U1LC: Cross Cultural Relations

Course: Cross Cultural Relations    

Unit:  Introduction to Sociology    

Deliverable Length:  see question details

#1 Length: 1 page, double spaced

-Psychology: study of human behavior and mental process 

-How does cultural influence this? How does cultural influence human behavior and mental process?  

#2 Lenght: 3 Paragraphs

– Define equivalence and describe the five bias of equivalence described in textbook.

-See attachment

BIAS AND EQUIVALENCE

In designing and evaluating cross-cultural research, no concepts are more important than equivalenceand bias. 

Bias

 refers to differences that do not have exactly the same meaning within and across cultures. 

Equivalence

 is a state or condition of similarity in conceptual meaning and empirical method between cultures that allows comparisons to be meaningful. These constructs go hand in hand; bias refers to a state of non-equivalence, and equivalence refers to a state of no bias.

In its strictest sense, if there is any bias in any aspect of a cross-cultural comparative study, then the comparison loses its meaning (and may in fact be meaningless). Bias (or lack of equivalence) in a cross-cultural study creates the proverbial situation of comparing apples and oranges. Only if the theoretical framework and hypotheses have equivalent meaning in the cultures being compared—and if the methods of data collection, management, and analysis have equivalent meaning—will the results from that comparison be meaningful. Apples in one culture can be compared only to apples in another.

Thus it’s important for cross-cultural researchers to understand the many aspects of their studies that may be culturally biased and work to establish equivalence in them. Below we discuss five major areas of bias: 

conceptual bias

, method bias, measurement bias, response bias, and interpretational bias.

Conceptual Bias

A major concern of cross-cultural research is the equivalence in meaning of the overall theoretical framework being tested and the specific hypotheses being addressed in the first place. If these are not equivalent across the cultures participating in the study, then the data obtained from them are not comparable because they mean different things. If, however, the theoretical framework and hypotheses are equivalent across the participating cultures, the study may be meaningful and relevant.

For example, people trained to do research in the United States or Europe may be bound by a sense of “logical determinism” and “rationality” that is characteristic of such formal and systematic educational systems. In addition, because we are so used to drawing two-dimensional theories of behavior on paper, that medium affects the way we think about people and psychology. People of other cultures who have not been exposed to such an educational system or who are not used to reducing their thoughts about the world onto a two-dimensional space may not think in the same way. If this is the case, then a real question arises as to whether a theory created within a Western European or American cultural framework is meaningful in the same way to people who do not share that culture. If the theory is not meaningful in the same way, then it is not equivalent.

Method Bias

Sampling Bias

There are two issues with regard to 

sampling bias

, which refers to whether cross-cultural samples can be compared. One concerns whether the samples are appropriate representatives of their culture. Most cross-cultural studies are, in fact, not just cross-cultural; they are cross-city, and more specifically, cross-university studies. A “cross-cultural comparison” between Americans and Mexicans may, for instance, involve data collected in Seattle and Mexico City. Are the participants in Seattle representative of American culture? Would they provide the same responses as participants from Beverly Hills, the Bronx, or Wichita? Would the participants in Mexico City provide the same results as those in San Luis Portosi, Guadalajara, or the Yucatan Peninsula? Of course the answer is “we don’t know,” and it is important for cross-cultural researchers, and consumers of that research (you) to recognize that sound cross-cultural comparisons would entail the collection of data from multiple sites within the same cultural group, either in the same study or across studies, to demonstrate the replicability of a finding across different samples within the same culture.

A second question concerning sampling bias concerns whether the samples are equivalent on noncultural demographic variables, such as age, sex, religion, socioeconomic status, work, and other characteristics. For example, imagine comparing data from a sample of 50 Americans from Los Angeles with 50 individuals from Bombay, India. Clearly, the Americans and the Indians come from entirely different backgrounds—different socioeconomic classes, different educational levels, different social experiences, different forms of technology, different religious backgrounds, and so on.

To deal with this issue, researchers need to find ways of controlling these non-cultural demographic factors when comparing data across cultures. They do this in one of two ways: experimentally controlling them by holding them constant in the selection of participants (e.g., conducting studies in which only females of a certain age can participate in the study in all cultures) or statistically controlling them when analyzing data.

