Personality and Individual Differences 47 (2009) 209–213

Contents lists available at ScienceDirect

Personality and Individual Differences

journal homepage: www.elsevier .com/locate /paid

Are there gender differences in work ethic? An examination of the measurementequivalence of the multidimensional work ethic profile

John P. Meriac a,*, Taylor L. Poling b, David J. Woehr b

a University of Missouri-St. Louis, St. Louis, MO 63121-4499, United Statesb The University of Tennessee, United States

a r t i c l e i n f o

Article history:Received 17 April 2008Received in revised form 28 January 2009Accepted 2 March 2009Available online 1 April 2009

Keywords:Work ethicMWEPGenderDifferential functioningMeasurement equivalence

0191-8869/$ - see front matter � 2009 Elsevier Ltd. Adoi:10.1016/j.paid.2009.03.001

* Corresponding author.E-mail address: meriacj@umsl.edu (J.P. Meriac).

a b s t r a c t

Previous research has indicated that males and females differ on their reported levels of work ethic. How-ever, previous studies have relied upon work ethic inventories with limited generalizability, and no studyhas evaluated the invariance of measures. This study examined measurement invariance by exploring thedifferential item and test functioning of one work ethic inventory, the multidimensional work ethic pro-file (MWEP; Miller, Woehr, & Hudspeth, 2002) for male and female respondents. Results did not indicatethat the MWEP functioned differently by gender at the test or item level. Hence, work ethic as measuredby the MWEP does not carry different socially constructed meanings for men versus women. We alsoconclude that in contrast with previous research, women do not have a higher level of work ethic thanmen.

� 2009 Elsevier Ltd. All rights reserved.

1. Introduction tied to any one set of religious beliefs. Miller et al. identified seven

The ‘‘changing nature of work” has become a ubiquitous themein the study of organizational behavior. Work roles and behaviorsare now typically associated with a state of flux more so than a pre-dictable set of tasks. As a result, organizations have become inter-ested in identifying employees who are committed to the inherentvalue of work in general (i.e., work ethic). By focusing on workethic, organizations aim to build a workforce that will proactivelyengage and persist in behaviors that promote the effectiveness ofthe organization over time, tasks, and situations (Ryan, 2002). Inparticular, the assessment of work ethic has become increasinglyimportant as it allows organizational decision-makers to buildand sustain a motivated and diligent labor force. For instance, ina study of American managers, Flynn (1994) reported that fornearly 60% of respondents, work ethic was the top-ranked factorwhen hiring administrative employees.

Based on previous literature and original empirical research,Miller, Woehr, and Hudspeth (2002) posited that work ethic isnot a unitary construct, but a constellation of attitudes and beliefspertaining to work behavior. Specifically, they state that work ethic:(a) is multidimensional; (b) pertains to work and work-relatedactivity in general, not specific to any particular job (yet may gen-eralize to domains other than work); (c) is learned; (d) refers to atti-tudes and beliefs (not necessarily behavior); (e) is a motivationalconstruct reflected in behavior; and (e) is secular, not necessarily

ll rights reserved.

conceptually distinct (i.e., divergent) dimensions that comprise thework ethic construct: Centrality of work, self-reliance, hard work,leisure, morality/ethics, delay of gratification, and wasted time(see Table 1). This measure displays strong relationships betweenthe dimensions of work ethic and work attitudes and outcomes.Specifically, Miller et al. (2002) reported that the MWEP subscalesdemonstrated significant multiple correlations with job satisfaction(R = 0.50), job involvement (R = 0.65), and organizational commit-ment (R = 0.48), as well as a multiple R of 0.37 between MWEPdimensions and supervisory performance ratings.

The results of several additional studies have supported positiverelationships between work ethic and work-related outcomes. Mer-rens and Garrett (1975) revealed a relatively large effect size for theimpact of work ethic on time spent on task (d = 1.10) and workcompleted (d = 1.58). Similarly, Greenberg (1977) found a moderateeffect of work ethic on task persistence (x2 = 0.47). Finally, Blood(1969) reported a mean correlation of 0.17 between work ethicand job satisfaction. However, most work ethic inventories usedin this stream of research are unidimensional, single-scale invento-ries. Subsequently, researchers have questioned the mismatchbetween the theoretical multidimensional conceptualization ofwork ethic and the widespread use of unidimensional scales whichhave limited implications for interpretation and validity.

