Kent Academic Repository Bardi, Anat, Calogero, Rachel M. and Mullen, Brian (2008) A new archival approach to the study of values and value-behavior relations: Validation of the value lexicon. Journal of Applied Psychology, 93 (3). pp. 483-497. ISSN 0021-9010. Downloaded from https://kar.kent.ac.uk/15454/ The University of Kent's Academic Repository KAR The version of record is available from https://doi.org/10.1037/0021-9010.93.3.483 This document version Publisher pdf DOI for this version Licence for this version UNSPECIFIED Additional information Versions of research works Versions of Record If this version is the version of record, it is the same as the published version available on the publisher's web site. Cite as the published version. Author Accepted Manuscripts If this document is identified as the Author Accepted Manuscript it is the version after peer review but before type setting, copy editing or publisher branding. Cite as Surname, Initial. (Year) 'Title of article'. To be published in Title of Journal , Volume and issue numbers peer-reviewed accepted version. Available at: DOI or URL (Accessed: date). Enquiries If you have questions about this document contact [email protected]. Please include the URL of the record in KAR. If you believe that your, or a third party's rights have been compromised through this document please see our Take Down policy (available from https://www.kent.ac.uk/guides/kar-the-kent-academic-repository#policies). JournalofAppliedPsychology Copyright2008bytheAmericanPsychologicalAssociation 2008,Vol.93,No.3,483–497 0021-9010/08/$12.00 DOI:10.1037/0021-9010.93.3.483 A New Archival Approach to the Study of Values and Value–Behavior Relations: Validation of the Value Lexicon Anat Bardi, Rachel M. Calogero, and Brian Mullen UniversityofKent Thepresenteffortemploysanewarchivalapproachtostudyvaluesandvalue–behaviorrelations,which islikelytobeparticularlyusefulinappliedsettings.Avaluelexiconwasdevelopedonthebasisofthe Schwartz (1992) value theory to extract lexical indicators of values from texts. The convergent, discriminant,andpredictivevalidityofthismeasurewasestablishedusingAmericannewspapercontent from1900to2000vis-a`-visexistingself-reportmeasuresofvaluesandobjectiveindicatorsofvalue- expressivebehaviors.Resultsprovideempiricalsupportfortheuseofthevaluelexicontostudyvalues and value–behavior relations. First, the value lexicon demonstrated convergence with self-report re- sponsesofvalues.Second,valuesinAmericannewspaperswereassociatedwithobjectiveindicatorsof their corresponding value-expressive behaviors compared with noncorresponding value–expressive behaviors. Third, patterns of values over this 101-year period exhibited meaningful fluctuations with majorhistoricalandpoliticalevents.Thediscussiondescribesnewpossibilitiesforfutureresearchon valuesinmanyappliedsettingswiththevaluelexicon.Thediscussionalsosuggeststhattheprinciples ofthevaluelexiconcouldbeadoptedtomeasureotherpsychologicalconstructsofinteresttoapplied psychology. Keywords:values,behavior,archivalresearch Rokeach(1973)proposedthatvaluescanbeusedasaunifying seems important to be able to measure values and behavior in a concept across all the sciences concerned with social behavior. way that would facilitate examination of these relations across Values (e.g., achievement, conformity) are defined as relatively appliedsettingsandacrosstimeandplace. stable,broadgoalsthatguidepeople’sperceptions,attitudes,and Much of the research on values and behaviors has relied on behaviorsacrosstimeandcontexts(e.g.,Allport,1961;Rokeach, self-reportresponses.Whereasaself-reportparadigmusuallycan 1973;Schwartz,1992).Oneofthemostcomprehensiveandrobust bejustifiedinstudyingvalues(e.g.,Borkenau&Ostendorf,1987; theoriesofvaluesistheSchwartz(1992)valuetheory.Thistheory Rohan, 2000), it has limitations in certain circumstances (see defines10broad,cross-culturallyrecognizedvaluesbasedonthe Podsakoff,MacKenzie,Lee,&Podsakoff,2003,forreview).First, particular motivational goal that underlies each value (see Table self-report questionnaires are prone to response biases, such as 1). The Schwartz value theory has established predictive validity social desirability and consistency biases (see Podsakoff et al., across samples, outcome variables, languages, and cultures (see 2003,foranextensivereview).Second,theyareanobtrusiveand review in Schwartz & Bardi, 2001). Furthermore, this theory has labor-intensive method for measuring values across time and been shown to significantly predict behavior across a variety of place. Finally, it is impossible to administer self-report question- applied settings. For example, values predicted making sugges- naires “to measure the subjective states of people long dead” tionsforimprovementatwork(Lipponen,Bardi,&Haapama¨ki,in (McGuire,1976,p.175),althoughthistypeofdatacouldinform press),reactionstoproceduraljusticeallocationatwork(Fischer& knowledgeaboutvalue–behaviorrelationshistoricallyandacross Smith, 2004), buying environmentally friendly products (Grunert timeandreal-worldsettings.Thus,developingameasureofindi- &Juhl,1995),politicalvoting(Caprara,Schwartz,Capanna,Vec- vidualvaluesthatdoesnotrelyontheadministrationofself-report chione, & Barbaranelli, 2006), and counselees’ behavior during questionnaires is crucial for contexts in which it is undesirable, careercounseling(Sagiv&Schwartz,2004).Clearly,onthebasis impractical,orimpossibletoadministeraquestionnaire. of this prior research, an understanding of values and value– Thecurrentarticlepresentsanewarchivalapproachthatcould behavior relations in applied settings is important. It therefore be particularly useful in such contexts. First, the rationale for developing a value lexicon is described, and the primary aims of this research are summarized. Next, the method used to validate Anat Bardi, Rachel Calogero, and Brian Mullen, Department of Psy- the value lexicon is delineated. Specifically, this section explains chology,UniversityofKent,Canterbury,UnitedKingdom. thedevelopmentofthevaluelexiconusingAmericanEnglishand WeacknowledgewithgreatsadnesstheuntimelydeathofBrianMullen the strategy used to establish convergent, discriminant, and pre- during work on this article. The article reflects his legacy regarding the dictivevalidityofthevaluelexicon.Next,weestablishthelexical importanceofarchivalresearch. co-occurrenceamongthesameanddifferentvaluewordscompris- WethankDominicAbramsandDirkJanssenfortheirinvaluablecom- ing the value lexicon. We establish the convergent and discrimi- mentsonpreviousdraftsofthisarticle. Correspondence concerning this article should be addressed to Anat nantvalidityofthevaluelexicontomeasurepatternsofindividual Bardi, Department of Psychology, University of Kent, Canterbury, Kent valuesbyexaminingitsconvergencewiththestructureofindivid- CT27NP,UnitedKingdom.E-mail:[email protected] ualvaluesmeasuredbytraditionalself-reportmethods.Inaddition, 483 484 BARDI,CALOGERO,ANDMULLEN Table1 Definitionsofthe10ValuesBasedonMotivationalTypesandItemsThatMeasureThem Value Definition Power Socialstatusandprestige,controlordominanceoverpeopleandresources(socialpower,authority,wealth). Achievement Personalsuccessthroughdemonstratingcompetenceaccordingtosocialstandards(successful,capable,ambitious,influential). Hedonism Pleasureandsensuousgratificationforoneself(pleasure,enjoyinglife,self-indulgent). Stimulation Excitement,novelty,andchallengeinlife(daring,avariedlife,anexcitinglife). Self-direction Independentthoughtandaction-choosing,creating,exploring(creativity,freedom,independent,curious,choosingowngoals). Universalism Understanding,appreciation,toleranceandprotectionofthewelfareofallpeopleandofnature(broadminded,wisdom,social justice,equality,aworldatpeace,aworldofbeauty,unitywithnature,protectingtheenvironment) Benevolence Preservationandenhancementofthewelfareofpeoplewithwhomoneisinfrequentpersonalcontact(helpful,honest, forgiving,loyal,responsible). Tradition Respect,commitmentandacceptanceofthecustomsandideasthattraditionalcultureorreligionprovidetheself.