Response Style and Cross-Cultural Comparisons of Rating Scales Among East Asian and North American Students

This report examines cross-cultural differences in response style regarding the use of rating scales Subjects were high school students 944 from Sendai (Japan), 1,357 from Taipei (Taiwan), 687 from Edmonton and Calgary (Canada), and 2,174 from the Minneapolis metropolitan area and Fairfax County, Virginia Responses to fifty-seven 7-point Likert-type scales were analyzed The Japanese and Chinese students were more likely than the two North American groups to use the midpoint on the scales, the U S subjects were more likely than the other three groups to use the extreme values Within each cultural group, endorsement of individualism was positively related to the use of extreme values and negatively related to the use of the midpoint These small, albeit statistically significant, differences in response styles generally did not alter cross-cultural comparisons of item means

Ratmg scales arc among the most widely used tools m psy· chology Nevertheless, systematic biases m responding to rat· mg scales have been reported (Berg, 1967, Couch & Keniston, 1960, Cronbach, 1946, 1950, Hamilton, 1968, Rorer, 1%5, Schuman & Presser, 1981 Recent research has pointed to pos· s1ble cultural differences m the extent of response biases (e g , Bachman & O'Malley, !984a, 1984b. Hui & Tnand1s, 1989, Mann, Gamba, & Mann, 1992 For example, O'Malley (1984a, 1984b) found that Afncan-Amencan adoles· cents were more likely than white adolescents to select extreme values such as "agree" or "disagree" as opposed to "mostly agree·• or "mostly disagree., They suggested that lh1s extreme response style might account for the differences m sclf-es1eem often reported between Afncan Amencans and whites Smularly, concerns have been expressed aboul possible contanunatmg influences of response biases m cross-natJonal compansons (e g , Leung, 1989, Leung & Bond, 1989 One sign of these influences 1s the consistent finding that compared with therr Western peers, Asian children rate themselves and are rated by theu mothers as havmg lower levels of abthty and less pos1t1ve personahty chatactcnst1cs (c g , Stevenson et al , 1990, StJgler. Smith, & Mao, 1985 It appears that Asians may be influenced by the vutues of moderation promoted by Confucian philosophy and believe they should not stand out from the group It 1s important to consider, therefore. the degree to which these cross-cultural differences tn ratings are the result of d1f· ferences m response style and the degree to which they reflect true cultural ddferences m the level of self-evaluat1ons An un· Copynght C 1995 Amencan Psychological Society derstandmg of the extent and the nature of ddferences m the use of ratmg scales has theoretJcal as well as methodological 1mpb· catwns for cross-cultural research Bias m ratmgs may be related to other charactenst1cs. such as the social onentat1on of md1v1duals w1thm a society Markus and K1tayama (1991), for example, have suggested that the apparent modesty of Asians reflected m their ratings may be related to the emphasis 1n Asian cultures on an mterdepcndency among md1v1duals that pro· motes other-enhancing biases This article thus has three purposes The first 1s to examine the extent and nature of cross-cultural differences m the use of ratmg scales We examine Chmese, Japanese, Canadian, and US high school students' responses to L1kert-hke items related to achievement attitudes and beliefs By mcludmg representatives of two East Asian and two North Amencan cultures, we are able to gam some ms1ght into the consistency with which the potential response biases may operate w1thm the two cultural groups We ask, therefore, whether the effects are greater between the two cultural groups than between the two representative cultures w1thm each cultural group In other words, we attempt to deterrmne whether response bias might have a consistent influence on East Asian and North Amencan com pan sons The second purpose 1s to te<;t !he relation w1thm each group belween the onentat1on toward md1v1duahsm or collect1v1sm and response style We hypothesize that members w1thm each group who endorse higher degrees of md1Vtduahsm are more likely to use the extreme values on a scale and less hkely to use the m1dpomt Those who arc more collect1v1st onented are hypothesized to do the opposite The final purpose 1s to examine the effects or cultural differences m response styles on item means m eross·cultural compansons We assess the magnitude of polent1al d1stortmg effects through two transformations of scales following the approach used by Bachman and O'Malley (1984b)