1.1. Gender differences in work ethic

Many studies have reported gender differences in work ethic(e.g., Furnham & Muhiudeen, 1984; Hall, 1990, 1991; Hill, 1997;

Table 1MWEP dimensions, dimension definitions, and sample items.

Dimension Definition Sample items

Centrality of work Belief in work for work’s sake and the importance of work Even if I inherited a great deal of money, I would continue to work somewhereIt is very important for me to always be able to work

Self-reliance Striving for independence in one’s daily work I strive to be self-reliantSelf-reliance is the key to being successful

Hard work Belief in the virtues of hard work If you work hard you will succeedBy simply working hard enough, one can achieve their goals

Leisure Pro-leisure attitudes and beliefs in the importance of non-workactivities

People should have more leisure time to spend in relaxationThe job that provides the most leisure time is the job for me

Morality/ethics Believing in a just and moral existence People should be fair in their dealings with others.It is never appropriate to take something that does not belong to you

Delay of gratification Orientation toward the future; the postponement of rewards The best things in life are those you have to wait forIf I want to buy something, I always wait until I can afford it

Wasted time Attitudes and beliefs reflecting active and productive use of time I try to plan out my workday so as not to waste timeTime should not be wasted, it should be used efficiently

210 J.P. Meriac et al. / Personality and Individual Differences 47 (2009) 209–213

Miller, 1980; Petty & Hill, 1994; Wollack, Goodale, Witjing, &Smith, 1971). Most have reported higher work ethic scores for wo-men than men. For instance, Spence and Helmreich (1983) andKirkcaldy, Furnham, and Lynn (1992) found evidence to supportthe tendency for women to obtain higher mean scores than menwith respect to work ethic across occupations. This finding wasreplicated across the vast majority of countries examined in a largeinternational study conducted by Lynn (1991). These studies allfailed to recognize the multifaceted nature of work ethic. Anotherissue however is a lack of information on the measurement equiv-alence/invariance (MEI) of the instrument used to draw conclu-sions regarding gender differences. As noted by Hill (1997),gender-role stereotypes may lead to different approaches to workethic endorsement for males and females, such that the work ethicconstruct may have different socially constructed meanings formen versus women. Without an analysis of the multiple dimen-sions that compose the work ethic construct, our understandingof potential differences is limited. Further, without evidence ofMEI, whether these findings reflect actual differences among menand women in espoused work ethic or simply different interpreta-tions of the construct remains unclear (Vandenberg & Lance, 2000).

In our review of the literature, we did not find any studies thatinvestigated the MEI between men and women with respect towork ethic measures, even though as noted earlier, failure to doso makes subsequent conclusions dubious. Typical approaches tothe analysis of MEI are based on classical test theory and includemean differences across groups, relationships with external vari-ables, internal covariance differences across item responses, ormulti-group confirmatory factor analysis (CFA). Vandenberg andLance (2000) advocated the multi-group CFA method as a usefulapproach. This method is based on classical true score theory,where variance in observed scores is comprised of true score vari-ance and error variance.

Another popular approach to examining MEI is item responsetheory (IRT; Embretson & Reise, 2000; Hambleton, Swaminathan,& Rogers, 1991). One primary distinction among these methodsis that IRT evaluates psychometric properties at the item-level,whereas CFA approaches primarily focus on the test as a whole.If an item-level examination is of interest, then IRT provides aclearer picture of test properties over CFA approaches. Also, CFAmethods provide indices of item characteristics such as difficulty(p-values) and discrimination (item-total correlations) that aredependent on group differences in traits (Raju & Ellis, 2002). IRTparameter estimation techniques yield values that are independentof the group on which the set of items was administered. In otherwords, they allow for the control of group differences in abilitywhen estimating parameters (difficulty and location) so they arenot confounded with group differences (Raju & Ellis, 2002).

