You are on page 1of 10

Are Surveys on Trust Trustworthy?

Author(s): Alan S. Miller and Tomoko Mitamura


Source: Social Psychology Quarterly, Vol. 66, No. 1 (Mar., 2003), pp. 62-70
Published by: American Sociological Association
Stable URL: http://www.jstor.org/stable/3090141
Accessed: 10-05-2015 23:03 UTC

REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/3090141?seq=1&cid=pdf-reference#references_tab_contents

You may need to log in to JSTOR to access the linked references.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to Social Psychology
Quarterly.

http://www.jstor.org

This content downloaded from 132.203.227.61 on Sun, 10 May 2015 23:03:39 UTC
All use subject to JSTOR Terms and Conditions
SocialPsychology
Quarterly
2003,Vol. 66,No. 1,62-70

Are Surveyson TrustTrustworthy?*


ALAN S. MILLER
TOMOKO MITAMURA
HokkaidoUniversity

In thispaper we explorethevalidityof a surveyquestioncommonlyusedformeasur-


inggeneralizedtrust.Trusthas become a keyvariablein a varietyoffields,including
sociology,social psychology, politicalscience,and economics;therefore theaccuracy
withwhichitis measuredhas profoundimplications formanystudies.Wesuggestthat
ambiguouswordingon thissurveyitemhas led to misinterpretations concerningactu-
al trustlevels,especiallyin a cross-culturalcontext.To testthisclaim,we conductan
extensivesurveyof studentsat UCLA and at Hokkaido University, Japan. Results
stronglysuggestthatthesurveyquestionmeasuresdifferences in cautionlevelsrather
thanin trust.Implicationsof thisresearchare discussed.

Problemswithusingsurveydata are well databases(e.g.,theAmericanGeneralSocial


known,especiallyin cross-national research Survey, theEuropeanValuesSurvey, and the
(for examples, see Box-Steffensmeier, WorldValues Survey).This questionalready
Jacobson,and Grant 2000; Rucinski 1993; has exerteda significant effecton the field
VijverandLeung1997).Giventhatquestions because manyinfluential studieshave relied
translatedinto otherlanguagesnecessarily on it to drawimportant conclusions.
Several
involveslightchangesin the meaning,and researchers, forexample,usingthequestion,
are readand answeredindifferent socialcon- have claimedthattrustis decliningamong
texts,greatcare is needed in measuringkey Americans(e.g.,Putnam1993;Robinsonand
concepts. Jackson 2001), and then have sought to
"Trust"is a key concept that plays an explain this decline. Following Putnam
important roleinmanysocialsituations rang- (1993), Inglehart(2000) rank-ordered coun-
ing frominterpersonal relationshipsto eco- triesbytrustlevelson thebasis ofthissingle
nomic exchange.Thus it is addressedwith questionfromtheWorldValues Survey.He
increasingregularityby sociologists,social then attemptedto relate thisrankingto a
psychologists,
economists, and politicalscien- country'shierarchicalstructure(but with
tists;frequentlytheyuse surveydata and onlylimitedsuccess).
oftentheyemploycross-nationalcompar- Unfortunately, it is unclearwhatexactly
isons.Unfortunately,muchresearchprevious thisquestionis measuring, and whetherthe
as well as current,relieson a singlesurvey above conclusionsare justified. In thisstudy,
questionand oftenhas producedconflicting to explore the limitationsof researchthat
and inconsistentresults.In this paper we uses responsesto the question,we examine
examinethisquestionin depthto understand whatis actuallybeingmeasuredand howthe
moreclearlywhatis beingmeasuredandwhy responsesmightcontaina systematicbias
pastresultshavebeen problematic. based on variouspopulationcharacteristics.
A standardquestionpurporting to mea-
suregeneralizedtrustappearsin manymajor MeasuringGeneralizedTrust

* This project was funded by a grant from the Generalizedtrusttypicallyis definedby


Russell Sage Foundation.We thankToshioYamagishi, researchersas a defaultexpectationof other
Masaki Yuki, Gi-Wook Shin, JohnHoffmann,and people'sgoodwill.In otherwords,itis similar
Kenji Sugiyamaforhelp and comments.Direct corre- to one'sviewofhumannature:someonewith
spondence to Alan S. Miller, Department of
Behavioral Science, Hokkaido University,N10 W7
highgeneralizedtrustbelievesthatmostpeo-
Kita-ku, Sapporo, 060-0810, Japan; amiller@ ple have benign intentions (Miller and
let.hokudai.ac.ip. Kanazawa 2000: Yamagishi 1998). The
62

