WIXLIB

INVESTMENT MODELinstructions.If individualsfailed to respondtothe first mailing, a second packet was sent,followed finally by a third set of materials.2Five hundred packets were sent via bulkmail, which constitutedthe most cost-effectivemethod of distribution. Items that are bulkmailed, however, are not forwarded or returned to the sender if they reach an invalidaddress. Therefore, to obtain an accurate return rate estimate, 100 packets were mailedfirst class. Of these 100, 11 were returnedbecause of invalid addresses and 38 were completed and returnedto the investigator.Thus,the estimated overall response rate was 43percent-38 out of 89.3 An additional171 persons responded to the bulk mailing. The totalresponse, then, was 209 persons, 130of whomwere currently involved in serious relationships (79 persons were not currentlyinvolved).These 130persons were the respondentsin thestudy. 58 percentwere female, 95 percentwerecaucasianand 74 percent were married.Theirmean age was around33, they had an averageof 3 years of college educationand their meanpersonal income was approximately 14,000.42 Dillman's (1978) "total design" techniques ofquestionnaireconstruction and follow-up mailingswere utilized in an attempt to maximize responseaccuracy and rate.I Our response rate estimate must be based onlyupon this 100packet samplebecause it is impossibleto determinehow many of the bulk-mailedpacketswere "terminated"due to changed or invalid addresses. Thoughwe knowfromourfirstclass mailingthat approximately11 percent of our packets wereterminateddue to invalidaddresses,we have no wayof determiningwhat percentagewere terminateddueto changedaddresses. Thus, we cannot directly extrapolateresponse rates from the first class mailingto estimate the response rate for the bulk-mailing.4 The representativenessof this sample is not aserious concern because we are testing theoreticalrelationshipsamong variablesratherthan assessingthe incidence of a phenomenonin the generalpopulation (Kidder, 1981).Nevertheless, these figuresdocompare favorably to those obtained for the samegeographicregion in a recent 51 percent responserate mailed questionnairereportedby Rusbult andZembrodt(1982) (53%female, 95%caucasian, 63%married,mean age of 37, and mean educationof 2.5years of college) and to those obtainedin a recent63percent response rate telephone survey conductedby the University of Kentucky Survey ResearchCenter(54%female, medianage between 25 and 40,median education between one and three years ofcollege). Percent marriedwas higher in the presentsurveybecausesurveythanin the Rusbult-Zembrodtthe present survey includedonly persons who werecurrentlyinvolved in serious relationships,whereasthe former survey included all persons who had atsome time been involved in serious relationships.Data regardingincome levels were not comparable,since the present survey assessed personal income83QuestionnaireThe questionnaireincluded measures of allmodel variables-reward value, cost value, investment size, alternativequality, satisfactionand commitment.5Respondentsalso providedinformation regarding demographic characteristics. Unless otherwise indicated, itemswere 5-point Likert scales. First, demographicinformationregardinggender, age (10categories), race, education (7 categories) andpersonal income (10 categories)was obtained.Respondentsalso indicatedwhetherthey weremarried, dating regularly or not involved.Those who were not currentlyinvolved wereasked to complete this generalinformationandreturn the questionnaire. Respondents whowere married or dating someone regularlycompleted the remainderof the questionnaire.In order to obtain comparablemeasures of allmodel variables, we asked that respondentsadopt a similar time perspective-to describetheir relationshipsas they stood just prior totheir most recent period of dissatisfaction.6Inaddition, we felt that respondents would notreadily understandand be capable of answering questions concerning rewards, costs, investments and alternatives, so both specificand generalitems were provided,followingtheprocedure utilized in Rusbult (1980a). In essense, the specific items "taught"respondentsthe meaning of the constructs tapped in thegeneral items.The specific items assessing rewards andcosts concerned, for example, the partner'spersonality, sense of humor and intelligence,and the partners'sex life, way of handlingconflicts and division of household tasks or childcare. Two general items assessed relationshipand the comparisonsurveys assessed householdincome. Thus, the sample obtained in the presentstudy appearsto be demographicallysimilarto thatobtained in comparable research using the samepopulation.S We will not examine the effects of variationsincomparisonlevel on satisfactionor commitmentbecause it is extremelydifficultfor personsto separatethat which exists objectively (reward value, costvalue) from that which they expect more generally(comparisonlevel).6 This time perspectivewas employedbecause:(a)it allowed us to assess respondents'feelings abouttheir relationships during similar relationship"states" (i.e., duringa nontroubledperiod);and (b)this survey was designed to provide data for twoprojects, and such a time perspective was essentialfor the second project. A full 60 percent of the respondents described problems that had occurredwithinthe last month,so we can feel relativelyconfident that findingsreportedbelow do not result fromthe "warm glow" of retrospective reports of thestatus of relationshipsin the distant past.