IRT also takes a different approach toward the examination ofMEI than CFA approaches, where IRT approaches view invariancein terms of ‘‘differential functioning” of items (DIF) and tests(DTF). When items demonstrate invariance across two groups, theypossess invariant item parameters (e.g., difficulty and discrimina-tion) given the same level of the latent trait. However, when indi-viduals possess the same level of the trait or ability and havedifferent probabilities of obtaining expected scores on an item,the item displays DIF (Hambleton et al., 1991; Hulin, Drasgow &Parsons, 1983). Specifically, DIF refers to group differences in theirresponses to test items when their ability (or trait-level) is heldconstant. Differential test functioning is the extension of this samephenomenon to test-level differences in response tendencies (Raju& Ellis, 2002). Raju, van der Linden, and Fleer (1995) proposed amethod for evaluating the ‘differential functioning of items andtests’ (DFIT). This approach further offers two indices of DIF: com-pensatory (CDIF) and non-compensatory (NCDIF). The NCDIF indexrepresents the average squared difference between the two sub-groups’ item-level true scores, and CDIF represents the relative im-pact in test-level functioning of individual items (Raju et al., 1995).

In other areas of individual differences research, DIF has beenidentified across male and female respondents. For instance, Reise,Smith, and Furr (2001) found gender DIF on items on several facetsof the NEO PI-R neuroticism scale. The implications for such find-ings are two fold: (a) gender by item content interactions may bepresent, and (b) comparisons of mean levels of males and femalesmay be inappropriate to the extent that items do not function thesame way for both subgroups. In the present study, we seek toexamine whether MWEP items function in the same manner formales and females.

1.2. Implications and contribution of current study

In sum, previous work on gender differences in work ethic hasbeen marked by construct deficiency coupled with a lack of mea-surement equivalence information regarding differences betweenmale and female respondents. To address these issues, the currentpaper presents MEI information between men and women with re-spect to each of the dimensions of the MWEP, as well as a compar-ison of differences using a multidimensional work ethic inventory(i.e., the MWEP).

2. Method

We examined DIF and DTF of the MWEP between responsesfrom 1122 men and 828 women. Data were gathered from 1996to 2002 in the United States, in both industrial (25.91%) and

J.P. Meriac et al. / Personality and Individual Differences 47 (2009) 209–213 211

university student samples (74.09%). The demographic composi-tion of the subgroups was comparable, where males were on aver-age 27 and females were 23 years old. All respondents had at leasta high school diploma, and in the male subgroup 6% had a bache-lor’s degree, and 1% had a master’s degree compared with 2% andless than 1% of females, respectively. In terms of race, 10% of maleswere African–American, 2% Asian, 78% Caucasian, and 8% Hispaniccompared with 18%, 2%, 72% and 7% for females, respectively. Fi-nally, 18% of males were Catholic, 48% Protestant, and 1% Jewishcompared to 21%, 37% and 1% of females, respectively. Hence, thegender subgroups were quite similar in terms of demographiccharacteristics.

2.1. IRT parameter estimation and DIF analyses

Unidimensionality is an assumption of the IRT model used inthe current analyses. If this assumption is not reasonably met, thenDIF analyses can be spurious and potentially result from additionallatent factors, rather than actual DIF. Unidimensionality is oftenevaluated using a full-information exploratory factor analysis ap-proach, using principal axis factoring. Here, we examined the uni-dimensionality of each MWEP subscale for men and womenseparately by examining the scree plot and percentage of varianceexplained by each component. Reckase (1979) proposed that if atleast 20% of the variance is explained by the first component, thena scale can be interpreted as effectively unidimensional.

Next, we proceeded to estimate item parameters using MULTI-LOG 7.03 (Thissen, 2003). MULTILOG employs polytomous (i.e.,three or more response options) IRT models which are applicablefor such data. In the current example we used Samejima’s (1969)graded response model (GRM) as the MWEP utilizes a Likert-typerating scale with five response options (Embretson & Reise,2000). For the comparison of item parameters across groups, it isnecessary to convert item parameters so that parameters for bothsubgroups are on the same metric. To equate parameters, we em-ployed the procedures developed by Stocking and Lord (1983),conducted using the EQUATE 2.1 program (Baker, 1995). Equatingparameters is an iterative process that necessitates choosing itemsthat do not exhibit DIF to serve as ‘linking’ items. Hence, theprocess begins by using all items in the subscale as linking items,

Table 2DIF and DTF estimates for MWEP subscales.