(humble, acceptingmyportioninlife,devout,respectfortradition,moderate) Conformity Restraintofactions,inclinations,andimpulseslikelytoupsetorharmothersandviolatesocialexpectationsornorms (politeness,obedient,self-discipline,honoringparentsandelders). Security Safety,harmonyandstabilityofsociety,ofrelationships,andofself(familysecurity,nationalsecurity,socialorder,clean, reciprocationoffavors). weexaminethepredictivevalidityofthevaluelexiconbycorre- the values and their structure are slightly different from the ones lating patterns of individual values, as portrayed in American used in the individual-level value theory. Hence, using the (U.S.)newspapers,withobjectiveindicatorsofbehaviorthrough- cultural-levelvalueswouldrequiredevelopingavaluelexiconthat out the 20th century. We note that the reliance on newspapers in reflectsthecontentsoftheculturalvaluedimensions. the United States alone is necessary given that the value lexicon was developed using American English. The utility of the value Rationale for Developing a Value Lexicon lexicon is illustrated by demonstrating the correspondence be- tween patterns of individual values as portrayed in American Theuseofarchivaldatasourcesinempiricalresearchhasbeen newspapers and major national events. Finally, we discuss the described as detecting a “cognitive footprint” (e.g., Waller & potentialapplicationsandlimitationsofthisnewarchivalapproach Zimbelman, 2003) or obtaining a “cardiogram of an age” (e.g., forthestudyofpatternsofindividualvaluesandvalue–behavior Ricci,2003).Oneofthedefiningattributesofanarchivalapproach relationsinsocial,organizational,andculturalcontexts. isitscapacitytoexaminepsychologicalconstructsintimesandin We rely on Schwartz’s (1992) individual-level value theory to places where self-report questionnaire data are not available. In- develop the value lexicon. Although the individual-level value deed,numerousexamplesexistdocumentingtheutilityofarchival theoryisusuallyusedtoexamineindividualdifferencesinvalues, data sources to study psychological processes and outcomes. For prior research has also used aggregated responses of individual example, archival research studies have established significant values to represent national patterns of individual-level values associationsbetweenthepersonalityofleadersandorganizational (Schwartz&Bardi,2001;Schwartz&Sagie,2000).Theseaggre- gatesofindividualindicatorscaninformknowledgeaboutwhatis performance(Peterson,Smith,Martorana,&Owens,2003),archi- importanttoindividualsatthestudiedtimeandplaceandhowthis val indicators of societal threat and authoritarian attitudes and corresponds with behavior en masse. Establishing the validity of behaviors (e.g., Doty, Peterson, & Winter, 1991; Peterson & the value lexicon in this way would advocate its use both for Gerstein, 2005; Sales, 1973), and prediction of various forms of measuring values of individuals in organizations (including lead- outgroup hostility (Mullen, 2001, 2004; Mullen, Calogero, & ers) and for measuring individual-level values that organizations Leader,2007). advocate (e.g., achievement, conformity, self-direction), as was Previous research has also utilized archival indicators to study donebyBardiandSchwartz(1996)inidentifyingindividual-level values,corroboratingthemwithself-reportsofvalues.Forexam- valuesincommunistideologicalwritings. ple, Eckhardt and Alcock (1970) identified the values that were Note, however, that when the research question is focused on central to communism by content analyzing communist texts. the structure of society, a group, or an organization, it would be Yearslater,BardiandSchwartz(1996)foundevidencethatthese more appropriate to employ the Schwartz (2004) cultural value valueswereunderstoodascommunistvaluesusingself-reportsin theory.ThisisbecauseaccordingtoSmithandSchwartz(1997),a teachersamplesinEasternEurope.Moreover,thepersonalimpor- culture-leveltheoryisusefulforstudyingtheeffectofcultureon tance of these values predicted voting for communist parties. In thestructureofsocietyorothergroupcharacteristics(e.g.,demo- another study, Schwartz and Ros (1995) identified differences in graphics, political characteristics). The cultural value theory thevaluesembeddedintheAmericanDeclarationofIndependence (Schwartz, 2004) is the appropriate theory for such purposes compared with values embedded in the motto of the French rev- becauseitviewsculturalvaluesasdifferentsolutionstoproblems olution. They found the same differences in values comparing andchallengesthatfaceallsocieties(e.g.,howtoensurerespon- American and West European teacher samples using self-reports siblebehaviorofmembersofsociety,determiningtheboundaries of values. Notably, these studies suggest that individual values betweentheindividualandthecollective,etc.).Hence,acultural embedded in texts can serve as good indicators of values in the valuetheoryistheoreticallyfocusedonvaluesthathavedeveloped relevant population. The present effort extends previous archival insocietiesinresponsetosuchsocietalchallenges.Thecontentsof workonvaluesbydevelopinganewapproachtomeasurepatterns ANARCHIVALAPPROACH 485 ofindividualvaluessystematicallyonthebasisofnaturallanguage behaviors,thenwewouldexpectthesetofvaluewordsrepresent- use. ing each value to be positively associated with the behavioral Natural language use provides a reliable and valid indicator of indicator expressing that value in comparison with behavioral basic personality, cognitive processes, and social processes (Pen- indicatorsthatdonotexpressthatvalue. nebaker&King,1999;Pennebaker,Mehl,&Niederhoffer,2003). Consequently,therehavebeenpreviouseffortstodevelopsystem- Method aticmethodstoanalyzethecontentofnaturaltexts(e.g.,Kulesa& Bishop,2006;Pennebaker,Francis,&Booth,2001;Winter,1994). The Value Lexicon Recent innovations in the study of associative meaning among wordsinnaturallanguageusehaveutilizedthebillionsofnatural The process of constructing the value lexicon began with an languagedocumentsavailableontheInternet(Gulli&Signorini, initial pool of all possible words that reflected each of the 10 2005; Heylighen, 2001; Turney, 2001). Specifically, the associa- values.ThesewordsweredrawnfromtheitemsontheSchwartz tivelinkagebetweenanyarrayofwordscanbeempiricallydefined (1992) value questionnaire and from the 6th edition of Roget’s in terms of their lexical