Subjects
The subJects constituted representative samples of l Ithgrade students m six metropolitan areas Sendai (Japan), Taipei (fai.wan), Edmonton and Calgary (Canada), and Mmneapohs and Fairfax County, V1rg1ma Of the 6,451 students who par-t1c1pated m the study, 80% responded to the items exammed m the present analyses The final sample mcluded 944 Japanese, 1.357 Chinese, 687 Canadian, and 2.174 U S students The percentages of males were 56% (Japan), 42% (Tai.wan), 51% (Can· ada), and 4990 (Urutcd States) IRVINE on June 18, 2015 at UNIV CALIFORNIA pss.sagepub.com Downloaded from PSYCHOLOGICAL SCIENCE Chuansheng Chen, Shm-ymg Lee, and Harold W Stevenson A questionnaire for 11th-grade students was designed s1multancously in English, Chinese, and Japanese by a group of b1tmgual researchers and graduate students The questmnnaire covered a broad range of topics concerning ideas, values, attitudes, beliefs, and self-evaluations related to school and dady hfe We focus here on the fifty-seven 7-pomt L1kert-hke 11ems included m the questionnaire Among them were 39 items that formed seven mult1ple-1tem scales, the re\t were smgle items The seven scales all had satisfactory mtemal con\1\tency (see Table I for the Cronbach alphas and sample items) Furthermore, there were few cross-cultum.I differences m the rehab1hty stattst1cs A subgroup of the US, Chmese, and Japanese \tudents was also given four items dealmg with their onentatmn toward m-d1v1dual1sm or collect1v1~m bee Table 2 for the item\) We used these items to examine the rela!mn w1thm each cultural group between the onentat1on toward md1viduahsm or colleet1v1sm and response style Although these items were mtercorrelatcd m the expected dtrcction, the mtemal consistency amona them was not as high as would be deslfCd (Cronbach alphas were 55, 36, and 52 for Japan, Taiwan, and the United States, respectively), perhaps because of the diverse aspects ofmd1V1dual1sm and collect1v1sm that were measured Thus, we considered these items both as md1vidual items and as a composite scale

Use of Midpoint and Extreme Scale Values
As shown m Table], on the one hand, Japanese students were more hkely than Chmese students to use the midpoint (4), and Chmese students m tum were more hkely to use the m1dpomt than e11her the Canadian or the U S students {Schcffe contrasts, ps < 001) On the other hand, the US students were more likely than all the other three groups to use the Generally, people my age should be l4 l 9 63 72 51 000 themselves rather than trying to act hke Nme Numbers m parentheses are siandard dev1a11on~ The anchor words were I "' strongly duagree, 4 "' ne11her agref' or disagree, and 7 = llrongty agrtt "Shown are the means after the scale was reversed extreme values, Schefff contrasts between the U S group and each of the other three groups were s1gmficant (ps < 001) Japanese. Chtnese, and Canadian students d1d not differ s1gruf-1cantly m their use of extreme values It seems dear that there were cultural differences m the overall response styles Were the differences consistent across all scales and items'> Figure I shows the mean percentages of students choosing the m1dpomt for the vanous scales and items Japanese students were more hkely to use the m1dpomt than the U S and Canadian students across all eight categones Chinese students used m1dpomts more frequently than the two North Amencan groups on all but one scale There was not a consistent difference m the use of the ex· treme scale values between the Japanese. Chinese, and Canadian students The ma1or differences were between those three groups and the U S students For four of the eight categones, the U S students were more likely to use the extreme scale values than the other three groups (sec Fig 2) On no scaJc were the U S students less likely to use the extreme values than the three other groups The data clearly md1cate small but s1gmficant cross-cultural differences m response style The interpretation of these differences may be related to the d1stmct1on often made between collect1v1st soc1et1es, such as the Chinese and Japanese, and mdtv1dual1st cultures, such as the North Amencan Accordmg to such a d1stmct10n, the results 1nd1cate that students from the two types of cultures display d1stmct preferences m their use of scale values Students from the two collect1v1st cultures demonstrate a greater preference for the midpoint and Jess preference for the extreme values than those from the two md1v1du-al1st cultures---espec1ally the U S students To test such a relation w1thm each cultural group, we exammcd the correlattons between students' response styles and the degree of therr endorsement of md1v1dual1sm or collect1v1sm The four items tapping an onentat1on toward md1v1duahsm or collect1v1sm revealed s1gmficant cross-.cullural differences The U S students' ratmgs were s1gmficantly higher on md1v1d· uahsm than the ratmgs of the lwo East Asian groups on each of the items (see Table 2) A combmed score was computed by summing all item means after the collect1v1sm item was reversed A higher mean md1cated a stronger endorsemenl of md1v1dual1sm Table 4 shows correla11ons between students' onentat1on toward md1v1duahsm and their use of scale values As 1s evident, the correlatmns show a clear pattern An md1v1duahsm onen~ tatmn was negatively related to the use of the m1dpomt and pos1t1vely related to the use of ex1reme scale values Ttus was !rue for all lhree groups for the cornbmed score and for mos! of the md1v1dual items As was the case with the cross.cultural ddferences, the magm1udes of the relalions between onentat1on toward md1v1duahsm and response o;tyle were small but statls· t1cally s1g01ficant

