Are you a researcher getting vastly different responses on client satisfaction scores? Take a look at this deck I put together for a dilemma we were attacking for our larger international audiences.
2. Our Questions
Japan - Are they always going to say “yes”?
Brazil - Are Brazilians just happy?
France - Is anything good enough for the French?
Germany - Are the Germans just not fun?
Why are your survey responses for Germany & France so
different from Brazil & Japan? Is your survey adapted for
culture?
3. How are they responding
currently?
Japan (-)
Brazil (+)
France (negative)
Germany (negative)
4. Maybe they have one or more
of these biases
Acquiescence Response Style (ARS)
Where is it a problem?
Agree/Disagree questions
Middle Response Style (MRS)
Extreme Response Style (ERS)
Response style bias: is there a tendency to respond to
survey questions in a particular way, regardless of
content?
5. ARS
Acquiescence Response Style (ARS)
Where is it Prevalent?
Cultures where being polite is important
Can we pinpoint the issue?
Agree/Disagree questions are tough
Some cultures interpret these questions as a double
negatives (agreeing with something negative is
doubly negative)
They may not fully disagree or agree
6. MRS
Middle Response Style (MRS)
Where are MRS prevalent?
Cultures where extremes are discouraged: Asian Cultures
Can we pinpoint the issue?
Confucious: No one answer is absolute
Collectivism: Avoid making others look bad
7. ERS
Extreme Response Styles
Where are ERS prevalent?
In cultures that value individualism, are masculine, and have a low
tolerance for ambiguity
Can we pinpoint the issue?
Individualistic, masculine cultures put a value on assertiveness and
decisiveness
8. Geert wants to remind us…
Power Distance (PDI)
Individualism (IDV)
Uncertainty Avoidance (UA)
Masculinity (MAS)
Long Term Orientation (LTO)
Response biases are related to all of these orientations
9. Brazil: High Acquiescence
Highest acquiescence cultures…
and may have a bit of SDR
Why?
High UA & Low IDV
Collectivism (low IDV)
Expected behaviors
More collectivist than Japan!
Aversion to confrontation
Politeness
10. Japan is a Middle Response
Culture
Why?
Collectivism
Personal opinions are not
strongly voiced
Value harmony
Dialectical Thinkers
No one, absolute truth
Long Term Orientation
LTO societies believe that
truth depends on the
situation
11. Germans are well balanced,
but can be a little negative
Why?
DRS or “disacquiescence” is associated
with low power distance and high
individualism. Germany is be a prime
example of this type of culture.
High masculinity and UA are also
associated negatively to acquiescence.
PDI: 35 IDV: 67 MAS: 66 UA: 65
12. The French are Extremists!
The French are the 3rd most extreme culture in Europe.
Why?
Rules (UA)
Societies higher in UA have many rules and have little tolerance
for ambiguity.
Direct Communication (IDV)
Individualistic societies are characterized by self-
expression, independence, and personal opinions.
Individual contributions expected, speaking up appreciated
– disagreement is normal.
Individualistic societies have absolute truths
UA: 86 IDV: 71
"You Can't Judge a Book by its Cover", Arun Joshi, Sagar Tamang, Himanshu Vashishtha, Research World, January 2009
International Marketing Research By C. Samuel Craig, Susan P. Douglas
14. How to balance acquiescent
cultures
What’s the Solution?
Give various levels of agreement and disagreement
Reverse order points of scale
Employ negatively worded statements vs positive
Or don’t use obvious highs/lows or negatives/positives
Use visual stimuli/pictures vs words or numbers
Test for acquiescence using a behavioral variable
Joshi-Tamang-Vashishtha Scale
Acquiescence is closely related to Social Desireability –
lower class people will give responses that reflects the
values of the interviewer in face to face interviews
15. Solutions for MRS Cultures
What’s the Solution?
Kill the middle response!
Use a multi-item scale
18. Survey Type
Not all surveys are transferable
Scale development is important
Measurement scales need to be investigated
19. Item Generation
Does the survey rely on commonalities or distinctions?
