The evidence - subjective well-being
Acquiescence and extreme responding have also been investigated on constructs related to subjective well-being among a limited number of cultural and linguistic groups. Looking across measures that included psychological distress and locus of control (both loosely related to the construct of eudaimonia or psychological well-being), Ross and Mirowsky (1984) reported that Mexican respondents were more likely to exhibit acquiescent response tendencies, when compared with both Mexican-Americans and non-Hispanic Whites living in El Paso.
It has been hypothesised that certain cultures tend to use response scales differently, and that this could lie behind some of the subjective well-being differences observed between different countries with similar objective life circumstances. For example, Chen, Lee and Stevenson (1995, cited in Schimmack et al., 2002) suggested that low levels of life satisfaction in Japanese cultures may reflect a “modesty bias”.
Tendencies either towards more extreme or more moderate responding could affect either the overall distribution of scores or the mean level of responses. Minkov (2009) explored differences among cultures in the extent to which strong and polarised (i.e. very good versus very bad) judgements tend to be reported across 17 different questions about life quality judgements (e.g. satisfaction with income, family life and job) and social opinions (e.g. views on immigration, use of military force and protection of the environment).25 Country scores for polarisation varied widely, with Middle Eastern Arab societies (such as Kuwait, Palestine territories, Egypt and Jordan) showing the greatest degree of polarisation in judgements, and East and South East Asian societies (such as Indonesia, Japan, China and Korea) the least polarisation.
Minkov’s polarisation measure could reflect a variety of different effects, including differences among countries in the likelihood of using scale extremes, or the genuine diversity of social opinions within countries or greater differences in objective life circumstances within countries. Again, unfortunately, there is nothing in the data per se that enables one to separate these effects out - no attempt is made to demonstrate that the degree of polarisation is related to error. Minkov did, however, find evidence that polarisation was strongly correlated with other national-level dispositional and attitudinal differences - such as “dialecticism” (the tolerance for holding beliefs and feelings that may seem contradictory to a Western mind), the importance of demonstrating high levels of personal consistency, and the extent to which active-assertiveness or compliance are perceived as national traits.
Other studies have also shown how challenging it can be to separate out valid differences on a variable from the presence or absence of response styles. Of course, more moderate responding will produce more moderate mean scores, and more extreme responding will either influence the distribution of results (with a greater number of responses falling towards scale end-points) or potentially draw the mean value of a scale upward or downward where higher or lower scale values tend to be the dominant response within a group.26 Illustrating this difficulty, Hamamura, Heine and Paulhus (2008) reported a clear tendency for North American students of European heritage to report higher self-esteem scores on the Rosenberg Self-Esteem Scale, as well as less “moderate responding” and less “ambivalent responding”,27 than either North Americans of East Asian origin or Japanese students (who had the lowest self-esteem and highest levels of response styles). However, across the whole sample (N = 4 835), there were strong and significant negative correlations between the self-esteem measure and moderate responding (r = -0.41); between the self-esteem measures and ambivalent responding (r = -0.65); and between self-esteem and East Asian ethnicity (r = -0.46).28 The authors thus concluded that, “mean differences in self-esteem are inextricably confounded with these two response styles” (p. 940). This reflects how difficult it is to demonstrate that response styles add error to the data.
In a second study, Hamamura et al. (2008) attempted to isolate the effect of response style by focusing their attention on scale items where there were no overall mean score differences between cultures. This study involved a smaller sample (N = 185) of Canadian student volunteers of either European or East Asian heritage. Analyses were limited to 26 personality scale items, and once again these highlighted a relationship between East Asian ethnicity and heightened levels of moderate responding (r = 0.15, N = 185, p < 0.05) as well as more ambivalent responding (r = 0.18, N = 185, p < 0.05).
Further complicating the picture, Diener et al. (2000) have also found differences between countries in terms of their ideal level of satisfaction with life. They asked a sample of over 7 000 undergraduate students in 41 different societies to indicate the level of overall life satisfaction the “ideal person” would have. The variation was striking, ranging from a mean average score of 19.8 (out of a possible 35) for China to 31.1 for Australia. Notably, those countries that typically score lower on evaluative measures reported relatively lower “ideal” scores (e.g. Japan, 25.8; Korea, 25.0; Hong Kong, China, 25.4), whereas some of those typically scoring higher on evaluative measures reported relatively higher “ideal” scores (e.g. Colombia, 31.0; Puerto Rico, 30.7). These differences between countries in the “ideal” level of life satisfaction could potentially influence the manner in which social desirability affects life evaluation data - with the most socially desirable response varying significantly between cultures.
Techniques based on item response theory have also been developed (e.g. Oishi, 2006; Vittersp, Biswas-Diener and Diener, 2005) to assist in the identification of differences in scale use (i.e. different patterns of number use). For example, Oishi (2006) found that Chinese and US respondents showed different patterns of item functioning on a set of life satisfaction questions, although this differential scale use only partly accounted for the mean differences between these respondent groups. Meanwhile, Vittersp, Biswas-Diener and Diener (2005) found that Greenlanders tended to use more extreme responding than Norwegians on the Satisfaction with Life Scale, but when this tendency was statistically controlled for Norwegians were found to be significantly more satisfied than Greenlanders.
Investigating national differences in patterns of responding can also help to identify subjective well-being questions that may not translate well between cultures, either linguistically or conceptually. For example, Vittersp et al. (2005) found that one item on the Satisfaction With Life Scale (If I could live my life over, I would change almost nothing) was particularly responsible for different response patterns between Norwegian and Greenlander respondents.
Chapter 4 discusses other techniques that have been used to explore the extent of cultural differences in scale use and the potential to adjust for their effects post hoc. This includes the use of counterfactuals (i.e. predicting levels of subjective well-being on the basis of observed determinants and comparing these estimates with the responses actually obtained); vignettes (short descriptions of hypothetical scenarios that respondents are asked to rate, and which may be used to identify differences in how respondents react to the same information); and migrant data (to test whether country-specific effects also apply to migrants in that country). These studies rarely focus on the identification of specific response styles per se (e.g. acquiescence or social desirability), but they seek to identify persistent upward or downward biases in how respondents from different countries self-report their own levels subjective well-being, which is very relevant to both moderate and extreme responding.