A conceptual problem arises in cross-cultural research in that some noncultural demographic characteristics are inextricably intertwined with culture such that researchers cannot hold them constant across samples in a comparison. For example, there are differences in the meaning and practice of religions across cultures that make them oftentimes inextricably bound to culture. Holding religion constant across cultures does not address the issue because being Catholic in the United States does not mean the same thing as being Catholic in Japan or Malaysia. Randomly sampling without regard to religion will result in samples that differ not only on culture but also on religion (to the extent that one can separate the influences of the two). Thus presumed cultural differences often reflect religious differences across samples as well. The same is often true for socioeconomic status (SES), as there are vast differences in SES across cultural samples from around the world.

Linguistic Bias

One arena in which potential bias in cross-cultural research becomes quickly apparent is in language. Cross-cultural research is unique because it often involves collecting data in multiple languages, and researchers need to establish the linguistic equivalence of the research protocols. 

Linguistic bias

 refers to whether the research protocols—items on questionnaires, instructions, etc.—used in a cross-cultural study are semantically equivalent across the various languages included in the study.

There are generally two procedures used to establish linguistic equivalence. One is known as 

back translation

 (Brislin, 

1970

). Back translation involves taking the research protocol in one language, translating it to the other language, and having someone else translate it back to the original. If the back-translated version is the same as the original, they are generally considered equivalent. If it is not, the procedure is repeated until the back-translated version is the same as the original. The concept underlying this procedure is that the end product must be a semantic equivalent to the original language. The original language is 

decentered

 through this process (Brislin, 
1970
, 

1993

), with any culture-specific concepts of the original language eliminated or translated equivalently into the target language. That is, culture-specific meanings and connotations are gradually eliminated from the research protocols so that what remains is something that is the closest semantic equivalent in each language. Because they are linguistic equivalents, successfully back-translated protocols are comparable in cross-cultural hypothesis-testing research.

A second approach to establishing language equivalence is the committee approach, in which several bilingual informants collectively translate a research protocol into a target language. They debate the various forms, words, and phrases that can be used in the target language, comparing them with their understanding of the language of the original protocol. The product of this process reflects a translation that is the shared consensus of a linguistically equivalent protocol across languages and cultures.

Researchers may combine the two approaches. A protocol may be initially translated and back-translated. Then the translation and back-translation can be used as an initial platform from which a translation committee works on the protocol, modifying the translation in ways they deem most appropriate, using the back-translation as a guideline.

Procedural Bias

The issue of bias and equivalence also applies to the procedures used to collect data in different cultures. For instance in many universities across the United States, students enrolled in introductory psychology classes are strongly encouraged to participate as research subjects in partial fulfillment of class requirements. American students generally expect to participate in research as part of their academic experience, and many American students are “research-wise.”

Customs differ in other countries. In some countries, professors simply collect data from their students or require them to participate at a research laboratory. In some countries, students may consider it a privilege rather than a chore or course requirement to participate in an international study. Thus, expectations about and experience with research participation may differ.

All the decisions researchers make in any other type of study are made in cross-cultural studies as well. But those decisions can mean different things in different countries. Laboratory or field, day or night, questionnaire or observation—all these decisions may have different meanings in different cultures. Cross-culturalresearchers need to confront these differences in their work and establish procedures, environments, and settings that are equivalent across the cultures being compared. By the same token, consumers need to be aware of these possible differences when evaluating cross-cultural research.

Measurement Bias

Perhaps the most important arena with regard to bias and equivalence may concern the issue of measurement. 

Measurement bias

 refers to the degree to which measures used to collect data in different cultures are equally valid and reliable. As mentioned above, validity refers to whether a measure accurately measures what it is supposed to measure; reliability refers to how consistently a measure measures what it is supposed to measure.

To be sure, one of the most important lessons to learn about cross-cultural research methods is that linguistic equivalence alone does not guarantee measurement equivalence. This is because even if the words being used in the two languages are the same, there is no guarantee that those words have exactly the same meanings, with the same nuances, in the two cultures. A successful translation gives the researcher protocols that are the closest linguistic equivalents in two or more languages. However, they still may not be exactly the same. In translating the English word anger, for example, we might indeed find an equivalent word in Cantonese or Spanish. But would it have the same connotations, strength, and interpretation in those languages as it does in English? It is very difficult to find exact translation equivalents of most words. Thus, cross-cultural researchers need to be concerned with measurement equivalence in addition to linguistic equivalence.