This content downloaded from 132.203.227.61 on Sun, 10 May 2015 23:03:39 UTC
All use subject to JSTOR Terms and Conditions
SURVEYS ON TRUST 63
AmericanGeneralSocial Survey(GSS) has significantly higherthanthose of Japanese.
been measuringtrustannuallysince 1972. Experimental findingsyieldedsimilarresults.
The GSS is used so widelythatit ofteninflu- Corresponding resultsfromtheWorldValues
ences what questions are asked on other Survey(WVS), however,reportedopposite
majorsurveys. Obviously,
theuse ofthesame results:Japaneseexhibitedmuchhigherlev-
variable across surveyspays dividendsby els ofgeneralizedtrustthandidAmericans.
allowingcomparisonsto be made; thusthe In thisstudywe considerthewordingof
same question on trustnow appears on a theabove questionand howit mightbe tap-
varietyof othersurveysincluding(as stated pingtwo conceptuallydistinctfactors:trust
above) theEuropeanValues Surveyand the and caution.We thentestforbiases in the
WorldValuesSurvey.In fact,itoftenappears reporting of trustrelatedto thisconceptual
as well on individuallydesignedquestion- problem,usingsurveyresultsfromAmerican
naires (e.g., Hayashi et al. 1982; Ho and and Japaneseuniversity students.
Kochen1987).The questionreadsas follows:
Generally speaking,wouldyousaythatmost Conflating TrustWithCaution
peoplecanbe trusted orthatyoucan'tbe too The question underinvestigationdoes
careful indealingwithpeople? notask respondents to choosebetweentrust
Mostpeoplecanbe trusted and distrust,as is impliedin virtuallyall of
Can'tbe toocareful theresearchcitingresultsfromthisquestion.
Rather,respondentschoose between trust
(Some questionnairesadd a thirdcategory: and caution.Yet thereis reason to believe
"Don't know"or "Depends.") thattrustand cautionare notopposites:cer-
This questionfirstappeared in a paper tainlyit is possible fora personto believe
published in the American Sociological mostpeople can be trusted, and at thesame
Review(Rosenberg1956),but notas a mea- timebelievethatitis prudentto be cautious.
sureofgeneralizedtrust. It was partofa five- It is reasonable,forexample,to believethat
questionindexthattheauthorcalleda "faith the overwhelmingmajorityof people will
in people" scale,whichincludeditemsabout nevertryto burglarize one's house,butstillto
basichumannature.Gradually, however, ithas choose to lock one's door. Yamagishiand
becomea stand-alonequestionthatpurport- Yamagishi (1994) foundthat measures of
edlymeasuresgenerallevels of trust.It has trustand measuresofprudencecreatedsepa-
beenusedin a widevariety ofstudiesoverthe ratefactors; thisfinding suggeststhatcaution
yearstoexploretherelationship betweentrust does notnecessarily implydistrust.
and such diverse topics as religiosity Thisdistinctionbetweencautionand dis-
(Schoenfeld1978), maritalstability(Yoder trustcan be understoodmostclearlybysepa-
and Nichols1980),social networks(Ho and ratingthe questionunderinvestigation into
Kochen 1987),job satisfaction (Liou, Sylvia, its two components."Would you say most
andBrunk1990),cross-national differencesin people can be trusted?"asks foran assess-
social structure(Yamagishiand Yamagishi mentof otherpeople's trustworthiness. It
1994),democracy(Fukuyama1995),culture does notask therespondent abouthisor her
(Inglehart2000),and socialorder(Millerand behavior,but seeks a general appraisal of
Kanazawa2000).Yetdespiteitspopularity, the otherpeople's behavior.The second halfof
validityof thisquestionhas not been con- the question,however,asks people whether
firmed, andresultsbasedon itsometimes have theybelievethat"you can'tbe too careful."
contradicted otherresults. Unlikethefirsthalf,thisportionasks about
ToshioYamagishiand his colleagues,for one's ownbehavioralpreference ratherthan
example,conductedbothsurveyand experi- thatofothers.In doingso,ittapstherespon-
mental studies comparingtrustlevels of dent'swillingness to be vulnerable.In other
Japaneseand Americans.Theirsurveyscon- words,it asks fora self-evaluation regarding
tained a varietyof questionsconstituting a therespondent's degreeof comfort in taking
"generalizedtrustscale,"whichshowedthat risks.Thusthetwohalvesofthisquestionare
Americans'levels of generalizedtrustwere conceptually distinct.Obviouslyitis possible