co-occurrence.1 The core idea of using InternationalThesaurus(Kipfer,2002)torepresenteachofthe10 lexicalco-occurrencetogaugeassociativestrengthisthat“aword individualvalues.Theaimwastoidentifythreewordsthat,when ischaracterizedbythecompanythatitkeeps”(Firth,1957,p.11). appearingtogether,wouldrepresenttherelevantvalue.Consistent For any given construct, a lexicon of words indicative of the with the use of multiple items to measure one value (Schwartz, construct can be developed. The relative co-occurrence of these 1992; Schwartz & Bardi, 2001), we reasoned that the occurrence wordsisusedtoindicatethedegreetowhichthesewordsconverge of three words on a single Web page to represent a given value ontheactivationoftheconstruct. wouldserveasamorereliableindicatoroftheprevalenceofthat This Internet-based approach for measuring lexical co- value than the occurrence of just one word to represent a given occurrenceisanimprovementoverextantsetsofwordassociation value.Wechosetousethreewords(ratherthananyothernumber) norms,whicharelimitedintheirrepresentationoftargetwordsfor asthisisthesmallestnumberofitemsthatmeasureavalueinthe every construct of interest and rely on restricted samples of par- most recent version of the Schwartz Value Survey (Schwartz, ticipants to generate the target words (e.g., Moss & Older, 1996; Sagiv & Boehnke, 2000). The set of three words for each value Palermo&Jenkins,1964).Instead,thisnewapproachreliesonthe were selected in order to maximize the co-occurrence of pairs of massive lexical corpus on the Internet to determine which words words for the same value while simultaneously minimizing the actually co-occur in use. Spence and Owens (1990) have shown co-occurrence of pairs of words for different values, as detailed that indicators of lexical co-occurrence in natural language use below. renderresultsthatarehighlyconsonantwiththepatternsobtained To narrow down the initial pool of words, we applied several from word association norms. Similar applications of this basic criteria.First,apreferencewasgiventousethevaluelabelitself approach have been reported by others (e.g., Fellbaum, 1998; (e.g.,power,security)asoneofthewordstorepresenteachvalue; Heylighen, 2001). In the original validation of this lexical co- however,thiswasnotalwayspossiblebecausesomeofthesevalue occurrence approach to measuring psychological constructs, Cal- labelsyieldedprohibitivelylowwordfrequencies(e.g.,universal- ogero(2007)usedthisapproachtodevelopalexiconofwordsto ism, hedonism). Second, in order to hold part of speech constant measureindividualdifferencesinthemotivationtoattainoravoid across different values, we considered only nouns. For example, cognitiveclosure.Goodtoexcellentconvergent,discriminant,and forthevalueHedonism,thenounpleasurewasusedratherthanthe predictivevalidityweredemonstratedforthisnewlexicalmeasure adjectivepleasing.Third,inordertominimizelexicalambiguity, of need for cognitive closure across eight studies. The present obviouspolysems(wordshavingmultiplemeaningsdependingon effortdrewonthislexicalco-occurrenceapproachtodevelopand context) were avoided. For example, the noun practice might be validate a new measure of individual values to study patterns of appropriateforthevalueTraditionindenotingtherelevantmean- valuesandvalue–behaviorrelationsovertime. ing of a custom or a habit, but practice could also denote the irrelevant meaning of a professional office. Finally, the words Summary of the Aims of the Current Research selected to represent each of the 10 values had to convey the In this research, the Internet-based approach described above 1Although beyond the scope of this article, there is an obvious reso- wasutilizedtodevelopalexiconofvaluewordstorepresentthe10 nance between the present approach and latent semantic analysis (LSA; distinct values comprising Schwartz’s (1992) individual-level e.g.,Deerwester,Dumais,Furnas,Landauer,&Harshman,1990;Landauer valuetheory.Thisvaluelexiconwassubjectedtotestsofconver- &Dumais,1997;Salton&McGill,1983).Theprimarydifferencebetween gent, discriminant, and predictive validity. First, if the value lex- thepresentapproachandLSAisthedirectionofconstructionofasemantic icon represents Schwartz’s 10 values, then the value words com- space:WhereasLSAattemptstodiscoveralowerdimensionalrepresen- prising the lexicon would be expected to differentially represent tationofasemanticspacefromhigherdimensionalvectorsofcountsof eachofSchwartz’s10values.Specifically,thesetofvaluewords co-occurrence,thepresentapproachattemptstovalidateahigherdimen- sionalrepresentationofasemanticspacefromlowerdimensionalvectors representingeachofthe10valuesshoulddiscriminateamongthe ofcountsofco-occurrence.Inotherwords,LSAmighthaveattemptedto 10valuesintheirrelativelexicalco-occurrenceinnaturallanguage distilltheconstellationof10valuesfromann-dimensionaltabulationof use on the Internet. Second, if the value lexicon represents the co-occurrences of all value-relevant terms, whereas the present ap- Schwartz’s 10 values, then the structure of relations among the proachattemptstoconstructatabulationoftheco-occurrencesofvalue- values as represented by value words would be expected to be relevant terms that maps onto the constellation of 10 values already significantly positively correlated with the structure of values establishedfromself-reportquestionnairestudies.SeeTurney(2001)for measuredbyself-report.Third,ifthevaluelexiconpredictsactual reviewofdifferencesbetweentheseapproaches. 486 BARDI,CALOGERO,ANDMULLEN meaning of the relevant value when they appeared together. For JP(kindness(cid:2)strength) example, the word respect does not necessarily reflect Tradition whenitappearsalone,butitdoesreflectTraditionwhenitispaired (cid:1)(2,360)/[(12,200(cid:3)130,000)/2](cid:1).03. with the words tradition and custom. The resultant lexicon of 30 If the value lexicon represents Schwartz’s 10 values, then we valuewordsispresentedinTable2. expectedthevaluewordsinthelexicontodifferentiallyrepresent Google represents the largest and most popular search engine each of Schwartz’s 10 values. Specifically, the mean JP for the availableontheInternet,andthesearchesoftermsreflectpopular co-occurrence of pairs of words for the same value should be usageofthoseterms(Blair,Urland,&Ma,2002).Googlesearches higherthanthemeanJPfortheco-occurrenceofpairsofwordsfor all documents, blogs, messages, and Websites to which it has differentvalues,thusdiscriminatingbetweenthe10valuesintheir access (e.g., HTML, PDF, Word, Macromedia Flash [SWF], relative lexical co-occurrence in natural language use. Conver- spreadsheets, plain text) and computes the number of sites that gence among the value words representing the same value and containtheenteredsearchterm(s).Therefore,searcheswerecon- divergence among the value words representing different values ductedforthese30wordswiththeGooglesearchengine,restricted would provide a measure of validity for the 10 distinct values toWebsitesinEnglishoriginatingfromtheUnitedStates.Each1 representedinthelexicon. of the 30 words was searched singly to obtain a frequency esti- It is important to note that the set of three words selected to mate,referredtoasaGooglehit.Replicatingpreviousresults(see