thefield of self-enhancement, there is a vast body of researchwhich has so far been ignored but might enrich future studiesconsiderably.

A path model of subjective ageThe path model of subjective age had an acceptable, but notvery high fit, and the item parcels showed higher variabilitythan would be wished for. In spite of these deficiencies, someconclusions are ventured here: the model showed that bodilyaspects explain the largest part of the variance of subjectiveage. The importance of the body in the experience of agingwas further shown in the factor analysis of the different aspectsof subjective age, where the general feeling of subjective ageloads higher on the factor comprising bodily aspects than onthe factor comprising cognitive and social aspects.

These findings confirmed the view of Öberg (1996) thataging is experienced to a large extent via the body, which is nottaken enough into account by social gerontologic research.

The structural equation model showed further that psycho-graphic variables such as satisfaction, personality, and attitudetowards aging are also valid predictors of subjective age. Thatis, people who feel especially young are more satisfied withtheir lives, have a high self-efficacy belief and are optimistic.Interestingly, they even have a more positive attitude towardsaging. The finding that a younger subjective age is positivelycorrelated with a positive attitude towards higher age corre-sponds to the unexpected findings of Gwinner and Stephens(2001), suggesting that their findings were not just, as theauthors had assumed, measurement artifacts due to sampleirregularities, but that they reflect correlations that exist in thepopulation. The present study tested therefore also whetherthis correlation can be explained by a higher life satisfaction ofboth people with positive attitudes towards higher age andsubjectively young people. For the associative ratings, thoughnot for the direct questions, the correlation between feelingyoung and viewing higher age positively indeed disappearedwhen controlled for satisfaction. To further elucidate thisrelation, it would be important for future studies to identifymeasures that make a clear distinction between self- and other-

related attitudes towards aging, which the present measurescannot assure.

The predictive value of subjective age for the number ofrecently learnt things can be interpreted in the framework ofsocio-emotional selectivity theory: people who perceive theirfuture time perspective as more limited have been shown tofocus more on emotional and less on knowledge-oriented goals(e.g., Carstensen, 1992; Carstensen et al., 1999; Carstensen,Fung, & Charles, 2003; Fung & Carstensen, 2004; Fung et al.,1999). Learning new things can clearly be seen as knowledge-oriented behavior. The fact that subjectively younger respon-dents specified more things that they had recently learntindicates that to feel young might be linked to a less limitedtime perspective – in other words, people who feel young mayindeed perceive death as further away than people who feel old.The findings are also consistent with findings from marketingpsychology, where subjectively young persons have been foundto be more willing to try new brands and to seek information(cf. Stephens, 1991; Szmigin & Carrigan, 2001).

Hopefully, the findings presented here will inform additionalresearch on age identity that not only builds on a solid theor-etical framework but also contributes to the further develop-ment of psychological theories. The topic of age identity meritsfurther research, in particular experimental research thatwould elucidate the direction of causal influence betweensubjective age and other variables.

References

Backhaus, K. (1996). Multivariate Analysemethoden. Berlin: Springer.Barak, B. (1987). Cognitive age: New multidimensional approach to measuring

age identity. The International Journal of Aging and Human Development, 25,109–125.

Barak, B. (2009). Age identity: A cross-cultural approach. International Journalof Behavioral Development, 33, 2–11.

Barak, B., & Gould, S. (1985). Alternative age measures: Research agenda.Advances in Consumer Research, 12, 53–58.

Barak, B., Mathur, A., Lee, K., & Zhang, Y. (2001). Perceptions of age-identity:Cross-cultural inner-age exploration. Psychology and Marketing, 18,1003–1029.

Barak, B., Mathur, A., Zhang, Y., Lee, K., & Erondu, E. (2003). Inner-age satis-faction in Africa and Asia: Cross-cultural exploration. Asia Pacific Journal ofMarketing and Logistics, 15, 3–26.

Barak, B., & Rahtz, D.R. (1999). Perceived youth: Appraisal and characteriza-tion. International Journal of Aging and Human Development, 49, 231–257.

Barak, B., & Schiffman, L.G. (1981). Cognitive age: Non-chronological agevariable. Advances in Consumer Research, 8, 602–606.

Barak, B., & Stern, B. (1986). Subjective age correlates: Research note. TheGerontologist, 5, 571–578.

Barnes-Farrell, J.L., & Piotrowsky, M.J. (1989). Workers’ perceptions of discrep-ancies between chronological age and personal age: You’re only as old as youfeel. Psychology and Aging, 4, 376–377.