This content downloaded from 132.203.227.61 on Sun, 10 May 2015 23:03:39 UTC
All use subject to JSTOR Terms and Conditions
64 SOCIAL PSYCHOLOGY QUARTERLY
fora risk-averse personto feelthatpeoplein STUDY DESIGN
general are trustworthy,but still to be
inclinedto be carefulin dealingwithothers. The above discussionsuggeststhatit is
Thus,how one answersthe above ques- critical to understanding moreclearlywhatis
tionwill depend on two factors:an assess- actually being measured when researchers
mentof otherpeople's trustworthiness, and use this standard question on generalized
an assessmentof one's own willingnessto trust. In this research we compare Japanese
take risks.The latterfactorgreatlycompli- with Americans to see whether the potential
cates interpretationof the response. For problemdescribedabove actuallyexists.
example,a wide arrayof studieshas shown A U.S.-Japancomparisonis appropriate
consistentlythat females are more risk- for two reasons.First,as discussed above,
averse than males (Gove 1985; Miller and contradictory resultshave been obtained
Stark2002; Powell and Ansic 1997;Veevers when different questionshave been used to
and Gee 1986). This findingmightexplain measure trust in thesetwocountries. Second,
whythe AmericanGeneral Social Survey, if the problem involves differences in caution
usingtheabove question,showsmalesto be levels rather than in trust, a U.S.-Japan com-
significantly higherin trustthan females, parison should reveal this problem because it
whileotherstudiesthatuse a varietyofques- is likely that Japanese are much less cautious
tions to measure trustshow no significant thanAmericans. Thislikelihoodis basedboth
genderdifference (YamagishiandYamagishi on the prevailing crime rates and on the
1994).Riskaversionand feelingsofpersonal social structure. The U.S. suffersfromvery
vulnerability, as well,also tend to increase high crime rates, while Japanhas the lowest
withage (Greve 1998) and withthe level of crime rates in the world (Miller and
in
crime a particular area (Perloff1983; L. Kanazawa 2000). Furthermore, theJapanese
Smithand Hill 1991).Thislatterobservation social structure is designed so thatpeople
is particularlytroublesomebecause changes spend much of their lives interacting primari-
in generalizedtrustovertimeor acrosssoci- ly with those persons with whom they havea
eties mightreflect,at least in part,demo- long-term relationship (Hechter and
graphicshiftsor differences in crimelevels Kanazawa 1993). In this type of closed social
ratherthandifferences in generalizedtrust environment, cautionis notnecessary: people
perse. do not cheat one another because the conse-
This distinctionhas profoundimplica- quences of harming those relationships are
tions because it could underminea great severe (Braithwaite 1989; Kollock 1994).
manypast studies.For example,studiesthat Yamagishi(1998) refersto thissocial struc-
haveused thisquestionto suggestthatsocial tureas assurance-based: thatis,itassuresthat
trustrecentlyhas declinedin the U.S., (e.g., people will be treated fairly by others.Thus
Putnam1993; Robinsonand Jackson2001) Japan's low crime rate and closed social
insteadmaybe witnessing an increasein cau- structurecreate an environmentin which
tion.Studiesclaimingthatone societyis high- cautionshouldbe low,and thislow level of
er in trustthananothermightbe misledby cautionmightbe misinterpreted as highlev-
differences in thedegreeto whichcautionis els of generalized trust.
more necessaryin one societythan in the The above theoreticaldiscussionsug-
other.If theseweremerelytwo sides of the gests thatthesurveyquestionunderinvesti-
samecoin(i.e.,iftrustand cautionrepresent- gation conflates trustwith caution, and
ed a singledimension),therewould be no because Americansprobablyare morecau-
If
problem. theyare conceptuallydistinct, tious than Japanese,theywill appear to be
however,interpretation becomesproblemat- less trusting. Iftrust/distrust is measuredas a
ic:itwouldthenbe possibleto be highinboth single concept, however,Americans will
trustand caution, or vice versa. In other appear to be more trusting than their
words,a personmightscorehighon a gener- Japanesecounterparts.
al trustscale,butifthatpersonis risk-averse, Hypothesis: Japanesewill score higheron
he or she mightchoose caution over trust trustwhentrustis measuredas a dichoto-
whenfacedwithonlythosetwochoices. mous trustversus caution variable, but