84SOCIAL PSYCHOLOGY QUARTERLYrewards: "The good traits your partner possesses and the good thingsaboutyour relationship are termed'rewards.'How rewarding[is]your relationship?"and "In general, how [do]the rewardsyou [get] out of your relationshipcompare to those of other people's relationships?" Two general items assessed costs:"The bad traitsyour partnerpossesses and thebad things about your relationshipare termed'costs.' How costly [is] yourrelationship?"and"In general,how [do] the costs you [get]out ofyour relationshipcompare to those of otherpeople's?"The specific items assessing investmentsizeconcerned, for example, shared friends andmaterial possessions, financial security, selfdisclosures, effort expenditureand numberofchildren.The generalmeasureswere: "Generally speaking,how much [have]you invested inyour relationship (e.g., time, energy, selfdisclosures, sharedexperiences, emotionalinvestments)?"and "All things considered,howmuch [have]you 'put into' your relationship?"The specific items assessing alternativequality concerned, for example, the importance of romantic involvement, estimatedtime to begin datingagain if relationshipwereto end, confidenceof findinganotherappealingpartner, enjoyment of time spent alone andlikelihood of finding an appealing alternativepartnerin regardto a variety of specific traits(e.g., personality,physical attractiveness,sexlife). The generalmeasuresof alternativequality were: "Generallyspeaking, how appealing[are]your alternatives(a differentrelationshipor spendingtime without a romanticrelationship)?", "Generally speaking, how [do] youralternatives compare to your relationship?",and "All things considered, how satisfyingwould it [be] to adoptyour alternativesinsteadof your relationship?"Only general measures of satisfaction andcommitment were obtained. The satisfactionmeasures were: "In general, to what extent[are]you attractedto your partner?";"In general, how [does] your relationshipcompare toother people's?"; and "All things considered,how satisfied [are] you with your relationship?". The general commitment measureswere: "For how much longer [do] you wantyour relationshipto last?"; "How committed[are] you to maintainingyour relationship?";"How likely [is] it that your relationship[will]end in the near future?";and "To what extent[do] you feel 'attached'to your partner?"Reliability and Validity of MeasuresReliabilitycoefficients calculatedfor the setof global items designed to measure eachmodel construct revealed sizeable alphas forthe measures of rewards (.85), costs (.63), alternatives (.81), investments (.77), satisfaction(.78) and commitment (.82). Therefore, thegeneral measures of each variable were averaged to form a single estimate of each factor.These averaged measures were employed inthe analyses. To assess the validity of our general measures, we computed the correlationsbetween each general measure and the specificitems associated with that measure. Theseanalyses revealed generally good convergence.The general rewards measure was significantlycorrelated with all specific reward/cost items(median r .29), the costs measure was correlated with 8 of 11 specific reward/cost items(median r -.23), the alternatives generalmeasure was correlated with 13 of 14 specificalternatives items (median r .22), and theinvestments general measure was correlatedwith 8 of 12 specific investments items (medianr .17).RESULTS AND DISCUSSIONPredicting SatisfactionThe investment model asserts that higherreward value and lower cost value should induce greater satisfaction with relationships. Toevaluate these relationships, zero-order correlations and multiple correlations betweensatisfaction and reward and cost value werecalculated for the sample as a whole and forselected demographic subsamples (for age,education, income and duration, a median splitwas used to divide the overall sample into subgroups). A summary of the results of theseanalyses is presented in Table 1.7 The combined impact of reward value and cost value onsatisfaction was significant for the sample as awhole and for all subsamples (refer to Mult. R,df, and F).8 Greater reward value consistentlyencouraged greater satisfaction (refer to REWr). Greater cost value was associated with reduced satisfaction among males, single persons, younger persons, persons with greater7To determinewhetherthe retrospectivenatureofrespondents'descriptionsof theirrelationshipscouldhave affected our findings, we divided our sampleinto two groups-those who describedtheirrelationships as they stood withinone monthand those whodescribedtheir relationships(and problemsin theirrelationships) in the more distant past. We performedzero-ordercorrelationsamongall investmentmodel variables separately for these two subsamples, and found virtuallyidenticalpatternsof resultsfor the two groups.8 The degrees of freedomfor the various analysesdifferslightlyacross analysesdue to missingdataonsome variables.

