Development of a Method of Measuring

7/30/2019 Development of a Method of Measuring

1/4

Section Four-Motivation, Attitudes, and Job Satisfaction

THE DEVELOPMENT OF A METHODOF MEASURING JOB SATISFACTION:THE CORNELL STUDIES *Patricia Cain Smith

CO~NELL TUDIES OF T OB SATISFACTION were initiated in 1959 with the-purpose of studying job satisfaction among a representative cross sectionof workers in the United States. One specific goal of these studies was torelate job satisfaction to nleasurable company and community character-istics and to characteristics of the individual worker. Since the success ofthe cntire study hinged on the nature and qualit y of the instrument to beusccl to measurc job satisfaction, a considerable amount of time and effortwent into constructing this device. This paper summarizes the rationalebehind the approach to measurement and the particular characteristics O Fthe measure finally adopted. Before turning to this, however, a briefdescription of what job satisfaction is and why it should be measured atall would seem to be in order.WHAT IS JOB SATiSFACTION?

25-The Development of a Method of Measuring Job Satisfaction 273or may not likc thcir jobs. Si~tisfactionand clissatislaction may or maynot result in ovcrt behavior (changes in productivity, gricvances, ab-sences, turnover, and so forth), depending upon the individual's person-aliry (whether he acts on or controls or represses his emotions), the op-portunities for self-expression on the job (closeness of supervision,company rules and regulations, and so forth), and thc othcr job altcrna-tives open to him (labor market for his particular skill, his financialcondition, and so forth). To repeat, job satisfaction is viewed primarilyas a consequence of job experience (and, in fact, high productivity mayproduce sati sfaction as much as the other way around). The causal cfficacyof job satisfaction is, then, problematic, rather than something to beassumed.However, there are still valid rcasons for wanting to study job satisfac-tion. Most obviously it can be viewcd as an end in itself. In fact, it is notreally meaningful to ask why pleasure or satisfaction are good or desir-able. They arc dcsirnblc by nature.Secondly, under certuitt circumstances job satisfaction and particularlyjob dissatisfaction may lend to overt bchavior which is of interest to


2/4

27 4 Section Four-Motivation, Attitudes, and Job Satisfaction

REQUIREMENTS OF A USEFUL MEASURE OF JOBSATISFACTIONA useful ineasorc of job satisfaction should be capable of bcing used

over a widc range of job clnssifications a nd wit11 pcoplc: of wr yi ng joblevels. I n othcr words, its content should b(: such that the meanings of thewords used arc con2rnon to workcrs on many cliffcrc~ltkinds of jobs, andthe verbal lcvcl should be low enough so that poorly cducatccl as wcll aswell-educated workers can un ders ta~l d he questions.

Practical considerations would demand that the measure be short,easily sdministn.ed (in groups ), and easily scorablc. Long, involvedmeasures with complicated scoring systems woulcl bc prccluclcd in alarge-scale study drle to time and financial considerations.

The measure should gone rate scores indicative of satisfaction with anumber of discrin~inablydifferent aspects of the work situation (pay,work, supervision, co-workers, and so forth). A measure of overa ll (global)job satisfaction may be sufficient for somc purposes b ut would be inade-quate for an intensive study aimed at identifying the relationships be-tween different aspects of the job situation an d individ ual and companycbaractcristics. The sa me variables may be re lated qu ite differently tosatisfaction with different aspects of tllc job, but these rclatio~>ships uddbe dilut ed if only a global measurc of satisfactioll were used.

Thc scale should llc frco from obvious biases, such as acquiesccncc-thc tcndcncy to "agree" with an iten1 indcpcnd cnt of item contcnt-SOthat pcople who agree with everything will not get artificially high scores.

The worker's f rame of reference, his standa rd of judgment, whenresponding to the items either should be taken into consideration whenconstructing and scoring thc measure or should he demonstrated not toaffect the answers markedly. To cite an obvior~sexample, if th e word''sinlple'' meant "good" to one worker but was interpreted as meaning"bad" by another worker, the item would not yicld uscful results.

The measu1.e should tlcmonstrate reliability: both internal consistency(agreement among items intended to measure the same thing) and stabil-ity over time in the same individual.

Finally, the measure should bc valid-it should measure what it isintcnded to measure.

THE CORNELL JOB DESCRIPTIVE INDEX (JDI)Areas Measured

To fulfill the criteria set out above, a numbel. of different types ofnlcasurcs were trivd. The onc, finally scttlccl upon mc:asurcd five area5 ofjob satisfactiun: satisfaction ~ i t b ork, satisfaction with pay, satisfac-

25-The Development of a Method of Measuring Job Satisfaction 27 5tion with opportunities for proinotion, satisfaction with supervision. andsatsfaction with co-workers. These categorics wcre arrived at afterconsidcrablc rcvicw of the factor-analysis literature on job satisfaction,and a fter an cxtcnsivc analysis of our own preliminary categories.