This content downloaded from 132.203.227.61 on Sun, 10 May 2015 23:03:39 UTC
All use subject to JSTOR Terms and Conditions
SURVEYS ON TRUST 65
Americans willscorehigherwhenit is mea- thinkyou can trustneighbors?(2) Do you
suredas a trust
versusdistrustvariable. thinkyoucan trustpeopleinyourschool?(3)
Do you thinkyou can trustcasual acquain-
We testthishypothesis usinga varietyof
tances?and (4) Do you thinkyou can trust
surveys.It mightseem inappropriate to use
otherAmericans/Japanese? The interperson-
surveyresearchas a methodforexploring
limitations insurveyresearch, butitis consid- al caution scale comprises four similarques-
erablyeasierto uncoverweaknessesin exist- tions:(1) Do you think you need tobe careful
ingsurveysthanto designsurveysthatdo not around neighbors? (2) Do you thinkyou
containthose weaknesses.For the current need to be careful around people in your
study,we test the hypothesisusingsurvey school? (3) Do you think you need to be
data collected at Hokkaido Universityin careful around casual acquaintances? and (4)
Japan,and at UCLA; bothare major,urban Do you think you need to be careful around
universities.The resultswillbe tentativeand other Americans/Japanese?We use both
not easily generalizable,but if significant Cronbach'salpha and principal-components
biases appear in suchhomogeneoussettings analysisto assessthedegreeto whichdistrust
(where all respondentsare similarin age, and cautionare different factors.We thentest
educationallevel, and even SES), it seems whether differences in levels of caution
likelythatbiasesalso willexistfora broader, betweenAmericansand Japaneseaccount
more heterogeneouspopulation.Certainly fortheresponsepatternsobserved.
any biases uncoveredin this studywould
warrantfurtherinvestigation.They also RESULTS
would serve as a warningto those people
who relyon this one question to measure Table 1 displays a comparison of
cross-nationaldifferencesin generalized Japanese and American responses to the
trust. question under investigation. Japanese
The sample consistsof undergraduate appear to have higherlevels of generalized
students in the Faculty of Letters at trustthan do Americans, and thedifference is
Hokkaido University(n = 293) and in the statistically significant. The levelsare higher
CollegeofLettersand Scienceat UCLA (n = thanin the nationalresultsobtainedin the
169). No significant sex differencesemerged 1997WVS, in whichtrustlevels are 43 per-
in the study;thus males and females are cent for Japanese and 36 percent for
pooled throughout theanalyses. Americans,but the difference betweenthe
To assess the validityof the question twocountries is aboutthesame.
underinvestigation, we beginbycomparing it Table 2, however,shows verydifferent
with a question that measures truston a results.As predictedin thehypothesis, when
seven-point scale and does notincludea ref- "caution" is removed and trust is measured
erenceto beingcareful:"Do you thinkmost on a scale,Americansappear to have higher
people can be trusted?"Responses range levels generalizedtrustthan do Japanese.
from1 (cannot trustat all) to 7 (complete Once again,thedifference is statistically
sig-
trust).In addition,we createan "interperson- nificant.
al trust"and an "interpersonal caution"scale These resultssuggestthattheGSS/WVS
to assess theirrelationshipto Rosenberg's question,indeed,presentsa problem.The
originalquestion.The interpersonaltrust resultsnot onlydifferwhencomparedwith
scale consistsof fourquestions:(1) Do you our Likert-scaled item,butalso are inconsis-

Table 1.The StandardMeasure of GeneralizedTrust

Japanese Americans
Generallyspeaking,would you say thatmostpeople can be trustedor thatyou
can't be too carefulin dealingwithpeople?a
Most people can be trusted 60% 50%
Can't be too careful 40% 50%
a
Chi-square= 3.8; df= 1;p < .05

This content downloaded from 132.203.227.61 on Sun, 10 May 2015 23:03:39 UTC
All use subject to JSTOR Terms and Conditions
66 SOCIAL PSYCHOLOGY QUARTERLY
Table 2. A LikertScale Measure of GeneralizedTrust

Japanese Americans
Do you thinkmostpeople can be trusted?
1. Not trustat all 3% 0%
2. 6% 3%
3. 24% 8%
4. 29% 36%
5. 30% 48%
6. 7% 4%
7. Completetrust 1% 1%
Mean 4.0** 45**
** p < .01 (t = 3.6)

tentwithfindings on everyotherquestionwe averagelevel of trustis stillabove the mid-