Emic or Etic?
Applicable universal concepts, values, truths
Frame of reference - not just language - should be
conceptually equivalent
Suggestion
Qualitative data (interviews) can be helpful, especially if taking
surveys from multiple countries at once
20. The Role of Language
English language by default elicits middle responses
Opposites in one language may not be opposite in another
language
Double negatives!
Intensity levels or words could be misinterpreted
Direct translation isn’t always the key
Scale anchors may not translate well
Harzing, 2006
21. The Role of Language
Language impacts surveys!
Cultural accommodation
Symbolic effect
Concepts and terms
Interpretation – culture specific connotations
Researcher needs cultural knowledge as well as
language knowledge
Use clear, simple, and familiar wording
Use natives in country where surveying
Careful translation + pilot testing with native speakers
22. Other Solutions
Mixture of Positive & Negative Statements
Mitigates acquiescence and disaquiescence
Linkert scales, use more than 5!
Reduces response and language bias
Scale anchors that have mutually exclusive constructs rather than
agreement level also reduce biases*
Do not compare loyalty scores across countries, instead
compare loyalty scores across companies in the same
country!
Less complex than statistical adjustments and probably more
accurate
23. NOT Good Solutions
Some organizations try to eliminate the influence of
response bias by calculating a cross-cultural adjustment
factor for each country or region in which they do business.
The adjustment factor is calculated as the ratio of the
average score for a comparison country over the average
score for a base country, which is typically the organization’s
largest, most important market.
In practice, however, this method is ill-advised unless the
adjustment procedure truly isolates opinion differences from
differences due to cultural response bias
24. Remember
The researcher needs to take culture into consideration in
the overall design, language, and item generation of the
survey.
Benchmarks should be within the same country
More points on the Linkert scale
Cross-country comparison helps to identify survey
issues
Native/local speakers can help with design
Editor's Notes
Response Styles
(ARS) acquiescence
The tendency to agree with an item regardless of the content
(DRS) disacquiescence
The tendency to disagree with an item regardless of the content
(ERS) Extreme Response Styles
The tendency to use the extreme response categories on ratings scales
(MRS) Middle Response Styles
The tendency to use middle response categories on ratings scales
(AMBRS) – Ambivalent Response Styles
(SD) – Social Desirability
Thurstone scaling?
Japan and Brazil are collectivist societies, making the group more
important than the individual. These societies tend to not speak directly
nor strongly voice personal opinions. Individualistic societies (USA,
Germany, France) put an emphasis on the individual and encourage
voicing and arguing opinions.
Underlying reasons for acquiescence bias1. Collectivity
One critical factor for acquiescence bias within a society is its level of collectivity.
In cultures that are marked by a high degree of collectivism, individuals feel it as
important to consider what people might think and say about their actions, and
this creates a bias. According to Hofstede, collectivist countries tend to be characterized by a preference
for harmony and an avoidance of confrontations, leading to more conformity in
behavior where opinions are predetermined by the ingroup . This tendency is reinforced by the fact that most collectivist countries are characterized by an indirect communication style, where the expression of strong opinions is avoided. Confucian teachings (predominant in East Asian countries) reinforce this communicationstyle by advising followers to keep themselves from extremes (
2. Politeness
Another factor is politeness, the level and nature of hospitality displayed in a
society can have an impact on acquiescence bias. If a face-to-face interview
takes place at the respondent’s place of residence, the respondent will treat the
interviewer as a guest and exhibit behavior that befits a good host.
3. Positivity
In a culture that preaches being positive, ‘negativity avoidance’ tends to create
acquiescence bias meaning that whilst a respondent might well have a negative
feeling or reaction, they would tend to avoid giving a negative response. Other
factors such as economic prosperity, consumer confidence and a high degree of
materialistic indulgence tend to add to acquiescence bias (Joshi, 2009).
It is important to note that all of the countries being compared are among
the top 10 economic performers in terms of GDP & PPP
(CIA World Factbook, 2012).