measureeachvaluearenotnecessarilysynonymsbutratherwere Blairetal.,2002),thenumberofsingleGooglehitsforeachword selected to give the best coverage of the value concept. For exhibited significant correlations with the word frequency esti- example, the words independence and freedom that measure the mates from Thorndike and Lorge’s (1944) 60–year-old norms, r(28) (cid:1) .41, p (cid:1) .01, and Kucera and Francis’s (1967) 40–year- valueSelf-Directionarenotsynonyms.Theyrepresenttwodiffer- oldnorms,r(28)(cid:1).57,p(cid:1).0005,indicatingthatthenumberof ent aspects of the Self-Direction value in the Schwartz Value Survey (Schwartz, 1992; see Table 1 above). Thus, their lexical Google hits for each word is an adequate indicator of word co-occurrencereflectsthebroadvalueconstructratherthanmerely frequency. Second, for each of the 10 values, the lexical co- expressing the same concept with multiple but equivalent terms. occurrence of word pairs for same and different values was cal- Thelexicalindicatorsselectedforeachvaluerepresentthebroad culated in terms of their joint probability. For example, kindness meaningofthevaluewhentheyappeartogether,inthesameway andmercyarewordsthatrepresentthesamevalue(Benevolence), thattheself-reportitemsforeachvaluerepresentthebroadmean- whereas kindness and strength are words that represent different ing of the value when averaged together. Because the words that values(BenevolenceandPower).Searchesweredonesingly(e.g., represent the same value are not synonyms, the expectation that “kindness”) and for every possible pair (e.g., “kindness mercy,” their lexical co-occurrence will be higher than their lexical co- “kindnessstrength”).Jointprobability(JP)isdefinedastheratio occurrencewithwordsthatmeasureothervaluesisnotanobvious ofthenumberofpairedhits(PH)totheaverageofthenumberof one. Furthermore, the values represented in the Schwartz (1992) singlehits(SH;Googlehitsexpressedin1,000s).Toillustrate,the value theory are interrelated with one another, creating a quasi- calculationoftheJPfortheco-occurrenceofkindnessandmercy circumplex (a circle without specified gradients, ordered in the isrepresentedas: same way as in the tables of this article) such that each value is JP(kindness(cid:2)mercy) more positively correlated with adjacent values in the circle and eachvalueislesspositivelycorrelatedwithnonadjacentvaluesin (cid:1)(kindness(cid:2)mercyPH)/[(kindnessSH(cid:3)mercySH)/2] the circle. Therefore, the differences between the lexical co- occurrenceofwordsforthesameanddifferentvalueswouldhave TheJPfortheco-occurrenceofkindnessandmercy(benevolence) to be quite robust in order to overcome the additional challenge is: presentedbythecircumplexstructureofthevalues. JP(kindness(cid:2)mercy) (cid:1)(1,560)/[(12,200(cid:3)21,800)/2](cid:1).09. Convergent and Discriminant Validity of the Value Lexicon The JP for the co-occurrence of kindness and strength (Benevo- lenceandPower)is: In order to test the convergent and discriminant validity of the valuelexicon,itwasfirstnecessarytoidentifyasourceofpeople’s Table2 naturallanguageusetowhichwecouldapplythevaluelexiconto WordsSelectedtoRepresentEachValue:TheValueLexicon measure the 10 values. Newspapers provide a rich, extensive sourceoflexicaloutputtowhichwecanapplythevaluelexicon. Value Lexicalindicatorsforeachvalue Althoughcautioniswarrantedwhenutilizingmediadatabecause Power power,strength,control of imperfect representativeness (Ortiz, Myers, Walls, & Diaz, Achievement achievement,ambition,success 2005),thereisevidenceinpreviousstudies,whichusedmethods Hedonism luxury,pleasure,delight different from our method, that popular textual media in demo- Stimulation excitement,novelty,thrill cratic societies largely represent the salient values, opinions, and Self-direction independence,freedom,liberty concerns of people. Specifically, in a comparison of newspapers Universalism unity,justice,equality Benevolence kindness,charity,mercy that either published all letters to the editor or published only Tradition tradition,custom,respect selected letters to the editor, individual letter writers to these Conformity restraint,regard,consideration newspapers did not differ in their support of the Equal Rights Security security,safety,protection Amendment (ERA; Hill, 1981). Furthermore, the percentage of ANARCHIVALAPPROACH 487 supportfortheERAindicatedinpublishedletterstotheeditordid Specifically, in contrast to the value questionnaire (Schwartz, notsignificantlydifferfromthesupportindicatedinpublicopinion 1992),thevaluelexiconrepresentsanindirectindicatorofvalues, polls that represent the general population. In addition, Inglis because values are reflected from the co-occurrence of value (1938)foundthatchangesintheportrayalsofagenticheroinesin words in texts, whereas in value questionnaires values are mea- TheSaturdayEveningPostfrom1901to1935covariedwith,but sured directly. However, we did expect a positive correlation didnotprecede,changesingainfulemploymentofwomeninthe betweenthetwopatternsofrelations. United States. More recently, Adelman and Verbrugge (2000) observedthatthecoverageofsixprominentdiseasesin34Amer- Predictive Validity of the Value Lexicon icannewspapersreflectedsocietaltrendsindiseaseprevalenceand mortality. These findings indicated that increases in newspaper The term value-expressive behavior refers to behaviors that coverageofcancerorAIDSreflect,ratherthancause,increasesin express or promote the attainment of primarily one value as theprevalenceofthesediseases.2 comparedwithothervalues(Bardi&Schwartz,2003).Forexam- Therefore,weexaminedtheoccurrenceofindividual-levelval- ple, dominating behaviors primarily express Power values, and uesinpopularAmericannewspapers.Tobegin,wemeasuredthe pleasure- seeking behaviors primarily express Hedonism values. prevalence of the value words on newspaper pages. A series of Only one previous study examined relations between Schwartz’s searcheswasconductedinU.S.newspapersfortheyears1900to 10 values and an array of behaviors that express the 10 values 2000onNewspaperArchive.com,anInternet-basedsearchengine (Bardi & Schwartz, 2003), demonstrating moderate relations be- withaccesstomillionsofnewspaperpages,fromeightcountries, tweenvaluesandbehaviorswhenbehaviorwasmeasuredbyself- for the years 1759 to present. First, for each year, the number of and peer-reports. For the purposes of this research, archival indi- U.S.newspaperpagescontainingallthreeofthewordsforeachof cators that were not derived from responses to self-report ques- the10valueswastabulated.Itwasreasonedthattheoccurrenceon tionnaires were specifically utilized because they circumvent the asinglenewspaperpageofallthreewords(e.g.,luxury,pleasure, knownlimitationsoftraditionalself-reportmeasures,spanhistor- and delight) for a given value (e.g., Hedonism) would serve as a ical periods and contexts, and provide greater ecological validity more reliable indicator of the prevalence of that value than the for the constructs under investigation. Whereas the indicators occurrenceofjustoneofthosewords.Second,foreachyear,the utilizedherearepotentiallyfallible,theseindicatorsarenobetter numberofAmericannewspaperpagescontainingthewordthewas