Item Self-reliance Morality/ethics Leisure Hard wor

NCDIF1 0.001 0.000 0.010 0.0062 0.003 0.006 0.000 0.0053 0.001 0.001 0.006 0.0024 0.015 0.001 0.008 0.0185 0.005 0.001 0.004 0.0046 0.002 0.004 0.002 0.0167 0.005 0.000 0.005 0.0028 0.005 0.000 0.020 0.0079 0.005 0.001 0.010 0.00410 0.009 0.000 0.001 0.001

CDIF1 0.003 �0.000 0.001 0.0032 0.006 0.003 0.000 0.0043 0.003 �0.001 0.002 0.0014 0.001 0.001 �0.008 �0.0005 �0.005 0.001 0.002 0.0026 0.005 0.001 0.004 0.0027 0.005 0.000 0.007 0.0038 0.007 0.001 �0.002 0.0029 �0.002 �0.001 0.008 0.00110 �0.006 0.000 0.002 0.000

DTF0.0173 0.0043 0.0159 0.0181

identifying items that exhibit DIF, and then re-scaling items usingthe non-DIF items to obtain linking constraints.

Finally, we evaluated differential functioning using Raju et al.’s(1995) DFIT procedure. The DFIT procedure is conducted using aprogram written by Raju (1999), which provides informationregarding the differential functioning of items and tests, and pro-vides indices for CDIF and NCDIF. The NCDIF index utilizes a cutoffvalue to determine whether items exhibit DIF. In the current anal-yses, the cutoff value of .096 was used for determining whether DIFwas present since the MWEP items were rated on a five-pointscale, based on recommendations by Raju (1999). CDIF valuescumulatively add to determine the DTF value. The DTF cutoff isdetermined by multiplying this cutoff score by the number ofitems, such that the cutoffs for DTF for five of the MWEP subscaleswere .960 (since they contain 10 items), .768 for the subscale witheight items (wasted time), and .672 for the subscale with sevenitems (delay of gratification).

After we examined the MWEP for DIF and DTF, we sought toexamine mean differences between men and women in their sub-scale scores. Differences were examined by conducting a one-wayANOVA across subgroups on each of the seven subscales. We alsocalculated the Cohen’s d effect size of the difference for each sub-scale since even small practical differences were likely to be statis-tically significant given our large sample.

3. Results

A visual examination of the scree plot for each dimension foreach subgroup indicated that one component explained most ofthe variance in each subscale analysis. The average percentage ofvariance explained by the first component was 59.74% (min =45.49%, max = 71.00%) for males and 56.75% (min = 35.69%, max =84.13%) for females, with the minimum being 35.70% and the max-imum being 84.13%. In each instance, the percentage of varianceexplained was well beyond Reckase’s (1979) 20% rule of thumb.Further, in comparison with the second component, it is evidentthat a dominant factor emerged in each of the analyses. Specifi-cally, in all cases there was a large ratio of the first to second eigen-value for males (mean = 7.50) and females (mean = 6.80). Hence,

k Centrality of work Wasted time Delay of gratification

0.002 0.013 0.0100.007 0.008 0.0000.008 0.016 0.0200.015 0.004 0.0000.007 0.011 0.0020.010 0.007 0.0080.009 0.004 0.0020.006 0.0170.0040.007

0.002 0.017 �0.0140.006 0.014 0.004�0.003 0.011 0.0220.008 �0.000 �0.0000.003 0.009 0.002�0.003 �0.003 0.014�0.006 �0.006 0.0050.000 �0.0130.0040.007

0.0179 0.0303 0.0337

Table 3Means, standard deviations, d, and Fvalues for the MWEP subscales.

Subscale Total Men Women d F-value

M SD M SD M SD

Self-reliance 29.14 (8.04) 28.85 (8.00) 27.81 (8.27) 0.13 7.70**

Morality/ethics 27.86 (15.24) 26.20 (15.02) 24.09 (14.96) 0.14 9.43**

Leisure 29.08 (6.26) 29.65 (6.49) 28.79 (5.89) 0.14 8.97**

Hard work 29.14 (11.27) 25.99 (11.54) 25.24 (10.65) 0.07 2.14Centrality 28.89 (9.59) 28.16 (9.97) 26.78 (9.25) 0.14 9.63**

Wasted time 23.07 (7.51) 22.46 (7.86) 21.67 (7.32) 0.10 5.08*

Delay of gratification 19.19 (6.41) 18.59 (6.73) 18.09 (6.02) 0.08 2.93

* p < 0.05.** p < 0.01.