Barnes-Farrell, J.L., Rumery, S.M., & Swody, C.A. (2002). How do concepts ofage relate to work and off-the-job stresses and strains? Field study of healthcare workers in five nations. Experimental Aging Research, 28, 87–98.

Baum, S.K., & Boxley, R.L. (1983). Age identification in the elderly. The Gerontologist, 23, 532–537.

Carstensen, L.L. (1992). Social and emotional patterns in adulthood: Supportfor socio-emotional selectivity theory. Psychology and Aging, 7, 331–338.

Carstensen, L.L., Fung, H.H., & Charles, S.T. (2003). Socio-emotional selec-tivity theory and the regulation of emotion in the second half of life. Motiva-tion and Emotion, 27, 103–123.

Carstensen, L.L., Isaacowitz, D., & Charles, S.T. (1999). Taking time seriously.Theory of socio-emotional selectivity. American Psychologist, 54, 165–181.

Catterall, M., & Maclaran, P. (2001). Body talk: Questioning the assumptionsin cognitive age. Marketing and Psychology, 18, 1117–1133.

Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.).Hillsdale, NJ: Erlbaum.

Dunning, D., Meyerowitz, J.A., & Holzberg, A.D. (1989). Ambiguity and

30 TEUSCHER / SUBJECTIVE AGE BIAS

022-031 099487 Teuscher (Q8D):210 x 280mm 18/12/2008 14:16 Page 30

INTERNATIONAL JOURNAL OF BEHAVIORAL DEVELOPMENT, 2009, 33 (1), 22–31 31

self-evaluation: The role of idiosyncratic trait definitions in self-serving assess-ment of ability. Journal of Personality and Social Psychology, 57, 1082–1090.

Felson, R.B. (1981). Ambiguity and bias in the self-concept. Social PsychologyQuarterly, 44, 64–69.

Fung, H.H., & Carstensen, L.L. (2004). Motivational changes in response toblocked goals and foreshortened time: Testing alternatives to socio-emotionalselectivity theory. Psychology and Aging, 132, 310–324.

Fung, H.H., Carstensen, L.L., & Lutz, A. (1999). The influence of time onsocial preferences: Implications for lifespan development. Psychology andAging, 14, 595–604.

George, L.K., Mutran, E.J., & Pennybaker, M.R. (1980). The meaning andmeasurement of age identity. Experimental Aging Research, 6, 283–298.

Gramzow, R.H., Elliot, A.J., Asher, E., & McGregor, H.A. (2003). Self- evaluation bias and academic performance: Some ways and some reasonswhy. Journal of Research in Personality, 37, 41–61.

Guiot, D. (2001). Antecedents of subjective age biases among senior women.Psychology and Marketing, 18, 1049–1071.

Gwinner, K.P., & Stephens, N. (2001). Testing the implied mediational role ofcognitive age. Psychology and Marketing, 18, 1031–1048.

Hu, L., & Bentler, P.M. (1998). Fit indices in covariance structure modeling:Sensitivity to underparameterized model misspecification. PsychologicalMethods, 3, 424–453.

Jerusalem, M., & Schwarzer, R. (1986). “Selbstwirksamkeit WIRK”. Skalen zurBefindlichkeit und Personlichkeit. In R. Schwarzer (Ed.), Forschungsbericht derFU Berlin, Institut für Psychologie. Berlin: FU Berlin.

Kastenbaum, R., Derbin, V., Sabatini, P., & Artt, S. (1972). “The ages of me”:Toward personal and interpersonal definitions of functional age. InternationalJournal of Aging and Human Development, 3, 197–212.

Kaufman, G., & Elder, G.H. (2002). Revisiting age identity. Journal of AgingStudies, 16, 169–176.

Kwan, C.M.L., Love, D., Ryff, C.D., & Essex, M.J. (2003). The role of self-enhancing evaluations in successful life transition. Psychology and Aging, 18,3–12.

Lang, F.R., & Carstensen, L.L. (2002). Time counts: Future time perspective,goals and social relationships. Psychology and Aging, 17, 125–139.

Markides, K.S., & Boldt, J.S. (1983). Change in subjective age among theelderly. The Gerontologist, 23, 422–427.

Markides, K.S., & Ray, L.A. (1988). Change in subjective age among the elderly:An eight-year longitudinal study. Comprehensive Gerontology Bulletin, 2, 11–15.

Mathur, A., Barak, B., Zhang, Y., & Lee, K.S. (2001). Cross-cultural procedureto assess reliability and measurement invariance. Journal of Applied Measure-ment, 2, 241–255.

Mathur, A., & Moschis, G.P. (2005). Antecedents of cognitive age: Replicationand extension. Psychology and Marketing, 22, 969–994.