INVESTMENT MODEL85Table 1. Correlations Between Satisfaction and Reward and Cost ValueaOverall SampleGenderFemalesMalesMarital StatusMarriedSingleAgeUnder 3535 or OlderEducationLess Than Bachelor'sBachelor's or MoreIncomeLess Than 15,000 15,000 or MoreDuration of RelationshipLess Than 10 Years10 Years or MoreaREW reward value and CST*p .05.*p .01. Mult. .622,672,4913.09**15.44**.50**.93**- 762,602,4540.86**30.87**REW rSatisfaction with:CST rcost value.education or income and for persons inlonger-term involvements (refer to CST r).These findingsare consistent with investmentmodel predictions.Costs were not significantlyrelated to reportedsatisfactionfor the overallsample,however,or for the remainingsix demographic subsamples (females, married persons, older persons, persons with lesser education, persons with lower incomes and persons in shorter-termrelationships).These nonsignificanteffects will be discussed below.Predicting CommitmentImpact of satisfaction, alternatives and in-vestments. The model asserts thatcommitmentshouldbe greatto the degreethat satisfactionishigh, alternativesare poor and investmentsizeis great. A summaryof zero-orderand multiplecorrelationsassessing these predictionsfor theoverall sampleand for subsamplesis presentedin Table 2. The combinedimpactof these variables on commitmentwas statistically signifi-Table 2. Correlations Between Commitment and Satisfaction, Alternative Quality and Investment SizeaCommitment with:SAT rALT rMult. RdfFOverall SampleGenderFemaleMaleMarital StatusMarriedSingleAgeUnder 3535 or OlderEducationLess Than Bachelor'sBachelor's or MoreIncomeLess Than 15,000 15,000 or MoreDuration of RelationshipLess Than 10 Years10 Years or .50**-.06.77.563,593,4428.18**6.59**aSAT satisfaction, ALT *p .05.**p .01.INV ralternative quality and INV investment sizc

SOCIAL PSYCHOLOGY QUARTERLY86cant for the overall sample and for all subsamples. Higher levels of commitmentwere consistently inducedby highersatisfactionand bypoorer alternatives. Greater investment sizeconsistently encouragedstrongercommitment(10 of 13 r's were significant), but theinvestment-commitmentrelationship was notsignificantfor single persons, youngerpersonsand persons in longer-termrelationships.(Weare not seriouslytroubledby these few nonsignificant effects of investment size, since theyare generally in the predicted direction).Impact of rewards, costs, alternatives andinvestments. Higher rewards, lower costs,poorer alternatives and greater investmentsshould also promote strongercommitment.Asummaryof correlationaltests of these predictions is presented in Table 3. The combinedeffect of these four predictorswas significantfor all groups. Of course, the correlationsbetween commitmentand alternativequalityandinvestment size remain as described eatercommitment,but this effect was not significant for males and persons with greatereducation. For the overall sample and for several subsamples, costs were not significantlyrelated to reportedcommitment.Examinationof Table 3 pairs of correlations reveals thatcosts exerted a negative impact on commitment for some subgroupswhile exertinga positive impactfor theircounterparts;highercostsencouraged lower commitment for youngerpersons, persons with greatereducation, persons with higher incomes and persons inlonger-term involvements, whereas highercosts encouraged greater commitment forolder persons, persons with less educationandpersons with lower income.The Impact of CostsThe nonsignificant and/or inconsistent effects of costs on satisfactionand commitmentbearfurtherexamination.Such findingsare notunique to this investigation:Several empiricalinvestigations have reported nonsignificanteffects of costs or related variables(e.g., conflict) on feelings for partners(e.g., Argyle andFurnham, 1983; White, 1983), and severaltheoristshave also suggestedthat such findingsshould not be surprising (e.g., Braiker andKelley, 1979; Scanzoni, 1979). Also, Rusbult(1980a)found that while variationsin costs affected satisfaction, the effects of