212 J.P. Meriac et al. / Personality and Individual Differences 47 (2009) 209–213

we proceeded to conduct the IRT analyses. The item parameterswere used as the input for the EQUATE and DFIT procedure. Param-eter estimates are available from the first author upon request.

3.1. DIF and DTF analyses

Results of the DFIT analyses indicate that no items were identi-fied as exhibiting DIF in any of the seven subscales. Further, none ofthe subscales exhibited DTF. The DIF and DTF values are presentedin Table 2. As none of the NCDIF indices were greater than the .096cutoffs recommended by Raju (1999), they were not flagged forDIF. Subsequently, none of the subscales exhibited DTF. These re-sults indicate that response tendencies are the same for both malesand females.

3.2. Examination of mean differences

Table 3 presents descriptive statistics and gender differences inthe mean scores expressed as d, and the F-values for the differencesas a test of statistical significance. These results demonstrate sig-nificant mean differences between men and women on five ofthe seven MWEP subscales. Differences were significant for self-reliance, morality/ethics, leisure, centrality of work, and wastedtime. Contrary to previous findings discussed above, men hadslightly higher means than women for these subscales. However,our effect size calculations show that the differences had littlepractical value as all values were well below the ‘small’ effect sizestandard associated with d (i.e., all effect sizes were below 0.20).

4. Discussion

4.1. Summary and conclusions

The results of the current study suggest that male and femalerespondents have the same response tendencies to MWEP items.Specifically, DFIT analyses did not show that the MWEP items func-tioned differently by gender at the test or item level. Given theinvariance of items by gender, the work ethic construct as mea-sured by the MWEP does not carry different socially constructedmeanings for men versus women. Thus, the perspective that menand women have been socialized to display different attitudes to-ward the value of work due to the different meanings theseattitudes have with respect to gender roles does not hold. Further-more, since we demonstrated that the MWEP displays measure-ment equivalence, we can conclude that when work ethic has thesame substantive meaning across genders, women do not have ahigher work ethic than men.

Effect sizes of the differences between males and females for allseven subscales were quite small. However, for all seven subscales,the pattern of mean differences was the same: Men demonstratedhigher mean scores than women. These findings contrast with

many studies in the literature, which reported higher mean scoresfor women compared to men on single-scale work ethic measures.Although actual differences were small across scales (.5–2 points),when distinctions are being made between individuals, even smalldifferences can have important implications for all parties involved(e.g., selection decisions, promotions, terminations, etc.). Impor-tantly, our analytic approach and findings with respect to theMWEP support that observed differences reflect true differencesand not measurement biases. Moreover, since these results werebased on IRT methodology, the measurement invariance cannotbe attributed to sample dependence, which is a major limitationof CFA approaches. Thus, researchers and organizational decisionmakers can have confidence in the use of the MWEP as a general-izable assessment tool.

A key contribution of the present research is its evaluation ofDIF and DTF. These IRT-based analyses represent an appropriateassessment of MEI between men and women for the MWEP thatprevious research on gender differences in work ethic has ne-glected to examine. This lapse potentially undermines the qualityand validity of the conclusions reached. To reiterate the perspec-tive of Vandenberg and Lance (2000), such neglect is akin to ignor-ing the importance of reliability and validity evidence. Bycapitalizing on the benefits of IRT, we have effectively demon-strated MEI for the MWEP.

Theoretical rationale for why gender differences have been evi-dent in previous studies has been sparse. However, one perspectivethat has been offered is rooted in social cognitive theory (SCT).Bandura’s (1991) SCT focuses on the dynamics of self-efficacy, out-come expectations, contextual factors, and social contexts in deter-mining behavior. Hill (1997) suggested that gender-rolestereotypes may account for differences in the endorsement ofwork ethic between men and women. In particular, the images ofwork portrayed by the media present most occupational informa-tion through male role characterizations (Signorielli, 1993), and of-ten depict work as a necessary evil and something people enjoyescaping from. As a result, men are socialized to think about theendorsement of work as a negative construct. On the other hand,starting in their formative years, females are treated differentlythan males; that is, females tend to be attributed with appropriatebehaviors (Sadker & Sadker, 1994). These appropriate behaviorsassociated with females are very similar to many of the character-istics associates with work ethic endorsement (Hill, 1997). Thus,females may interpret the endorsement of the work-ethic con-struct as a means to endorse their gender-role. In sum, this per-spective asserts that differences in response tendencies mighthave been expected, as the American male gender-role depictswork as something men enjoy escaping from, and that womeninterpret work ethic as a construct that should be endorsed in or-der to align with the expectations, socialized since childhood, forfemales to depict appropriate and positive attitudes and behavior.