Mezulis, A., Abramson, L., Hyde, J., & Hankin, J. (2004). Is there universal posi-tivity bias in attributions? Meta-analytic review of individual, developmental,and cultural differences in the self-serving attributional bias. PsychologicalBulletin, 130, 711–747.

Montepare, J.M. (1996a). Actual and subjective age-related differences inwomen’s attitudes toward their bodies across the life span. Journal of AdultDevelopment, 3, 171–183.

Montepare, J.M. (1996b). Variations in adults’ subjective ages in relation tobirthday nearness, age awareness, and attitudes toward aging. Journal of AdultDevelopment, 3, 193–203.

Montepare, J.M. (1996c). An assessment of adults’ perceptions of their

psychological, physical, and social age. Journal of Clinical Geropsychology, 2,117–128.

Montepare, J.M., & Clements, A.E. (2001). “Age schemas”: guides to process-ing information about the self. Journal of Adult Development, 8, 99–108.

Montepare, J.M., & Lachmann, M.E. (1989). “You’re only as old as you feel”:Self-perceptions of age, fears of aging, and life satisfaction from adolescenceto old age. Psychology and Aging, 4, 73–78.

Moschis, G.P., Lee, E., Mathur, A., & Strautman, J. (2000). The maturingmarketplace: Buying habits of baby boomers and their parents. Westport, CT:Quorum Books.

Mutran, E., & George, L.K. (1982). Alternative methods of measuringrole/identity. Social Forces, 60, 866–876.

Myers, D.G., & Diener, E. (1995). Who is happy? Psychological Science, 6, 10–19.Newby, R., Watson, J., & Woodliff, D. (2003). SME survey methodology:

Response rates, data quality, and cost effectiveness. Entrepreneurship Theoryand Practice, 28, 163–172.

Öberg, P. (1996). The absent body – a social gerontological paradox. Ageing andSociety, 16, 701–719.

Öberg, P., & Tornstam, L. (2001). Youthfulness and fitness – Identity ideals forall ages? Journal of Aging and Identity, 6, 15–29.

Scheier, M., & Carver, C. (1985). Optimism, coping, and health: Assessmentand implications of generalized outcome expectancies. Health Psychology, 4,219–247.

Sedikides, C., Gaertner, L., & Toguchi, Y. (2003). Pancultural self-enhancement.Journal of Personality and Social Psychology, 84, 60–79.

Sedikides, C., Herbst, K.C., Hardin, D.P., & Dardis, G.J. (2002). Accountabilityas deterrent to self-enhancement: The search for the mechanisms. Journal ofPersonality and Social Psychology, 83, 3.

Silvia, P.J., & Duval, T.S. (2001). Predicting the interpersonal targets of self-serving attribution. Journal of Experimental Social Psychology, 37, 333–340.

Sirgy, M.J., Mentzer, J.T., Rahtz, D.R., & Meadow, H.L. (1991). Satisfactionwith health care services, consumption, and life satisfaction among the elderly.Journal of Macromarketing, 11, 24–39.

Staats, S. (1996). Youthful and older biases as special cases of self-age optimiz-ation bias. International Journal of Aging and Human Development, 43,267–276.

Staats, S., Heaphey, K., Miller, D., Partlo, C., & Romine, N. (1993). Subjectiveage and health perceptions of older persons: Maintaining the youthful bias insickness and in health. International Journal of Aging and Human Development,37, 191–203.

Stephens, N. (1991). Cognitive age: Useful concept for advertising? Journal ofAdvertising, 20, 37–48.

Szmigin, I., & Carrigan, M. (2001). Time, consumption, and the olderconsumer: An interpretive study of the cognitively young. Psychology andMarketing, 18, 1091–1116.

Taylor, S.E., Lerner, J.S., Sherman, D.K., Sage, R.M., & McDowell, N.K.(2003). Portrait of the self-enhancer: Well adjusted and well liked or malad-justed and friendless? Journal of Personality and Social Psychology, 84, 165–176.

Van Auken, S., Barry, T.E., & Bagozzi, R.P. (2006). Cross-country constructvalidation of cognitive age. Journal of the Academy of Marketing Science, 34,439–455.

Wieland-Eckelmann, R., & Carver, C.S. (1990). Dispositionelle Bewaltigungs -stile, Optimismus und Bewaltigung: Ein interkultureller Vergleich. Zeitschriftfür differentielle und diagnostische Psychologie, 3, 167–184.

022-031 099487 Teuscher (Q8D):210 x 280mm 18/12/2008 14:16 Page 31