For cach area thcrc: is a list of acljcctivcs or short phrases, cach witha blank spacc hcsidc it. The respondent is jnstructccl to show how wclleach word or phrase describes thc aspcct of his job in question (forinstance. his pay). If a word describes the pay on his pc sc nt jol ~ or hissi~l~c~.vision,uid so forth) , hc is instructed to writc t l ~ c :1ottc:r "Y" for "Ycs"bcsidc that word or phrase. If the word does not dasc rihc his presentpay (or supcr~~ision,ncl so for th) , he is asked to writc "N" for "No" besidethat word or phmsc. If 11c cannot dccide, hc is asked to placc a "?" in theblank for "Cannot dcci dc." The scales for cach of th e fiva job ar eas arcshown in Table 25-1.

Developing the Sc oring ProcedureTo investigate the bcst scoring procedures to use, in our early studies

thcse scales were administered to each person a second and a third time.The second time each person was asked to dcscribe the "best" job he couldthink of or the "bcst" job hc l ~ dwc ~ ad. Tho third timo thv individiialwas asked to describc th c "worst" prcvious job he had had or could thinko f This madc it l~ossiblc o score the qoastion~l:~ircsn four diffcrcnt ways:1. Satisfaction could be infcr rcd from thc similarity of his rcsponseswhcn describing his present job to his rcsponses wllcn describing his"bcst" job; th at is, if he described his best an d present jobs in the sameway, one could infer th at he was satisfied with the job.

2. Satisfaction could be infe rred from the dissimilarity of his rcsponscswhen describing his present job to his responses whcn drscribing his"worst" job; tha t is, if he described his present and worst jobs complctelydifferen tly, it could be infcrr cd th at the ind ividual was satisfied with hisjob.

3. Satisfaction could be infcrrccl simultaneousl~~rom thc similarity ofhis present job responses to his 'best" job responses and their dissiini-la it y to his "worst" job responses. F or instancc, if a person described hisbest and present jobs as "challenging," whereas he described his worstjob as not challenging, it could be inferred t hat h e was satisfied with hisjob in that respect. On the other hand , if he descr ibed his prescnt, best,and worst jobs as "challenging," satisfaction would not bc inferred fromthat item; in fact. the item wo~dcl ot be used in scoring at all, 1)ecausc itwould indicate that for him that cliamctcristic was not important.

4. Satisfaction could b(, inferrcd from dircct, a priori scoring of theitems under tlic assumpt ion that 11io.st inclividuals would intcl prct theitem in th e samc way and would see thc same things as desirable and un-


3/4

'76 Section Four-Motivation, Attitudes, and Job Satisfactionlesirable on a job; tha t is, it would be assumed that all people wouldGee a "challenging" job as des irable.The rationale behind the first three scoring methods was to eliminateIre possible effects of dif ferent frames of ref erence of diRercnt individualsn answering the itcms; that is, to control for t l l ~act that sonle peoplenight not see "challenging" work as a good thing. The question then>ecame how to decide which of these scoring methods was the bes t. Sincethere are no behavioral or perf orman ce criteria of job satisfaction (recallthat our definition of satisfaction indicated that it was a response to thework situati on, not a determ inant or cause of performance), other meth-ods of co mparing the validity of th e various scales had to b e used.An alternative method of validation in a situation like this is to choosethe method that ( a ) is most represen tative of a ll the methods used, thatmost coilsistently agrees with tlre uther inethods, and (G) shows the bestiiiscri~ninntion mong the diff eren t job areas. For ins tancr, if a iuethodyicldcd scores oil tho fivc arcas of job s;~Maction work, ply, promo-tions, sopervision, and co-workers) and all these scores currelated veryhighly with each other, hut did not correlntc at all with scores on the sameareas as mcasured by other methods, we would conclude that this particu-lar measure of satisfaction contained some special bias and was not veryuseful.Using tllcsc criter ia, it wils found that thc a priori scoring of the JD Iscales gavc the best rcsults; that is, it yiclded scores that agrced mosthighly with scores dcrived from other scoring methods and yielded theclearest discriniination or independence among thc five job-area scales.The direct JD I measures also correlated highly with several entirelydifferent sets of measures whic h asked the individual to rate his jobsatisfaction directly, which gave added credence to its validity.Thus, the TDI yields five scores, one for each scale. These scores areobtained. by add ing up the num ber of responses with in eac h scale, accord-ing to the kcys provided in Table 25-1.Selection of Items