Effects of Response Style on Cross-Cultural Differences m Mean Rattngs
The prevalence of cross-cultural differences m response style nuses a cnt1cal question abou11he degree to which cultural response styles alter conclusmns based on cross-cultural compansons of Ltkert+ltke scales This questmn can be answered by companng the magnitude of cross-cultural ddierences m mean VOL 6, NO 3, MAY 1995 ratings before and after the response biases are controlled 1hrough the reducuon of 7-pomt scales to 3-pomt or 2-pomt scales To control for avoidance of extreme vaJues, the ongmal 7-point scales were transformed into 3-pomt scaJes by comb1nmg the extreme vaJue with the two mtermedLate vaJues at each end of the scales (see Bachman & O'Malley, 1984b) In a second analysis, we controlled for the effect of choosing the m1dpomt by further 1ransfonmng the scales mto 2-pomt scales by om1ttmg responses to lhe m1dpomt Previous research has found that such a transforma11on provides data equivalent to what would be obtamed if the midpoint were actually omitted dunng the data collec11on (Schuman & Presser, 1981) Reduction or disappearance of cross-cultural ddferences m averaae ratings after mstitut1ng these controls would pomt 10 the importance of cons1denng response biases m 1nterpret1ng raongs made by persons from dtfferenl cultures Table 5 shows a summary of cross-cultural differences before and afler the transformation of !he scaJes There was hnle change m the cross-cullural differences among the four cultures Only 1 of the 57 items shifted from reprcsentmg nonsigmficant to signlficanl differences after the transformallon into 3·pomt scales That IS, when the four cultures were compared, the magmtude of cross-cultural ddierenccs was not attenuated by the control for response biases In fact, the average eta.squared (an mdex of overall cross.cultural differences) changed ilttle 20 for the ongmal 7-pomt scales, 22 for the 3-pomt scales, and 20 for the 2-pomt scales We interpret this result as md1catmg that although the cultural groups ddfered 1n thei.r response styles, these dlfferences  Table I for the complete names of the scales exerted a very modest mfluence on the degrees of difference found m compani\Ons of the groups' mean ratmgs Such a small magmtude of mfluence 1s due, m part, to the large differences between group means For a maJonty of the compansons, the differences between the means of East Asian and North Amer- tcan student~ were between 0 5 and I 0 on 7-pomt scales, which represented difference~ between about a third and a half of a standard deviat10n When the differences were smaller, such as those between Canadian and U S students, control ofresponsc styles did reduce the number of sigmficant differences Table 5 shows that the largest decline m number of sigmficant differ· ences (9 items) mvo!ved compansons between Canadian and U S students A dechne m the number of s1gmficant differences between the U S and the East Asian groups occurred on no more than 3 of the .' 57 items Therefore, the results for the two types of transformat10n (to 3-pomt and 2-pomt scales) revealed little evidence that differences m response style affected the cross-cultural differences m the ratmgs made by East Asian and U S respondents If anything, the response biases most strongly affected the compansons between the two North Amencan groups

CONCLUSION
The pnmary purpose of this study was to examme possible cross-cultural differences m response styles and their effects on cross-cultural compansons Respondents from four cultures were found to make differential use of certain pomts on the scales Japanese and Chinese students were more hkely than the U S and Canadian students to select rrudpomts, U S students, more frequently than Japanese, Chmese, or Canadians, selected the extreme values The dd'ference m response style between North Amencans and East Asians was in !me with the dtstmct1on often made between md1v1dua1Jst and collecuv1st cultures The latter would be assumed to be more likely to respond on the basis of group no nm and the fonner, on the basis ofmd1v1dual preference These tendencies would generate differences m the use of the m1dpomt and the extreme value~ Furthermore, w1thm each group, a small but significant relation was found between the endorsement of md1vidual1sm and response style Stronger endorsement of md1v1dual1sm was accompanied by more frequent use of extreme values and le~~ frequent use of the midpoint These small but consistent cultural differences m response styles could not account for the large difference~ found m the comparisons of group means After we controlled for response styles, few cross-cultural differences m average ratmgs were changed When changes did occur, they were more frequent m the U S --Canadian compansons than m the Ea~t Asm!l--North Amencan compansons To conclude, the present study confirmed a re!at10n between culture and response style, but offered no evidence for the suggestion that a response style m which extreme value~ are avmded and the midpoint 1s preferred provides a meaningful explanatton for !he cross-cultural differences obtained between the respomes of East Asian and North Amencan students