asked, which consistentlyshow Japanese pointof4 on thescale.
respondents to be lower in trust than Is theproblemdue primarily to theuse,
Americans. To explore this problem in in the question,of thewordcaution,as pre-
greaterdetail,we considerin moredepththe dicted in the hypothesis?To resolve this
relationshipbetweenthe dichotomousitem question,we asked a seriesof morespecific
and our Likert-scaleditem (Table 3). The questionsthatdisaggregatetrustfromcau-
uppersectionofthetablecomparesJapanese tion; the resultsare presentedin Table 4.
and Americanrespondentswho chose the Whenaskedaboutspecificgroupsofpeople,
midpoint(4) on the seven-pointtrustscale. Americansconsistently displaysignificantly
AmongJapaneserespondents, 58 percentfelt higher levels of trustthan do Japanese.
Indeed,the levels of trustamongJapanese
that4 was highenoughto justifychoosing
respondents are extremely lowwhenwe con-
truston the dichotomousmeasure.In sharp
siderthat60 percentindicateda generaltrust
contrast, only 29 percent of American
in people.The lowerpart of Table 4 shows
respondentswho chose 4 feltthatit was a
thatalthoughAmericansare more trusting
highenoughlevelof trustto justifychoosing
thanJapanese,theyare also morecautious:
truston the dichotomousvariable.Clearly, theyconsistently displayhighlevels of cau-
Americansrequirea higherlevel of trustin tion despite theirrelativelyhighlevels of
orderto choose thatoptionin the dichoto- ThusthechoicefacingAmericansinthe
trust.
mousquestion. initial question-between trustand cau-
The lowersectionofthetableselectsout tion-is difficultbecause Americanstendto
onlythoserespondents who chose"caution" be highin both.
ratherthan"trust"on theinitialdichotomous To confirm thatdistrust and cautionare
variable. Even among those "cautious" conceptuallydiscrete,Table 5 presentsthe
Americans, levelsofgeneraltrustare signifi- resultsof a principal-components analysis
cantlyhigherthan among theirJapanese usingthe eightquestionsshownin Table 4.
counterparts.Indeed,evenin thissubset,the The resultsare unambiguous:two separate
Table 3. How Much TrustIs Needed to Overcome Caution?

Japanese Americans
Japanese(N = 79) and American(N = 41) respondentswho chose the
midpoint(4) on the Likertscale of trusta
Most people can be trusted 58% 28%
Can't be too careful 42% 72%
Mean trustscoresforJapanese(N = 110) and American(N = 60) who
chose "caution"on the dichotomousvariable
Mean 3.3*** 4.1***
a Chi-square= 10.3;df= 1;p < .001
b Measured on a seven-point Likertscale (see Table 2)
***p< .001 (t = 4.5)

This content downloaded from 132.203.227.61 on Sun, 10 May 2015 23:03:39 UTC
All use subject to JSTOR Terms and Conditions
SURVEYS ON TRUST 67
Table 4. CautiousTrust:A Comparisonof Japaneseand Americans

Japanese Americans
(% Yes) (% Yes)

Can you trustthefollowingpeople?


Neighbors 20 52
People in school 36 47
Acquaintances 10 28
Strangers 4 24
Do you feelyou need to be carefularoundthe followingpeople?
Neighbors 42 43
People in school 32 68
Acquaintances 56 78
Strangers 57 69

AnalysisofTrustand Caution
Table 5. Principal-Components

Factors

Variables Trust Caution

TrustNeighbors .761 -.039


TrustPeople in School .697 -.128
TrustAcquaintances .837 -.056
TrustOtherAmericans/Japanese .817 .054
CautiousWithNeighbors -.158 .638
Cautious at School -.030 .774
CautiousWithAcquaintances -.055 .693
Cautious WithAmericans/Japanese .064 .666

factorsemerged,one fortrustand one for Resultsare presentedinTable6. Model 1


caution.Therefore, can we concludethatthe showstherelationship betweencountryand
addeddimensionofcautionis thereasonwhy generalizedtrust.The negativecoefficient
trustlevels differwhen this dichotomous means that Japanese respondentsexhibit
variableis used ratherthanothervariables? higherlevelsof trust,whenmeasuredby the
To testthisclaimdirectly,we createan inter- dichotomousvariable, than do American
personaltrustscale and an interpersonalcau- respondents.These resultsreplicatethose
tion scale, usingthe variablesfromTable 5 presentedinTable 1. Model 2 adds theinter-
(Cronbach's alphas are .80 and .69 respec- personaltrustscale comprisingthetrustvari-
tively).Thenwe testwhethertheU.S.-Japan ables showninTable 5. Inclusionofthisscale
differences are reducedby includingthese does not attenuatethe cross-nationaldiffer-
scales, particularlythe caution scale, in a ences.Inclusionofthecautionscale inModel
logisticregressionmodel withthe dichoto- 3, however,reducesthecross-national differ-
mousmeasureoftrustas thedependentvari- ences to statisticalinsignificance.
In other
able. words,thecross-national differences
record-