T l ~ rtcms for the JDI were selected by a thee-stage proccss. First,items were selected from 0 t h nventories and by conimon sense whichseemed to be relevant to cach of thc five area scalcs . This original searchyielded from 30 to 40 itcms per scale. Ncxt we looked at how frequentlyeach item was usrd to dexribe "bcst" and "worst" jobs. Items which wereused equally frcquently to describe "best" and "worst" jobs were dis-carded on the grounds tlrst they probably wcic not important in deter-mining job satisfaction. Finally, the scales wcrc administered to severalsaniplcs of cmployerr and t he subjects were div ided (on each scale) intoa "satisfied" half and n "dissatisficd" half on thc basis of thcir total scores011 each scalr Proportional differcnccs in itcw responses bctwcen high

25-The Development of a Method of Measuring Job Satisfaction 277TABLE 25-1Items in Fi nal Version of JD I *

Each of the five scales was prescnted on a separate page.Th e instru clions Tor each scale asked the su bject to pu t "Y" bcside all item ~f he iten,described the part icular aspec t o f Ius job (work, pay, and so forth ) , "Nu if t h e i f e m d id n o tdcscribe lha t aspec t , or "1" if he could not decide-..The rcsponse sho wr~ eside each i tem is the one scored in the "salisfied" directio n for

each scale.Work

Y Fascinating-R o u t in e- Satisfying- Boring-G o o d~-Creative- Rcspccted-Ho t-Plcasa n-Useful-Tiresome- Healthful\ Challenging-On your fee t-Frustra t ing-Simple-N Endless- ives sense of

Y accomplishment-

SupervisionY Asks my adviceN Hard to pleaseN Impoli te- Praites good rvor k--Y Tactfrd- Influential- t l p l o d a e

Doesn't superviscN cnuugh- Quick-tempered

Tells me where IY standPN Annoying- Stu b b o rn--Y Knows job well--N Bad- Intelligent-

eaves me on m yY ow n-

PeopleY Stimula t ingN Boring-Slow--Y Ambit ious-- .-N Stnpid- Responsi l~ lc--Y Fnsl- Intelligent- a sy t o m a k eN enemies- Ta lk t o o m u c h- Y Sm a r t--N Lazy- Unpleasant-N o p r iw c yY Active-

Narrow interests- Loyal-Hard to meet-ro u n d w h e nY needed

Income adequate for normalY expenses- bornor onsGood opporluni ty for advance-Y m e n t-Satisfac tory profi t sharing- N O p p o r tu n i t y ~ o r r~ e w l l a limileilN Barely live on incom e- Y Promotion on abi l i ty-Bad- N Dead-end job-Income provides luxuries- Y Good chance for promotion-Insecure- N Unfa ir pron ~otio n ol icy--Less than I deservew N Infrequent prom otions-Highly paid- Y Regular y roniot ions-Underpaid .- Y Fairly good chance fur promotion-crmissioll 10 I S C hcsc cm~m hoilld hr obfaiocd from Dr. pat ric ia r. rnirh. cpL, psvchnlop,.,owli~ r e e n Slate U n wc r s i t y . Doalinp ~ r c ~ n ,h,,,, 43403 .


4/4

27 8 Section Four-Motivation, Attitudes, and Job Satisfactionand low halves in each sample were computed for each item. Only itemswere retained which show ed a clear diffcrentiatio n between satisfied anddissatisfied workers. This process was repeated over five different samplesof workers and only those items which showed consistcnt discriminationwere retained.

The final p y and promotion scales included nine items each and thework, supervision, and co-workers scales included eighteen items each.About half the items chosen for each scale are positive, so that a "Y"rcsI~onscwould indicate satisfaction, and abo ut half are ncgative, so that- 1an "N" esponse would indicate satisfaction. Thus a person who put a"Y" before every item would not get a high (satisfied) score simplybecause of a tendency to say yes.

