What is the role of culture in international comparisons, and can data be "corrected" for "cultural bias"?
The fact that the global distribution of both life satisfaction and affect is wide and varied suggests that differences in country-level life circumstances are likely to produce differences in country-level subjective well-being, and this has been confirmed empirically in a wide variety of studies based on large international datasets (e.g. Helliwell and Barrington-Leigh, 2010; Helliwell et al., 2010; Helliwell, Layard and Sachs, 2012; Deaton, 2010; Boarini et al., 2012; Fleche, Smith and Sorsa, 2011). For example, in the Cantril Ladder life evaluations data set examined in the World Happiness Report (Helliwell et al., 2012), the countries with the top four rankings are reported to have average incomes 40 times higher than those countries with the bottom four rankings.
On the other hand, countries with relatively similar levels of economic development can sometimes report quite different mean levels of subjective well-being.20 Inglehart, Foa, Peterson and Welzel (2008) illustrate how international measures of subjective well-being data can diverge from the pattern that might be predicted based solely on their level of economic development (see also Figure 4.7). This indicates that Latin American countries
Figure 4.7. Subjective well-being (SWB) and per capita gross domestic product (GDP)
Mean life satisfaction1 versus log of GDP per capita - OECD and selected countries, 2010
1. Mean Cantril Ladder score.
Source: OECD (2011), How's Life?, OECD Publishing.
in particular tend to report higher levels of subjective well-being than might be expected based only on their GDP per capita, whilst ex-Communist countries appear to report lower subjective well-being than one might expect.
While income may be an important determinant of country differences in subjective well-being, a number of other non-economic factors are also important, many of which relate to measurable differences in life circumstances (such as health and social context, see Helliwell et al., 2012 and the discussion below). As with interpersonal comparisons of subjective well-being, there may also be some differences between countries in terms of the frames of reference used by individuals to report on their own well-being, as well as differences in how questions are understood and interpreted, and how response formats are used. Chapter 2 described the evidence around hypothesised cultural response styles and the methodological steps that can be taken to reduce the risk of differences in how scales are understood by respondents, and Chapter 3 covered the issue of scale translation in more detail. The purpose of the present discussion is therefore to focus on the interpretation of observed international differences in average levels of subjective well-being, possible sources of those differences, and whether data can and should be “corrected” for linguistic or cultural bias after it has been collected.
Cultural impacts versus cultural “bias”. Before attributing differences in average subjective well-being between countries at similar levels of economic development to “cultural bias”, it is important to remember that these differences may have many sources. A helpful distinction can be made between cultural impact, which refers to valid sources of variance between cultures, and cultural bias, which refers to inter-cultural differences that result from measurement artefacts21 (Van de Vijver and Poortinga, 1997).
If we assume that an identical measurement approach has been adopted across all countries, and therefore that observed differences cannot be attributed to the methodological differences described in Chapter 2, differences in average levels of subjective well-being between countries may have at least four different sources:
• Life circumstances
In addition to income and other economic variables, there may be other important differences between countries in terms of social context and other life circumstances. As noted above, levels of economic development are just one group of potential drivers of subjective well-being, but a very wide variety of others exist, often playing a more substantial role than income (e.g. health, social relationships, unemployment rates, freedom of choice and control). These drivers can include valid country differences in subjective well-being connected to levels of democracy, tolerance of outgroups, strength of religiosity, or trust in others (Inglehart et al., 2008; Bj0rnskov, 2010), and perceived freedom, corruption and the quality of social relationships (Helliwell, 2008; Helliwell et al., 2010; Helliwell et al., 2012). The socio-demographic structure of countries may also contribute to mean differences observed between countries. Because of the very wide range of factors that impact on average levels of subjective well-being, comparing countries on the basis of income alone is insufficient.
• Differences in how people feel about their life circumstances
There may be differences between countries in how people feel about their current life circumstances. Many factors may potentially influence how life circumstances are appraised, including an individual’s reference group (i.e. frame-of-reference effects, discussed just above), past life experiences, the past or present political and economic situation, the policy environment and the country’s religious, cultural and historical roots. These differences may contribute to appraisal styles that influence the connection between objective life circumstances and subjective feelings - for example, the degree of optimism or pessimism individuals feel about the future. Rather than representing cultural “bias”, these should arguably be regarded as valid sources of difference between countries - because they influence the level of subjective well-being actually experienced by individuals, even if this does not mirror exactly the measures of their objective life circumstances.22
• Language differences that influence scale use
Systematic differences between countries may also arise as a result of imperfect translatability of subjective well-being constructs. For example, Veenhoven (2008) has shown differences between the (0-10) numerical ratings that respondents assigned to English and Dutch translations of verbal response categories (very happy, quite happy, not very happy and not at all happy). In this instance, linguistic differences would produce biases in how people respond to a verbally-labelled scale that bear no relation to how individuals actually feel about their lives - and thus it would be desirable to remove this bias from the data.
• Cultural response styles or biases
There may be country-specific differences in how individuals report their feelings, regardless of their actual experiences. For example, a “modesty” or moderate-responding bias might have a downward influence on self-reports, without having a negative impact on private feelings of subjective well-being. Similarly, tendencies towards “extreme responding” (i.e. using scale end-points) or more socially desirable responding could imply differences in modes of cultural expression, rather than substantive differences in the subjective well-being actually experienced. These effects could be described as group differences in scale use, or cultural response styles. If differences in scale use do not represent differences in how people feel about their lives, they can be regarded as a source of bias that it would be desirable to remove.
It is important to distinguish between these four potential sources of country differences, because they have different implications for the validity of between-country comparisons, and for the actions one might take to address country-specific differences in subjective well-being. In the case of unmeasured life circumstances, there is a country-specific effect that may or may not be related to culture. In the second case, differences between countries can reflect cultural impact - i.e. differences in how respondents genuinely feel, and which would add to the predictive validity of the overall subjective well-being measure (e.g. in its association with future behaviour or health states). One would not necessarily want to correct subjective well-being scores for either the first or the second of these country-specific differences. Linguistic differences or cultural response styles, on the other hand, can be expected to add bias to the data, reducing its overall validity and predictive ability. In these instances, it would be desirable to find a way to either minimise the problem at source through survey design (Chapter 2) and translation (Chapter 3), or to adjust the data ex post to remove the bias and enhance the overall usefulness of the measures.