Table 6. U.S.-JapanDifferencein GeneralizedTrust(As Measured by the DichotomousVariable),With


ControlsforInterpersonalTrustand Cautiona

Model 1 Model 2 Model 3


Country(1 = U.S.) -.48** -.82*** -.24
(.18) (.20) (.21)
TrustScale .29***
(.04)
Caution Scale -.28***
(.06)
aLogisticregressionestimateswithstandarderrorsin parentheses.
** p < .01; ***p < .001

This content downloaded from 132.203.227.61 on Sun, 10 May 2015 23:03:39 UTC
All use subject to JSTOR Terms and Conditions
68 SOCIAL PSYCHOLOGY QUARTERLY
ed bythedichotomous variablereflectdiffer- BYU students.Also like the Japanesestu-
encesin cautionratherthanin trust. dents,the BYU studentsdiffermost from
Althoughourprimary purposehereis to UCLA students in their overall caution
testthevalidity ofthequestionunderinvesti- levels.
gation,it is usefulto understandwhatenvi-
ronmental characteristics are likely to DISCUSSION
producebiased responses.In thispaper we
We beganbynotingthata growingliter-
suggestthatthetrustfiguresare inflatedfor
aturefocuseson generalizedtrust,and that
Japanbecause people therelive in a safe,
muchof thisis based on a popular survey
closedenvironment, whichleadsto unusually
research question. We then pointed out
low levelsof caution.If thisis true,themea-
inconsistencies in findingswhenthatques-
surement problemis notlimitedto Japan:for
tion,and other,moreexhaustivemeasuresof
anygrouplivingin suchan environment, lev-
generalizedtrustwereused.We hypothesized
els ofgeneralizedtrustwillappearto be high-
that the differencesmightbe due to the
er thantheyactuallyare whenmeasuredby
dichotomous natureofthequestion,inwhich
thisdichotomous variable. trustis pittedagainstcaution.Because ofthis
To testthisclaim,we conducteda small wording,we believed responses could be
follow-up surveyamongstudentsat Brigham interpretedas measuringcaution levels as
Young University(BYU). We chose BYU wellas trustlevels.Despite limitsin size and
because its social environment mostclosely demographicdiversityin the sample, we
resemblesthatofJapanin thecharacteristics found strong differencesin how people
relevant to this study: the crime rate is answeredthe questionunderinvestigation,
extremely low in Provo,Utah,and thepopu- and thehypothesis was stronglyconfirmed.
lationis ethnicallyand religiously homoge- As we stated at the beginningof this
neous. The religiouselementprovidesthe paper,the actual questionnaireused in this
typeof assurancenetworksfoundin Japan. studycontainedmorethan100 measuresof
That is,maintaining a good reputationand trust, and in all instances except one,
goodrelationships withothersis quiteimpor- Americansscored significantly higherthan
tantina closed,religiouscommunity; thusthe Japanesein trust.The order was reversed
likelihoodofbeingcheatedbyothersis rela- onlyin the questionunderinvestigation. A
tivelysmall. wide varietyof testsstronglysuggeststhat
summarized
The results, inTable7,paral- thisdiscrepancy is due to theuse oftheterm
lel thosefortheJapaneserespondents. When careful. Whenthetrust-versus-caution option
trustis measuredwiththedichotomous vari- was replacedwithtrustversusdistrust, the
able,BYU studentsappearto have extreme- cross-national differences disappeared.More
lyhighlevelsofgeneralizedtrust. Whentrust important, trustandcautionprovedto be dis-
is measuredon a simpleLikertscale,howev- cretefactors,and cautionlevels accounted
er,whichremovescautionas an optionand forthe main differencebetweenJapanese
replacesitwithdistrust,we findno statistical- andAmericanrespondents. We replicatedthe
lysignificant differencebetweenUCLA and results with students at Brigham Young

Table 7. Measures of GeneralizedTrust:UCLA and BrighamYoung University

BYU UCLA
Generallyspeaking,would you say thatmostpeople can be
trustedor thatyou can't be too carefulin dealing
withpeople?a
Most people can be trusted 72% 50%
Can't be too careful 28% 50%
Do you thinkmostpeople can be trusted?b
Mean 4.63 4.46
a Chi-square= 14.7;df= 1;p < .001
b
Measured on a seven-pointLikertscale (See Table 2); t = 1.62 (not significant)