or worse than more traditional data extraction methodologies, as tabulated in order to provide a baseline of the total number of documented in the extant archival research literature, (e.g., Sim- newspaper pages printed during that year. This was based on the onton, 2003; Webb, Campbell, Schwartz, Sechrest, & Grove, assumption that every page of a newspaper in English would 1981). The purpose was to correlate these behavioral indicators containthewordthe,asweindeedfound.Third,foreachyear,the with value occurrences in newspapers across 101 years (1900– relative prevalence of each of the 10 values was defined as the 2000). number of American newspaper pages containing all three of the Each of the behavioral indicators was selected on the basis of wordsforagivenvaluedividedbythenumberofnewspaperpages several criteria: First, the behavioral indicator needed to be con- in that year. On the basis of these prevalence figures, it was sistent with the conceptual definition of the value based on the possible to compute the intercorrelations among the 10 values in Schwartz (1992) value theory. Second, the behavioral indicator Americannewspaperspagesacross101years. hadtobeconsideredanexpressionofprimarilyonevalue,andnot To test the convergent and discriminant validity of the value an expression of the other values. Third, the behavioral indicator lexicon, we compared the lexical occurrence of the 10 values in hadtobeconsideredanexpressionofavaluethatcanvaryinits popularAmericannewspaperswithself-reportedresponsesofthe frequency of expression over time while retaining functionally 10values.Specifically,weexaminedtheconvergencebetweenthe equivalent meanings across time and social change (Doty, Peter- intercorrelations among the 10 values in newspapers and the son, & Winter, 1991). Fourth, the behavioral indicator needed to intercorrelations among the 10 values based on self-report re- bequantifiableinameaningfulwaywithoutdistortingthenature sponsestotheSchwartzValueSurvey(Schwartz,1992).Bardiand ofthebehaviorexpressed.Finally,thebehavioralindicatorhadto John (2007) computed intercorrelations among the 10 different providedataformostoralloftheyearsfrom1900to2000.The values on the basis of the self-report responses of 545 American objective indicators selected to represent the 10 value-expressive students. The 10 (cid:4) 10 matrix produced by computing the inter- behaviorsaredescribedbelow. correlations among the 10 different values yields a total of 55 Power. ThevalueofPowerstemsfromthemotivationtoexert correlations. If the value lexicon serves as an indicator of dominanceoverpeopleandresources(Schwartz,1992).Whenpeople Schwartz’s(1992)10values,thenthepatternofintercorrelations value Power, they are likely to engage in activities that will enable among the 10 values reflected in American newspaper pages themtohavedominanceoverothers.Takingpartinmilitaryactionis shouldconvergewiththepatternofintercorrelationsamongthe10 abehaviorthatenablespeopletopursuetheirPowervaluesbecause valuesinparticipants’self-reports.Thatis,thegeneralpatternsof theaimofmilitaryactionistoestablishdominanceoverpeopleand/or relations among the values observed in the 55 correlations ob- resources. Indeed, military cadets valued power much more highly tained from the lexical co-occurrence of the 10 values in U.S. newspaper pages should exhibit the same general patterns as 2Ofcourse,value-ladenmessagescanalsoaffectvalues,aswasshown observedinthe55correlationsobtainedfromU.S.students’self- regarding the influence of school textbooks on children (McClelland, reportsofthe10values(Bardi&John,2007).Wedidnotexpect 1961).However,childrenaremoreamenableforvaluechange(see,e.g., to find exactly the same structure of relations among the values Bardi & Schwartz, 1996). Although the media is likely to have some whencomparingthetwopatternsofrelations,astheyrepresented influenceonitsaudience,inademocraticsocietyitismoreplausiblethat markedly different operationalizations of the Schwartz values. themediareflectswhatisimportanttoitsaudience. 488 BARDI,CALOGERO,ANDMULLEN than a comparable sample of adults (Lo¨nnqvist, Verkasalo, Bez- ulationvaluesbecausemoviesprovideopportunitiesforexcite- menova,&Helkama,2007).Wethereforereasonedthatthepercent- mentandarousal.Whenthereisanincreaseddemandtowatch ageofthepopulationonactivemilitarydutywouldbeoneobjective movies, it is likely that the film industry will respond by indicatorofbehaviorthatexpressesthevalueofPower.Thatis,the releasingmorefilms.Wethereforereasonedthatthenumberof greatertheextenttowhichPowervaluesaresalientandimportant,the entertainment films released per capita would be one objective morepeopleinthepopulationwillenlistinthearmy.Dataforthis indicator of behavior that expresses the value of Stimulation. objectiveindicatorofbehaviorwereavailablefortheyears1900to Dataforthisindicatorwereavailablefortheyears1900to2000 2000(Kurian,2004).Weexpectedthatincreasesintheprevalenceof (Schmidt, 2006). We expected that increases in the prevalence PowervaluesinAmericannewspapersduringthisperiodwouldbe of stimulation values in U.S. newspapers during this period associatedwithincreasesinthepercentageofthepopulationinvolved would be associated with increases in the number of entertain- inactivemilitarydutyduringthesameperiod. ment films released per capita during this period. Achievement. The Achievement value emphasizes demonstrat- Self-Direction. Thisvalueemphasizesindependenceofthought ing competence according to social standards of success, and the and action, and includes the specific values freedom and choosing pursuitofthesevaluesentailsinvestingtimeandenergyintasksthat owngoals(Schwartz,1992).WhenpeoplevalueSelf-Direction,they serve group interests (Schwartz, 1992; Schwartz & Bardi, 2001). are likely to engage in activities that enable them to demonstrate WhenpeoplevalueAchievement,theyarelikelytoengageinactiv- independence of thought and action. Voting in national elections is itiesthatenablethemtodemonstratecompetenceandsuccess.Indeed, onebehaviorthatenablespeopletopursuetheirSelf-Directionvalues Bardi and Schwartz (2003) demonstrated that valuing achievement becausevotingprovidesanopportunitytodemonstratethefreedomto wasassociatedwithinvestingtimeandeffortinachievinghighgrades makeindependentchoices.Wethereforereasonedthatthepercentage at universities among student samples. Patenting an invention is ofvoterturnoutforPresidentialandCongressionalelectionswouldbe anotherbehaviorthatenablespeopletopursueAchievementvalues oneobjectiveindicatorofbehaviorthatexpressesthevalueofSelf- because it is an obvious demonstration of competence and success. Direction.Dataforthisindicatorwereavailablefortheyears1900to Indeed,priorresearchhasdemonstratedthathigherneedforachieve- 2000 (Woolley & Peters, 2005). We expected that increases in the mentscores,asmeasuredinU.S.