Reliability an d Validity o f the JDIThe internal consistency reliabilities of the fivc JD I scales rangc from

3 0 to 88 , as dctcrlnined by correctrd split-half correlations based on thercy onxs of eighty innlc cmp1oyr.c~ rom two dilfcrcnt rlc ctmn ic plants.There is no single general criterion mcasllre wllich can be uscd tovalidate a measure of job satisfaction. Wha t is needed is evidence t hat thescales rclate to othcr indc pend rnt n icmin gful indices of satisfaction in thesituation. The approaclr uscd was to study (o) rclations of th e variousJDI scales to other measures of job satisfaction, (b) the influence ofsituational characteristics on thcsc scales, and (c) thc relations betweenthe scales and individual differences thought to be rela ted to job satisfac-tion. These studies have been rcported elsewhere.' Briefly these studiesindicate that the ID 1 yields measures of satisfaction with five differentaspects of jobs which a re discriminably diff erent from e,ach othe r; theaverage correlation between the different scales is approximately .37which is b w enough to indica te a great deal of discrimination among thefive areas. The scales correlate high ly with other measures of satisfaction(average r = .70) and are affc cted in the expected directions by worker,job, and situational differences. In this sense the JDI has validity as ameasure of job satisfaction.

In sum, the JDI, using the direct a priori scoring method, appears tomeet the criteria set out initially. The JDI appears to be valid in thesense that it is representa tive of othe r types of measures of satisfactionand discriiuinates well among the various job areas. It demonstratcsadeq uate internal reliability (althoug h no test-retest studies have been

C. L. Hulin and Patricia C. Smith, "Sex Differences in Jo b Satisfaction," Jorrrrlnlof Applied Psy cl~ dogy 8 (1964): 88-92; idem, "A Linear Model of Job m at is faction:'ibid. 49 (1965): 209-16; L. M. Kendall, "Canonical Analysis of Job Satisfaction andBehavioral, Personal Background , an d Situational Data," (P1l.D. diss., Cornell Uni-versity, 1983); E . A. Locke et a]., "Convergent and Discriminant Validity for Areas andMethods of Rating Jo b Satisfaction," Jotirnal of Applied Psychology 48 (196 4): 313-19.

26-The Effectof Performance o n Job Satisfaction 279done as yet) and is relatively free fruin obvious ~ C S ~ O I I S Ciascs such asacquiescencc . It yields scores on five different areas of job satisfactionand it is short, easily administercd, and easily scored.

There is another interesting characteristic of the JDI th at should benoted. It does not ask the employee to indicate directly how satisfied heis, but rathcr asks him to describc his job ( tha t is, his pay, his wo rk) byputting the appropriate symbol ( Y , N, r ?) in the blank besidc each item(for example, "boring"). It was fclt that such a task would l x easier,particularly for morc poorly cd ncatcd workrrs, tl1a11 dwcribjn g intcrnalfceling statrs. Satisfaction is thus infcrrcd froni these job descriptions;for instance, if an indivi dual describes his work as "boring," "frustrat ing,"and not "pleasant," dissatisfaction with the work is inferred on thoseitems. Actually, as indicated above, scores on the JDI scales agree wellwith more direct mcasures of job satisfaction.

All told, over nine hundred peoplc in seven differentorganizations wereused in the dcvclopmcnt of the [DI. Thus far, thc JD I has bcrn adnlinis-tcred to over two tbousa~ld&~ploy ees in niorr thl n twcnty differentcoinpanics in e nlrrnlwr o f rlilkwnt lypcs of co~nrnr~~liti(~slnd gc,ogr;lpl~-ical locntio~lsn tllv U~~itccltales. Th c ID1 scalcs have s l~o\v n ul)st;ulitjalrelationships to intlivitlr~al, onlpnny. and coninlunity charactcri~tics.~l-though the data analyses are not complete as yet, it appears that the TDIhas adcqua tcly fi~lfillcd he pt ~rpo scs or which it was drsigncd .

26. THE EFFECT OF PERFORMANCEON JOB SATISFACTION "Edward E. Lowler I I I and L g m n W .Porter

THE HUMAN RELATIONS MOVEMENT with its emphasis on good interper-sonal relations, job satisfaction, and the importa nce of informa l groupsprovided an importan t initial stimulant for the study of job attitud es andtheir relationship to human behavior in organizations. Throagh the 30sand ~OS,many studies were carried out to determine the correlates ofhigh and low job satisfaction. Such studies rc lated job satisfac tion toseniority, age, sex, education, occupation, and income, to mention a few.Why this grcat intcrcst in job satisfaction? Undoubtedly some of itstemmed from a simplc desire on the part of scientists to lcarn moreabout job satisfact ion, but much of the intcrcst in job satisfactioo sccms

'C . L. Hulin, "T he Effects of Community C haracteristics on Measures of Job Satis-' fact~on,"Journal of Applied Psychologr 50 (1966): 185-92; idem, "Job Satisfactionk L and Turnover in a Female Clerical Population," ibid., pp. 280-85.From Industrid Relotions 7 (1967): 20-28.

Documents

Development of a Method of Measuring