Middle Response Style (MRS): Japan
The reasoning behind Japan’s MRS – and slightly acquiescent style – is due to at least three major factors. These three major factors in Japanese and East Asian cultures are influenced by Confucian teachings.
Japan is collectivist (low IDV)
The collectivist mindset may cause Japanese participants to choose the middle response to avoid any upset. Confucian teachings reinforce this communication style by advising followers to keep themselves from extremes (Hertzing, 2012).
Japan is a country of dialectical thinkers: No one, absolute truth
East Asian countries are characterized by dialectical thinking - or arguing against oneself, seeing multiple solutions.
Western notion of ‘truth’ has different connotations in Confucian
Philosophy (Harzing, 2012).
Japanese are long-term oriented (LTO)
In societies with a long-term orientation, people believe that truth depends very much on situation, context and time (Hofstede).
Dialectical thinking also falls in line with Confucian teachings –
Dialectical thinking: Cultures differ indialectical thinking, that is, the tolerance for holding apparently contradictory beliefs (Peng & Nisbett, 1999). In particular, dialectical thinking appears to be
more common among East Asians than North Americans. For example, Peng and Nisbett (1999)
found that Chinese often endorsed both of two arguments that Americans viewed as incompatible.
Dialectical thinking is also evident in the way that East Asians think of themselves. For example, Choi and Choi (2002) showed that, compared to Americans, Koreans displayed self-beliefs
that were less consistent (e.g., I’m outgoing but somewhat shy). Spencer-Rodgers, Peng, Wang,
and Hou (2004) found that, compared to Americans, Chinese self-evaluations are more ambivalent in the sense of incorporating both positive and negative self-evaluations.
Joshi-Tamang-Vashishtha Scale
Survey Type
Scales
Many scales are available, but may not be transferable to a different cultural context. The use of multinational research teams whose members are familiar with the respective local cultures may help to overcome problems related to adapting measurement scales. Key to a meaningful modification of existing measurement scales is a sound process of scale development.
Item Generation
Construct Equivalence
Choose whether or not your survey construct relies upon the commonalities between the benchmark culture and the country to be studied (etic) or the intrinsic distinctions between the two cultures (emic).
Emic/Etic: A construct can only be meaningfully measured across cultures if it is based on universally applicable concepts in these cultures, that is, the construct is conceptually equivalent. In this regard, whereas questionnaire translation is necessary to ‘clarify’ construct elements in the local language and frame of reference it is an insufficient condition for establishing construct equivalence (Harzing, 2012).
Triangulation: The feasibility of modifying existing scales to accommodate for cultural specificities is therefore limited if data from a larger amount of countries are to be compared. In that case, the collection of qualitative data, for example through interviews, can compensate for the inherent limitations of survey data. Triangulation can thus increase the robustness of the data gathered.
English vs Native Language
English-language questionnaires are shown to elicit a higher level of middle responses, while questionnaires in a respondent’s native language result in more extreme response styles. Finally, English language competence is positively related to extreme response
styles and negative related to middle response styles. An important explanation for response style differences across languages might be differential interpretation of equivalent scale anchors in different languages. Even though scale anchors
might translate into appropriate local equivalents, the intensity associated with these equivalents
might be different from the original language. (Harzing, 2006).
Most studies that show response bias used Likert scales with ordered scale anchors, e.g.
“strongly disagree” to “strongly agree” or “of little or no importance” to “extremely important”.
These anchors might be vulnerable to acquiescence bias as respondents are keen to agree,
whether this is caused by high levels of power distance, collectivism, uncertainty avoidance or
extraversion. Our detailed analysis showed that scale anchors referring to the level of importance are even more problematic in this respect than scale anchors referring to the level of agreement.
A related problem in this respect – discussed above with special reference to Japan – is that scale anchors are often difficult to translate and that translations might not result in metric equivalence. An alternative would be to use scale anchors as part of the question and let them reflect opposites rather than level of agreement. This would make the “right answer” less obvious and would also force respondents to carefully consider each question as most scale anchors would be different. Of course this technique would increase the level of cognitive involvement required and might lead to lower response rates.