-publishedreadingtextbooks,were prevalenceofSelf-DirectionvaluesinAmericannewspapersduring associated with increased patents granted per capita in the United this period would be associated with increases in the percentage of Statesfrom1800to1950(deCharms&Moeller,1962).Wetherefore reasonedthatthenumberofpatentedinventionspercapitawouldbe voterturnoutforPresidentialandCongressionalelectionsduringthis one objective indicator of behavior that expresses the value of period. Achievement. Data for this objective indicator of behavior were Universalism. Universalismisaprosocialvaluethatincludesthe availablefortheyears1900to2000(UnitedStatesPatentandTrade- specificvaluesofsocialjusticeandbroadminded(broadmindedbeing markOffice,2005.)Weexpectedthatincreasesintheprevalenceof definedastolerance).Thisvalueemphasizesafocusontheprotection Achievement values in American newspapers during this period ofallothersoutsidetheimmediatein-group(Schwartz,1992).When wouldbeassociatedwithincreasesinthenumberofpatentedinven- peoplevalueUniversalism,theyarelikelytoengageinactivitiesthat tionspercapitaduringthesameperiod. enable them to promote social justice and the protection of others. Hedonism. The value of Hedonism emphasizes gratification of Indeed,instudentsamples,universalismvalueswerecorrelatedwith self-orientedneedsanddesires,andthepursuitofthesevaluesentails donations made to disadvantaged populations in foreign countries seeking pleasure and gratification of desires (Schwartz, 1992; (Bardi&Schwartz,2003).Donationstocharitiesisonebehaviorthat Schwartz & Bardi, 2001). When people value Hedonism, they are enables people to pursue their Universalism values because these likelytoengageinactivitiesthatenablethemtoexperiencepleasure donations protect the welfare of disadvantaged populations and at- andgratifynonessentialneeds.Indeed,valuinghedonismwasasso- tempttorectifysocialinjustice.Wethereforereasonedthatthepro- ciatedwithconsumingfoodordrinksthatthepersonlikes,evenwhen portionoftotalpersonalincomedonatedtotheUnitedWaywouldbe heorsheisnothungryorthirsty(Bardi&Schwartz,2003).Alcohol oneobjectiveindicatorofbehaviorthatexpressesthevalueofUni- consumptionisabehaviorthatenablespeopletopursuetheirHedo- versalism.3NotethatthisbehaviorisabetterreflectionofUniversal- nismvaluesbecausealcoholismostoftenconsumedforpleasureand ismvaluesthanofBenevolencevaluesbecauseoftheemphasison nottorelievethirst.Wethereforereasonedthatalcoholconsumption helping people whom one does not know personally and who are percapitawouldbeoneobjectiveindicatorofbehaviorthatexpresses outsideone’simmediategroup.Dataforthisindicatorwereavailable thevalueofHedonism.Dataforthisindicatorwereavailableforthe for the years 1920 to 1997 (Putnam, 2000). We expected that in- years1900to2000(NationalInstituteonAlcoholAbuseandAlco- creasesintheprevalenceofUniversalismvaluesinAmericannews- holism,2006).WeexpectedthatincreasesintheprevalenceofHe- papers during this period would be associated with increases in the donismvaluesinAmericannewspapersduringthisperiodwouldbe proportionoftotalpersonalincomedonatedtotheUnitedWayduring associated with increases in the amount of alcohol consumed per theperiod. capitaduringthesameperiod. Stimulation. The value of Stimulation emphasizes the pur- 3UnitedWayisoneofthelargestcharitiesintheUnitedStates,working suit of opportunities that bring excitement, arousal, and novel with,forexample,businesses,communitydevelopmentcorporations,vol- experiences(Schwartz,1992).WhenpeoplevalueStimulation, untaryassociations,andvariousfaithcommunities(reflectingthespecific they are likely to engage in activities that enable them to have universalism value of broadminded) to redress various themes of social excitingandarousingexperiencesindailylife.Indeed,valuing injustice.Thesethemesincludepromotingself-sufficiencyandfacilitating stimulation was associated with the frequency of watching women’s leadership (reflecting the specific universalism value of social thrillers (Bardi & Schwartz, 2003). Watching entertainment justice). Thus, this large charity is a particularly suitable charity for the films is one behavior that enables people to pursue their Stim- expressionofuniversalismvalues. ANARCHIVALAPPROACH 489 Benevolence. The Benevolence value emphasizes a prosocial Security. ThevalueofSecurityemphasizessafetyandstability motivation to enhance and protect the well-being of people with of society, including maintaining the social order (Schwartz, whomoneisinfrequentcontact(Schwartz,1992).Itincludesthe 1992). When people value Security, they are likely to engage in specific values helpful and forgiving. When people value Benev- activitiesthatwillenablethemtohaveand/orprovidesafetyand olence,theyarelikelytoengageinactivitiesthatwillenablethem security in situations. Employment in security-based occupations toprotectthewell-beingofpeopleintheirimmediateenvironment. isonebehaviorthatenablespeopletopursuetheirSecurityvalues Indeed, in student samples, Benevolence values were correlated becauseitinvolvespeopledirectlyinmaintainingthesocialorder. withhelpingneighbors(Bardi&Schwartz,2003).Decidingnotto Indeed,amongworkersacross32occupations,peoplewhoworkin deport individuals from the country is one behavior that enables securityoccupationstendedtovaluesecurityhighlyascompared the pursuit of Benevolence values because it provides an oppor- with people in other occupations (Knafo & Sagiv, 2004). We tunity to protect and/or help individuals who are portrayed con- therefore reasoned that the percentage of the working population cretely (with a name, picture, and personal story). We therefore employed as police officers, guards, or public safety officials reasonedthattheproportionofthepopulationdeportedwouldbe would be one objective indicator of behavior that expresses the one objective indicator of behavior that expresses the value of value of security. Data for this indicator were available for the Benevolence (reverse scored). Data for this indicator were avail- years1900to1997(Putnam,2000).Weexpectedthatincreasesin ablefortheyears1900to2000(Kurian,2004).Weexpectedthat theprevalenceofSecurityvaluesinAmericannewspapersduring thisperiodwouldbeassociatedwithincreasesinthepercentageof increases in the prevalence of Benevolence values in American the working population employed as police, guards, or public newspapersduringthisperiodwouldbeassociatedwithdecreases inthepercentageofthepopulationdeportedduringthisperiod.4 safetyofficialsduringthesameperiod. Tradition. The value of Tradition emphasizes commitment andacceptanceofthecustomsandideasoftraditionalculturesor Considerations in the Selection of Behavioral Indicators religions and includes the specific values devout and respect for Because of the obvious challenges in selecting and finding tradition(Schwartz,1992).WhenpeoplevalueTradition,theyare available behavioral indicators for each of the 10 values, it was likelytoengageinactivitiesthatwillenablethemtodemonstrate necessary to utilize one behavioral indicator for each value to theircommitmenttoandrespectfortraditionalcustomsandprac- maintain consistency and make valid comparisons among the 10 tices. Indeed, valuing tradition was associated with observing value–behaviorrelations. traditionalcustomsduringreligiousfestivities(Bardi&Schwartz, Indeed, the behavioral indicators selected may not be the ideal 2003). Moreover, across religions, valuing tradition was highly correlated with religiosity (r (cid:1) .54) and frequency of church representations of each value-expressive behavior. Other