costs oncommitmentwere weak. And finally, Rusbult(1983) found that variations in costs affectedneither satisfaction nor commitment. Why?We explore four possible explanations.First, adherence to the romantic ideal-thebelief that one accepts a partnerfor richer orpoorer, in sickness and health (given rewardsand costs)-may prevent individualsfrom becoming less satisfied with and committed totheir relationshipsas the costs of doing so increase. If this were so, we should observe anonsignificantor positive effect of cost valueon satisfaction and commitment among subgroups who are known to believe morestrongly in the romantic ideal-males andTable 3. Correlations Between Commitment and Reward Value, Cost Value, Alternative Quality and Investment SizeaCommitment with:REW rCST rALT rMult. 36**.21*.33**- .29**-.35**- *4.64*Overall SampleGenderFemaleMaleMarital StatusMarriedSingleAgeUnder 3535 or OlderEducationLess Than Bachelor'sBachelor's or MoreIncomeLess Than 15,000 15,000 or MoreDuration of RelationshipLess Than 10 Years10 Years or MoreaREW reward value, CST*p .05.**p .01. cost value, ALT INV ralternative quality and INV investment size.

INVESTMENT MODELyoungerpersons (Hatkoffand Lasswell, 1979;Rubin, 1973).Both of these relationshipswerenegative, however, in the present study.Second, it is possible that the reward-costdifferential predicts satisfaction and commitmentmoreeffectivelythando rewardsandcostsindividually;that is, thatit is not absolutelevelof costs, but level of costs relative to level ofrewards that predicts satisfaction and commitment. However, we calculated a rewardcost differential score for each respondent,performedfurtheranalyses, and found that:(a)rewardvalue alone predictedsatisfactionbetter than did the reward-costdifferentialfor theoverall sample(the respectiveR's were .67 and.24) and for most subsamples (9 of 12), andreward value alone predicted commitmentbetter than did the reward-costdifferentialforthe overall sample (R's were .29 and .17) andfor most subsamples(7 of 12);and (b) the prediction of satisfactionfrom reward value andcost value was superior to that from thereward-costdifferentialfor the overall sample(R's were .55 and .24) and for most subsamples(9 of 12), and the prediction of commitmentfrom reward value and cost value (plus alternatives and investments)was superiorto thatfrom the reward-costdifferential(plus alternatives and investments) for the overall sample(the R's were .62 and .61) and for most subsamples (10 of 12).Third,it is possible that there is a thresholdeffect for cost value; perhaps variations incosts do not affect degree of satisfactionandcommitmentwhen costs are low, but come toexert negative effects at higher levels. We divided our sampleinto low and highcost groups(median split) and calculated correlationsbetween cost value and satisfaction and commitment for both groups. Indeed, we foundthat for the low cost group, variationsin costvalue did not significantlyaffect satisfaction(r -.13) or commitment (r -.03),87tion and commitmentin high cost-low rewardinvolvements (r's -.45 and -.36), these re-lations were not significant in high cost/highreward involvements (r's -.12 and -.27),low cost/low reward involvements (r's -.35and -.13), or low cost/high reward involvements (r's - .01 and - .07).Thus, two of our four potentialexplanationsfor the aberranteffects,of cost value on satisfaction and commitment seem promising:There appearsto be a thresholdeffect for theimpact of costs on satisfaction and commitment, and reward and cost value appear tointeract in affecting satisfaction and commitment, high costs producingreduced satisfaction and commitmentespecially given the preconditionof low rewardvalue. This phenomenon, and particularlyth

Review of Personality and Social Psychology, edited by Ladd Wheeler. Beverly Hills, CA: Sage. Tetlock, Philip E., Kristen A. Hannum and Patrick M. Micheletti. 1984. "Stability and change in the complexity of senatorial debate: Testing the cog- nitive versus rhetorical style hypothesis." Journal