In addition, social cognitive career theory (SCCT; Brown & Lent,1996) is an extension of SCT that describes how these factors

J.P. Meriac et al. / Personality and Individual Differences 47 (2009) 209–213 213

influence work-related attitudes and choices, and has been offeredas another related reason for explaining findings of higher levels ofwork ethic among women relative to men (Hill & Rojewski, 1999).This view suggests that when faced with the widespread percep-tion of a ‘glass ceiling’ hindering advancement of women in theworkplace, women may respond by embracing the belief in theirability to succeed and overcome barriers by working hard and fos-tering the expectation that persistent hard work will bring good re-sults. According to this explanation, women and men interpret themeaning of work ethic the same way, but environmental barriershave made a higher degree of hard work necessary for theadvancement of women in the workplace. Thus, the SCCT-basedexplanation asserts that men and women interpret the work ethicconstruct similarly, but for women work environment barriershave caused them to place a higher value on work ethic in orderto succeed. Hence, the items used to measure work ethic shouldtap the same latent constructs as the present study demonstrates,but the mean levels of endorsement among women should behigher than men. Specifically, in the present study, one might ex-pect higher levels of the MWEP dimension hard work for females.However, the present study also found no support for this idea, aseffect sizes were not practically different between males and fe-males. Further, males actually had slightly higher scores than fe-males across all work ethic dimensions.

Future research on the MWEP should extend these techniquesto examine other potential subgroup differences of interest (e.g.,generational differences or race). Although we found no evidenceof differential functioning based on gender, it is possible that othersubgroups may have different levels of work ethic or may have dif-ferent socially constructed meanings of work ethic based on otherfactors. Moreover, previous gender differences reported with re-spect to single-scale instruments should be viewed as suspect untilappropriate evaluations of MEI have been conducted. Attemptsshould be made to evaluate the invariance of questionnaires thathave been used in previous gender comparison studies, as thisstudy only evaluates the invariance of the MWEP. It is possible thatthe item content of other measures may result in differential func-tioning for male and female respondents. For instance, Furnhamand Rajamanickam (1992) reported differences for males and fe-males; based on a review of item content, they concluded thatthese differences may be a function of religious content. As theMWEP was designed to avoid religion-based work ethic, perhapsthis feature has reduced differences in response tendencies formales and females. Additionally, the implications of the smallbut real subscale differences between men and women revealedin this study deserve attention. The extent to which these differ-ences may contribute to gender differences in administrative deci-sions is a question that must be addressed by future research.

References

Baker, F. (1995). EQUATE 2.1: Computer program for equating two metrics in itemresponse theory (Version 2.1). Madison: University of Wisconsin, Laboratory ofExperimental Design.

Bandura, A. (1991). Social cognitive theory of self-regulation. OrganizationalBehavior and Human Decision Processes, 50, 248–287.

Blood, M. R. (1969). Work values and job satisfaction. Journal of Applied Psychology,53, 456–459.

Brown, S. D., & Lent, R. W. (1996). A social cognitive framework for career choicecounseling. The Career Development Quarterly, 44, 354–366.

Embretson, S. E., & Reise, S. P. (2000). Item Response Theory for Psychologists.Mahwah, NJ: Lawrence Erlbaum and Associates.

Flynn, G. (1994). Attitude more valued than ability. Personnel Journal, 73, 16.Furnham, A., & Muhiudeen, C. (1984). The protestant work ethic in Britain and

Malaysia. Journal of Social Psychology, 122, 157–161.Furnham, A., & Rajamanickam, R. (1992). The protestant work ethic and just world

beliefs in Great Britain and India. International Journal of Psychology, 27,401–416.

Greenberg, J. (1977). The protestant work ethic and reactions to negativeperformance evaluations on a laboratory task. Journal of Applied Psychology,62, 682–690.