indicators attendance(r(cid:1).37;Schwartz&Huismans,1995).Theprevalence mayserveasreasonableapproximationsofthevalue-expressivebe- haviors as well. For example, it could be argued that alcohol con- of churches is one indicator that enables people to pursue their sumption is not the best indicator of Hedonism values. During this Traditionvaluesbecausethebuildingandprevalenceofchurches timeperiod,thereweremanyexternalfactorsthatcouldhaveinflu- demonstrates commitment to traditional customs and religious enced alcohol consumption (e.g., Prohibition, wars), and therefore practices.Thegreaterthenumberofpeoplewhoattendchurch,the people may not have been able to primarily express the value of greaterthenumberofchurchesneeded.Wethereforereasonedthat Hedonismbyconsumingalcohol.Analternativereasonableindicator thenumberofchurchespercapitawouldbeoneobjectiveindicator couldbetheamountofmoneyortimepeoplespentontheirvacations. of the expression of the value Tradition. Data for this indicator However,spendingtimeonavacationwouldalsobeexpectedtobe were available for the years 1900 to 1977 (Melton, 1977). We influencedbysimilaranddifferentexternalfactors.Moreimportantly, expected that increases in the prevalence of Tradition values in similartootherplausiblealternativeindicators,datafortheamountof Americannewspapersduringthisperiodwouldbeassociatedwith timespentonvacationwerenotavailableforthetimeperiodunder increasesinthenumberofchurchespercapitaduringtheperiod. investigationhere. Conformity. ThevalueofConformityemphasizesmaintaining In addition, some of the indicators selected may seem to be the status quo, and the pursuit of this value entails restraint of reasonable expressions of other values. For example, it could be impulses in order to avoid violations of social norms (Schwartz, argued that the indicator of Power values (the percentage of the 1992; Schwartz & Bardi, 2001). When people value Conformity, populationonactivemilitaryduty)couldalsobeanexpressionof they are likely to engage in activities that will enable them to Security values. However, the United States was not under con- adheretosocialnormsandcontributetothestatusquo.Indeed,in tinuous threat between 1900 and 2000, and therefore the popula- student samples, Conformity values were correlated with obedi- tion in active military duty would be more likely to vary as a ence to parents and with avoiding confrontations (Bardi & functionoftheimportanceofdominance(Power)comparedwith Schwartz,2003).Givingbirthaftergettingmarriedenablespeople theimportanceofsafety(Security).Wemaintainthatanyindicator to pursue their Conformity values because marriage before chil- of objective behavior will have a particular set of strengths and dren represents a pervasive social norm in the American culture. weaknesses that may or may not overlap with the strengths and Wethereforereasonedthatthenumberofunwedbirthspercapita weaknesses of alternative indicators. Notably, a primary purpose would be one objective (reversed) indicator of behavior that ex- ofthisresearchwastoexaminewhetherthebehavioralindicators presses the value of Conformity. Data for this indicator were availablefortheyears1940to2000(Kurian,2004).Weexpected 4It was impossible to find a behavior that could be measured on a thatincreasesintheprevalenceofConformityvaluesinAmerican national level that would clearly reflect helping behavior toward people newspapersduringthisperiodwouldbeassociatedwithdecreases withwhomoneisinfrequentcontact.Hence,weusedthepercentageof inthenumberofunwedbirthspercapitaduringthesameperiod. populationdeportedasthebestapproximation. 490 BARDI,CALOGERO,ANDMULLEN we selected do relate to their respective values more than they .93, p (cid:5) .0001,6 demonstrating convergent and discriminant va- relate to the other values when both variables could plausibly be lidityforthevaluelexicon. affectedbyothersocietalfactors(e.g.,wars,economiccycles). Asweexpected,thestructureofrelationsamongthevalueselicited fromthelexicalco-occurrenceofvaluesonnewspaperpageswasnot Results identical to the structure of self-reported values. Similar to self- reported values, the intercorrelations among the lexical co- Lexical Co-occurrence of Value Words in the Value occurrencesofvaluesdeclineastheybecomemoredistantfromeach Lexicon otherinthecircle.However,incontrasttoself-reportedvalues,there is also a slight increase in correlations with opposite values. This WeexpectedthatifthevaluelexiconrepresentsSchwartz’s10 probablyreflectstheconcurrentuseofantonyms,suchas“getoutof values,thenthevaluewordswoulddifferentiallyrepresenteachof yoursafetyzone,havesomeexcitementinlife!”(Securityvs.Stim- these 10 values in their relative lexical co-occurrence in natural ulationvalues)oradebatethatpresentsconflictingvalues,suchasthe language use on the Internet. Table 3 presents the mean JPs of abortion debate in the United States that confronts Tradition values pairs of words representing the same value and the mean JPs of withSelf-Directionvalues. pairsofwordsrepresentingdifferentvalues(i.e.,onewordrepre- senting the target value and one word representing a different value)foreachofthe10values.5Asexpected,themeanJPforthe Predictive Validity of the Value Lexicon co-occurrence of pairs of words for the same value, M (cid:1) .09, is significantlyhigherthanthemeanJPfortheco-occurrenceofpairs BardiandSchwartz(2003)demonstratedmoderate,positivevalue– of words for different values, M (cid:1) .05, F(1, 53) (cid:1) 12.91, p (cid:5) behaviorcorrelationsbetweenself-reportedvaluesandself-andpeer- .0004, corresponding to a medium effect size, r (cid:1) .44. This reportsofactualbehaviors.Therefore,weexpectedthesamegeneral provides stronger support than expected given the nature of the patternsinusingthevaluelexiconandobjectiveindicatorsofvalue- words and the value structure, as noted in the Method section expressivebehaviors,althoughitisreasonabletoexpectmorerandom above.However,thereweretwoexceptionstothegeneralpattern variancewitharchivalindicatorsofactualbehaviors.Table4presents ofresults.TheJPsforAchievementandTraditionsuggestthatthe thecorrelationsbetweentheprevalenceofvaluesinAmericannews- wordsrepresentingAchievementandTraditionco-occurredsimi- papersandtheobjectiveindicatorsofbehaviors.Thecorrelationsof larly with the words representing the other values. We return to values with their corresponding value-expressive behaviors are pre- these exceptions in the Discussion. Overall, the value lexicon sented on the diagonal, in boldface. The mean correlations of the represents words that can significantly discriminate between values with their noncorresponding value-expressive behaviors are Schwartz’svaluesinnaturallanguageuseontheInternet. presentedonthebottomrow.Asexpected,mostofthecorrelationson thediagonalaremorepositivethanthecorrelationsonthesamerow Convergent and Discriminant Validity of the Value andcolumn.Thefewcorrelationsthatdivertfromthispatterncould Lexicon be attributed to random variance that, as noted above, should be expectedwitharchivalindicatorsofactualbehavior.Yet,importantly, If the value lexicon serves as an indicator of individual-level themeancorrelationforthe10correspondingvalue–behaviorpairs, valuesasdelineatedbySchwartz(1992),thenthegeneralpatterns