In addition, careful translation and pilot testing would become even more crucial as a respondent’s interpretation of the questions would be framed by single words, whereby words that are seen as opposites in some countries might not be opposites in other countries. However, if translation problems can be solved, responses might be more
meaningful (Harzing, 2006).
Survey Languages
If distributing a survey in English, it is important to consider language proficiencies. In some cases, English may in fact be adequate. This would be the case for very specific populations.
Research shows that English-language questionnaires may create a bias
Translation is crucial
Survey translation & interpretation is crucial.
Language of the questionnaire can impact attitudes and behaviors (cultural accommodation).
The translation and interpretation into a local language – even when the respondent has a high level of English - has a symbolic effect on the participant.
The fact that the researcher took the time and effort to make the survey as easy as possible for the participant has an impact.
Concepts and terms differ across cultures.
Asian countries regard feedback as personal, indirect, and informal procedure.
The US & UK regard feedback as an open, direct, and formalized procedure (Hofstede, 1998)
Interpretation is necessary as many concepts and terms entail culture-specific connotations, their mere direct translation is unlikely to transport the intended meaning
A meaningful translation of the original version of the questionnaire requires a researcher not only to ensure overall conceptual equivalence
but also to consider vocabulary, idiomatic and syntactical equivalence
Suggestions: It is suggested to use simple sentence structures as well as clear and familiar wording as much as possible to facilitate translation. In addition, by adding redundancy and necessary context for difficult phrases, the researcher is able to clarify the intended meaning (Harzing, 2012).
Pilot testing is particularly important in international research. One way to achieve this is to use a committee to review the survey to see if the interpretations of the survey are the same.
In terms of the choice of survey type, it is advisable to obtain help from native speakers in
designing the survey, ask locals to check and test the survey, and combine both paper-andpencil and online surveys to increase response rates. To generate survey items, it is
important to fist decide whether the research project is emic or etic in nature. Whereasthe
former case requires the adaptation of existing and development of new items, in the latter
case this may be unfeasible when interested in cross-cultural comparability. Instead,
researchers should collect additional qualitative data to triangulate the study results. Finally,
several translation techniques exist to adapt the survey to the local language of the
countries it is diffused. Simple sentence structures should be used to facilitate translation,
and additional clarifications provided to better convey the intended meaning of certain
questions. It is also important to pilot test the survey in each country (Harzing, 2012).
Other solutions all relate to initial questionnaire design. A use of a mixture of positive
and negative statements will mitigate both acquiescence and disacquiescence. Likert scales
with a larger number of scale points and the use of ranking have also been shown to reduce 31 | Page
both response and language bias (Harzing et al., 2009) as have scale anchors that refer to
mutually exclusive constructs, rather than to level of agreement (Harzing et al., 2012).
These are not solutions:
For example, researchers will typically assess customer opinions across a minimum of three areas unrelated to the area of actual interest, by adding additional questions to the survey. Surveys are fielded to an international panel of consumers and the results are used to construct structural equations or constrained factor analyses which help to isolate the cultural response bias from differences in true opinion (Brunso & Guerent,1995; Weijters, et. al., 2004; King, et.
al., 2004; Keillor et. al., 2001).
Some organizations try to eliminate the influence of response bias by calculating a cross-cultural
adjustment factor for each country or region in which they do business. The adjustment factor is calculated as the ratio of the average score for a comparison country over the average score for a base country, which is typically the organization’s largest, most important market.
In practice, however, this method is ill-advised unless the adjustment procedure truly
isolates opinion differences from differences due to cultural response bias. If this is not
the case (and short of the statistical modeling described above, it is difficult to do so), the
adjustment procedure is likely to eliminate true differences in opinion across countries.
the exact information the customer survey is intended to measure. Further, this approach
does not permit accurate statistical tests of observed differences across countries,
making it difficult to assess what constitutes a truly .significant. discrepancy in scores
(Scholderer et. al., 2005).