This content downloaded from 132.203.227.61 on Sun, 10 May 2015 23:03:39 UTC
All use subject to JSTOR Terms and Conditions
SURVEYS ON TRUST 69
University,and showed that this type of however,tend to be more tenuousbecause
responsepatterncan be generalizedto other this surveyis designed for cross-national
populationsthatdifferin overall levels of comparison and contains only the one,
caution. dichotomous question.
Obviously the next step is to test the
CONCLUSION validityof this question more completely
usingsurveyresearchthatsamplesrandomly
Our studyis preliminary; resultsbased a broaderrangeofpeople,preferably in sev-
exclusively on a studentsampleare difficult eral societies.If similarbiases emerge,find-
to generalize.Our results,however,suggest ingsthatreliedexclusivelyon thisquestion
thatdifferences in levelsofsafetyin an envi-
mustbe reconsidered, and a new,moreaccu-
ronmentstronglyaffectsurveyresults.If ratemeasureoftrustmustbe developed.Our
these differencesappear among relatively
resultssuggestthatmeasuring truston a scale
homogeneoussamplessuchas university stu- withtrustat one end
and distrustat theother
dentsin two highlymoderncountries(and
elicitsunambiguous responses.
even among universitystudents at two
On a more positivenote,our research
Americanuniversities), the problemalmost
suggeststhattrustlevelscan be generalized.
certainly is magnified bycomparisonsacross In
ourquestionnaire, we askeda widevariety
50 different societies,as in theWorldValues
of questionson manytypesof generaland
Survey.
interpersonaltrust(e.g., trustin relatives,
Thisstudy, therefore, shouldbe viewedas
friends, neighbors, classmates,people of the
a warningabout drawingconclusionsbased
samereligion, peopleofthesamenationality,
on a single surveyquestion,especially in
strangers,those who look kind). In all
studiesthatseek to addresschangeovertime
instances, patterns were consistent:
or differences across societies.At the very
Americansexhibitedhigherlevels of trust
least, studies of trustshould use multiple
thantheirJapanesecounterparts. Thisfinding
indicators.Afterall,thepurposeofmultiple-
suggeststhatif one measures"trustin peo-
indicatormodels is to attenuatemeasure-
ple" accurately, one can generalizethosetrust
mentand conceptualproblemsassociated
levelsacrosscategoriesofpeople.Suchinfor-
witha singlevariable.Fortunately, manyof
mationshouldgreatlyaid futurestudieson
themorethorough paststudiesused multiple
thistopic.
indicators. In addition,althoughcautionand
trustare conceptuallydistinct,fortunately
REFERENCES
theyare relatedtopicsthatcan have similar
implications.For example, both concepts Box-Steffensmeier, JanetM., Gary C. Jacobson,
mightbe relatedto topicssuchas socialcapi- and Tobin Grant.2000."QuestionWording
tal:one can arguethatsocialcapitalis a func- and the House Vote: Some Experimental
tionnotonlyof generalizedtrustbutalso of Evidence." Public Opinion Quarterly
caution.Justas a reductionin trustis regard- 64:257-70.
ed as a reductioninsocialcapital,so mightan Braithwaite, John. 1989. Crime, Shame and
Reintegration. Cambridge,UK: Cambridge
increasein cautionrepresentsuch a reduc- University Press.
tionbecause it also inhibitsthe creationof Fukuyama,Francis.1995.Trust:The Social Virtues
new relationships. Thus our resultsdo not and Creationof Prosperity. New York:Free
necessarilyunderminesome of the recent Press.
studies on the topic. For example,Paxton Gove, WalterR. 1985. "The Effectof Age and
(1999) andTomSmith(1997) bothexaminea Gender on Deviant Behavior: A
wide arrayof variables,use multiple-indica- Biopsychosocial Perspective."Pp. 123-46in
tormodels,and assess trendsrespectively in Genderand theLifeCourse,editedbyAlice
S. Rossi.NewYork:Aldine.
social capital and misanthropy. There is no
Greve,Werner.1998."Fear of CrimeAmongthe
reasonto suspectthattheconflating of trust Elderly:ForesightNot Fright."International
withcaution on a single indicatorin their ReviewofVictimology 5:277-309.
models affectstheir conclusions greatly. Hayashi,Chikio,TatsuzoSuzuki,GiichiroSuzuki,
Studiesbased on the WorldValues Survey, and MasakatsuMurakami.1982.A Studyof