r (cid:1) .56, was much higher than the mean correlation for the 90 observed in the 55 correlations obtained from the lexical co- noncorrespondingvalue–behaviorpairs,r(cid:1).07.Theeffectsforthe occurrenceofthe10valuesinAmericannewspaperpagesshould correspondingandnoncorrespondingvalue–behaviorpairsweresig- exhibitthesamegeneralpatternsasobservedinthe55correlations nificantly different, Z (cid:1) 4.80, p (cid:5) .0001, demonstrating that the obtained from American students’ self-reports of the 10 values values in American newspapers indicated by the value lexicon are (Bardi&John,2007).Indeed,theconvergencebetweenthesetwo very different operationalizations was remarkably high, r(53) (cid:1) stronglyandconsistentlyassociatedwithobjectiveindicatorsoftheir value-expressivebehaviors.Infact,exceedingexpectations,themean correlation for the 10 corresponding value–behavior pairs based on objective indicators of values and behaviors during this 101-year Table3 periodisstrongerthanthemeancorrelationforthe10corresponding MeanJointProbabilities(JPs)ofCo-occurrenceAmongWords value–behavior pairs based on students’ self-reports of values and intheValueLexicon behaviorsreportedbyBardiandSchwartz(2003;acrossthreestudies, r (cid:1) .49). It is also important to note that the correlation between MeanJPofpairs MeanJPofpairs ofwords ofwords Power values and the number of people on active military duty representingthe representing remainedsignificantafterexcludingtheyears1944and1945,which Value samevalue differentvalues werestatisticaloutliersbecausetheywerethepeakyearsoftheU.S. military’sactiveparticipationinWorldWarII.Similarly,thecorre- Power .17 .07 Achievement .06 .05 lationbetweenHedonismvaluesandtheamountofalcoholconsump- Hedonism .06 .03 tion per capita remained significant after excluding the 13 years of Stimulation .04 .02 Prohibition. Self-direction .11 .05 Universalism .07 .05 Benevolence .06 .03 5The complete matrices of 10 (cid:4) 10 correlations of all 10 values Tradition .05 .06 throughoutthearticleareavailablefromtheauthorsonrequest. Conformity .11 .05 6All calculations on correlations in this article are performed with Security .21 .05 Fisher’sZtransformationandbacktransformation. ANARCHIVALAPPROACH 491 Table4 CorrelationsBetweenValuesandObjectiveIndicatorsofBehaviors Values Behavior Self- (correspondingvalue,df) Power Achievement Hedonism Stimulation direction Universalism Benevolence Tradition Conformity Security Militaryparticipation (Power,99) .33 (cid:2).28 (cid:2).26 (cid:2).20 .09 .33 (cid:2).23 (cid:2).26 (cid:2).28 .06 Patentedinventions (Achievement,99) (cid:2).36 .35 .03 .09 (cid:2).05 (cid:2).20 .28 .11 .25 (cid:2).19 Alcoholconsumption (Hedonism,99) (cid:2).41 (cid:2).23 .50 (cid:2).75 (cid:2).24 (cid:2).24 .01 (cid:2).09 .10 .24 Moviesreleased (Stimulation,99) .72 .76 (cid:2).48 .89 .44 .25 .42 .51 .82 (cid:2).17 Votingparticipation(Self- Direction,99) .54 .32 .12 .15 .57 .32 .44 .39 .49 (cid:2).22 UnitedWaydonations (Universalism,76) .79 .28 (cid:2).26 .48 .35 .47 .19 .31 .44 (cid:2).17 Deportations(Benevolence, reversed,99) .41 .48 .14 (cid:2).02 .34 .26 .48 .17 .57 (cid:2).52 Numberofchurches (Tradition,76) .04 .15 (cid:2).17 .11 .13 (cid:2).09 .17 .09 .14 (cid:2).22 Unwedbirths(Conformity, reversed,59) .85 .37 (cid:2).07 .58 .36 .45 (cid:2).12 (cid:2).04 .70 (cid:2).47 Policeandguards employed(Security,96) (cid:2).67 (cid:2).75 .06 (cid:2).41 (cid:2).31 (cid:2).29 (cid:2).57 (cid:2).28 (cid:2).71 .76 Meancorrelationwith noncorresponding behaviors .30 .13 (cid:2).11 (cid:2).01 .13 .05 .06 .10 .25 (cid:2).20 Note. Degrees of freedom refer to correlations between values and corresponding value-expressive behaviors. Correlations of values with their correspondingvalue-expressivebehaviorsareinboldface. Itisimportanttonotetwoexceptionstothisgeneralpattern.The and troughs, which correspond to major historical events during JPs for Power and Tradition suggest that the words representing this101-yearperiodinAmerica.Forexample,Stimulationexhibits Power and Tradition values were similarly associated with corre- adramaticupswingduringthe“RoaringTwenties,”followedbya spondingandnoncorrespondingvalue-expressivebehaviors.Were- precipitousdownswingatthestartoftheGreatDepression.Secu- turn to these exceptions in the discussion. Overall, patterns of indi- rityexhibitsadramaticupswingatthebeginningofWorldWarII, vidualvaluesweresignificantlycorrelatedwiththeirvalue-expressive adramaticdownswingattheconclusionofWorldWarII,thena behavior and more strongly correlated with their value-expressive general increase that lasts until the conclusion of the Cold War. behaviorascomparedwithallotherbehaviors. This illustration demonstrates the utility of the value lexicon for Aplausibleconceptualquestionmightbeposed:whethersome identifyinguniquerelationsbetweenthevaluesandcorresponding oftheseassociationsbetweenvaluesandbehaviorareactuallyjust societal changes compared with noncorresponding societal areflectionofchangesovertime.7Therefore,theanalysesreported abovewerereplicatedwithregressionanalyses,whereinbehaviors weresimultaneouslyregressedonbothyearandvaluesinorderto 7Aplausibleanalyticsuggestionmightbethatthesedatashouldmore determinewhetherthepatternsreportedabovecouldbedismissed properlybesubjectedtoARIMA(autoregressiveintegratedmovingaver- asanartifactofchangesovertime.Asexpected,themean(cid:6)forthe age) analyses (e.g., Box & Jenkins, 1976; McDowell, McCleary, 10correspondingvalue–behaviorpairs,(cid:6)(cid:1).33,ishigherthanthe Meidinger,&Hay,1980).ARIMAanalyseswouldallowtheexamination mean (cid:6) for the 90 noncorresponding value–behavior pairs, (cid:6) (cid:1) ofthepredictionofbehaviorsbyvalueswhiletakingintoaccountcyclical .02, F(1, 98) (cid:1) 10.34, p (cid:5) .0009, indicating that the value– fluctuations, periodicity, and autoregressive nonindependence in these behaviorrelationsremainsignificantwhenaccountingforchanges data.Atatheoreticallevel,wewouldbereluctanttotreatanyvariability overtime.8 indicative of cyclical fluctuations or periodicity as “white noise” to be filteredoutofthemodel.Forexample,todate,therehasbeennotheoretical orempiricaldeterminationastowhetherthefluctuationsinvaluesdepicted An Illustration of the Utility of the Value Lexicon: Value in Figure 1 represent genuine, meaningful period fluctuations or merely Patterns in the United States randomnoisetobefilteredout.Atamechanicallevel,thesedatadonot conform to the requirements of ARIMA modeling. For example, each TheportrayalofvaluesinAmericannewspapersthroughoutthe archival indicator of behavior covers a slightly different portion of the 20thcenturyalsocanbeusedtoexaminechangeinvaluesatthe 101-yearperiodbeingexaminedintheseanalyses,andsomeoftheindi- grouporsocietallevel.Figure1depictsatotalof1,010datapoints catorshavesomemissingvaluesforsomeoftheyearlyincrementsacross capturing variations in the prevalence of values in American the101-yearperiodbeingexaminedintheseanalyses. newspapers. Inspection of this figure reveals several interesting 8Zero-ordercorrelationsamongvaluesandyearareavailablefromthe societalpatterns.Thesevariationsinvaluesexhibitseveralpeaks authorsonrequest.