Hall, G. S. (1990). Work attitudes of traditional and non-traditional technicalcommunity college students. Master of Science thesis, Knoxville: The Universityof Tennessee.

Hall, G. S. (1991). Do older college students have different attitudes about work ascompared with younger traditional students? Tennessee Education, 21, 27–29.

Hambleton, R. K., Swaminathan, H., & Rogers, H. J. (1991). Fundamentals of itemresponse theory. Newbury Park, CA: Sage Publications Inc.

Hill, R. B. (1997). Demographic differences in selected work ethic attributes. Journalof Career Development, 24, 3–23.

Hill, R. B., & Rojewski, J. W. (1999). Double jeopardy: Work ethic differences inyouth at risk of school failure. The Career Development Quarterly, 47, 267–280.

Hulin, C. L., Drasgow, F., & Parsons, L. K. (1983). Item response theory. Homewood, IL:Dow Jones-Irwin.

Kirkcaldy, B. D., Furnham, A., & Lynn, R. (1992). National differences in workattitudes between the UK and Germany. The European Work and OrganizationalPsychologist, 2, 81–102.

Lynn, R. (1991). The secret of the miracle economy. London: The Social Affairs Unit.Merrens, M. R., & Garrett, J. B. (1975). The Protestant ethic scale as a predictor of

repetitive work performance. Journal of Applied Psychology, 60, 125–127.Miller, D. (1980). Differences in the protestant work ethic values of selected freshman

and senior students at a land grant university. Unpublished doctoral dissertation,Oregon State University.

Miller, M. J., Woehr, D. J., & Hudspeth, N. (2002). The meaning and measurement ofwork ethic: Construction and initial validation of a multidimensional inventory.Journal of Vocational Behavior, 60, 451–489.

Petty, G. C., & Hill, R. B. (1994). Are women and men different? A study of theoccupational work ethic. Journal of Vocational Education Research, 19, 71–89.

Raju, N. S. (1999). DFITPS6: A Fortran program for calculating polytomous DIF/DTF[computer program]. Chicago: Illinois Institute of Technology.

Raju, N. S., & Ellis, B. B. (2002). Differential item and test functioning. In F. Drasgow& N. Schmitt (Eds.), Measuring and analyzing behavior in organizations: Advancesin measurement and data analysis (pp. 156–188). San Francisco, CA: Jossey-Bass.

Raju, N. S., van der Linden, W., & Fleer, P. (1995). An IRT-based internal measure oftest bias with implications for differential item functioning. AppliedPsychological Measurement, 19, 353–368.

Reckase, M. D. (1979). Unifactor latent trait models applied to multifactor tests:Results and implications. Journal of Educational Statistics, 4, 207–230.

Reise, S. P., Smith, L., & Furr, R. M. (2001). Invariance on the NEO PI-R neuroticismscale. Multivariate Behavioral Research, 36, 83–110.

Ryan, J. J. (2002). Work values and organizational citizenship behaviors: Values thatwork or employees and organizations. Journal of Business and Psychology, 17,123–132.

Sadker, M., & Sadker, D. (1994). Failing at fairness: How our schools cheat girls. NewYork: Touchstone.

Samejima, F. (1969). Estimation of latent ability using a pattern of graded scores.Psychometrika. Monograph supplement no. 17.

Signorielli, N. (1993). Television and adolescent’s perceptions about work. Youthand Society, 24, 314–341.

Spence, J. A., & Helmreich, R. L. (1983). Achievement related motives and behavior.In J. A. Spence (Ed.), Achievement and achievement motives: Psychological andsocial approaches. San Francisco, CA: Freeman.

Stocking, M. L., & Lord, F. M. (1983). Developing a common metric in item responsetheory. Applied Psychological Measurement, 7, 201–210.

Thissen, D. (2003). MULTILOG 7.03: A computer program for multiple, categorical itemanalysis and test scoring using item response theory. Chicago: Scientific Software,Inc.

Vandenberg, R. J., & Lance, C. E. (2000). A review and synthesis of the measurementinvariance literature: Suggestions, practices, and recommendations fororganizational research. Organizational Research Methods, 3, 4–69.

Wollack, S., Goodale, J. G., Witjing, J. P., & Smith, P. C. (1971). Development of thesurvey of work values. Journal of Applied Psychology, 55, 331–338.