This content downloaded from 132.203.227.61 on Sun, 10 May 2015 23:03:39 UTC
All use subject to JSTOR Terms and Conditions
70 SOCIAL PSYCHOLOGY QUARTERLY
JapaneseNationalCharacter. Vol. 4. Tokyo: Community:Social Capital and Public
Idemitsushoten. Affairs." The American Prospect
Hechter,Michael and Satoshi Kanazawa. 1993. (Spring):35-42.
"Group Solidarity and Social Order in Robinson,RobertV.andEltonF.Jackson. 2001."Is
Japan." Journal of Theoretical Politics Trustin OthersDecliningin America?An
5:455-93. Age-Period-Cohort Analysis." Social
Ho, Edricand ManfredKochen.1987."Perceived ScienceResearch30:117-45.
Acquaintanceshipand Interpersonal Trust: Rosenberg, Morris. 1956. "Misanthropy and
The Cases ofHong Kongand China."Social PoliticalIdeology."AmericanSociological
Networks 9:153-69. Review21:690-95.
Inglehart, Ronald. 2000. "Culture and Rucinski, Dianne. 1993. "A Review: Rush to
Democracy."Pp. 80-97 in CultureMatters, Judgment? Fast ReactionPolls in theAnita
editedbyLawrenceE. Harrisonand Samuel Hill-ClarenceThomasControversy." Public
Huntington. NewYork:Basic Books. OpinionQuarterly 57:575-92.
Kollock, Peter. 1994. "The Emergence of Schoenfeld,Eugen. 1978. "Image of Man: The
Exchange Structures:An Experimental Effectof Religion on Trust."Review of
Study of Uncertainty,Commitment,and ReligiousResearch20:61-67.
Trust." American Journal of Sociology Smith,LynnNewhartand Gary D. Hill. 1991.
100:313-45.
"Perceptions ofCrimeSeriousnessand Fear
Liou, Kuo-Tsai,Ronald D. Sylvia,and Gregory
ofCrime."SociologicalFocus24:315-27.
Brunk.1990. "Non-WorkFactors and Job
Smith, Tom W. 1997. "Factors Relating to
SatisfactionRevisited."Human Relations
Misanthropyin ContemporaryAmerican
43:77-86.
Society."Social ScienceResearch26:170-96.
Miller,Alan S. and SatoshiKanazawa.2000.Order
Veevers,JeanE. and Ellen M. Gee. 1986."Playing
byAccident:The Originsand Consequences
of Conformityin ContemporaryJapan. It Safe: Accident Mortalityand Gender
Boulder:Westview. Roles."SociologicalFocus19:349-60.
Miller,Alan S. and RodneyStark.2002."Gender Vijver,Fons J.R. van de and Kwok Leung.1997.
and Religiousness: Can Socialization Methods and Data Analysis for Cross-
Arguments Be Saved?"AmericanJournalof CulturalResearch.NewburyPark,CA: Sage.
Sociology107:1399-1423. Yamagishi,Toshio.1995."Have AmericansReally
Paxton,Pamela.1999."Is Social CapitalDeclining BecomeDistrustful?" Presentedat theannu-
in the UnitedStates?A MultipleIndicator al Meeting of the American Sociological
Assessment." AmericanJournal ofSociology Association,August19-23,Washington, D.C.
105:88-127. . 1998. The Structure of Trust(Shinraino
Perloff, Linda S. 1983. "Perceptions of Kozo). Tokyo:University ofTokyoPress.
Vulnerability to Victimization."
Journalof Yamagishi,Toshio and MidoriYamagishi.1994.
Social Issues39:41-61. "TrustandCommitment intheUnitedStates
Powell,Melanie and David Ansic.1997."Gender and Japan." Motivation and Emotion
Differencesin Risk Behavior in Financial 18:129-66.
Decision-Making: An Experimental Yoder,JanD. and RobertC. Nichols.1980."A Life
Analysis."Journalof EconomicPsychology PerspectiveComparison of Married and
18:605-28. DivorcedPersons."Journalof Marriageand
Putnam, Robert. 1993. "The Prosperous theFamily(May):413-19.

Alan S. Miller is professorof social psychologyat Hokkaido University,Japan.His main


researchinterestis religion,and his mostrecentarticle,"Genderand Religiousness:Can
SocializationExplanationsBe Saved?" withRodneyStark,was publishedin The American
Journalof Sociology (May,2002). He is authorof Order By Accident:The Originsand
Consequencesof Conformity in Contemporary JapanwithSatoshiKanazawa. His current
research,withJohnHoffmann, explorestherelationship
betweenreligionand trust.

TomokoMitamurarecently receiveda Master'sdegreein social psychology


fromHokkaido
University,Japan,and is now a researcher
for theelectric
powerindustry'sCentralResearch
Institute,
SocioeconomicResearchCenter, in Tokyo.

This content downloaded from 132.203.227.61 on Sun, 10 May 2015 23:03:39 UTC
All use subject to JSTOR Terms and Conditions

You might also like