One way to think about measurement equivalence is on the conceptual level. Different cultures may conceptually define a construct differently and/or measure it differently. Just because something has the same name in two or more cultures does not mean that it has the same meaning (Wittgenstein, 1953/1968, cited in Poortinga, 

1989

) or that it can be measured in the same way. If a concept means different things to people of different cultures, or if it is measured in different ways in different cultures, then comparisons are meaningless. Cross-cultural researchers need to be keenly aware of the issue of equivalence with regard to their conceptual definitions and empirical 

operationalization

 of the variables (the way researchers conceptually define a variable and measure it) in their study.

Past debates concerning cross-cultural studies of intelligence highlight issues concerning conceptual equivalence. Intelligence typically is thought to consist of verbal and analytical critical-thinking skills and tests such as the Wechsler Adult Intelligence Scale (WAIS) have been widely used to assess IQ. Different cultures, however, may have a different conception of what constitutes intelligence. For example, a culture may consider nobility of character and sincerity to be markers of intelligence. Another culture may consider the ability to have smooth, conflict-free interpersonal relationships a marker for intelligence. Yet another culture may consider creativity and artistic abilities to be indices of intelligence. Comparisons of WAIS data from all of these cultures may not be a meaningful cross-cultural comparison of intelligence.

Another way to think about measurement equivalence is on the statistical level—that is, in terms of 

psychometric equivalence

. Psychometric equivalence can be ascertained in several different ways. One of the most important ways, especially when using questionnaires to collect data (which is used in many cross-cultural studies), is to determine whether the questionnaires in the different languages have the same structure. For example, researchers often use a technique called 

factor analysis

 to examine the structure of a questionnaire. Factor analysis creates groups of the items on a questionnaire based on how the responses to them are related to each other. The groups, called factors, are thought to represent different mental constructs in the minds of the participants responding to the items. Scores are then computed to represent each of these mental constructs.

When using questionnaires across cultures, one concern that arises is whether the same groups of items, or factors, would emerge in the different cultures. If so, then the measure is said to have 

structural equivalence

. If not, however, the measure is structurally nonequivalent (biased), which suggests that people of different cultural groups have different mental constructs operating when responding to the same questionnaire. Thus, their responses may not be comparable to each other.

Another way in which psychometric equivalence can be ascertained is by examining the 

internal reliability

 of the measures across cultures. Internal reliability can be assessed by examining whether the items on a questionnaire are all related to each other. If they are supposed to be measuring the same mental construct, then items should be related to each other; that is, they should have high internal reliability. If the items are working in the same way across cultures, then they should have high internal reliability in each of the cultures being tested.

Response Bias

In addition to the methodological issues concerning bias and equivalence described above, cross-cultural researchers need to be aware of the fact that different cultures can promote different types of response biases. A 

response bias

 is a systematic tendency to respond in a certain way to items or scales. If response biases exist, then it is very difficult to compare data between cultures because it is not clear whether differences refer to “true” differences in what is being measured or are merely differences in how people respond using scales.

There are, in fact, several different types of response biases. 

Socially desirable responding

, for instance, is the tendency to give answers that make oneself look good (Paulhaus, 

1984

), and it may be that people of certain cultures have greater concerns that lead them to respond in socially desirable ways than people of other cultures. There are two facets of socially desirable responding, which include self-deceptive enhancement—seeing oneself in a positive light—and impression management. Lalwani, Shavitt, and Johnson (

2006

) demonstrated that European American university students score higher on self-deceptive enhancement than both Korean Americans and students from Singapore, but the latter score higher on impression management than do European Americans (

Figure 2.2

).

Figure 2.2 Socially Desirable Responding

(a) Comparison of European Americans and Singaporeans, (b) Comparison of European Americans and Korean Americans. SDE = self-deceptive enhancement; IM = impression management.

(Source: Adapted from Lalwani et al., 
2006
.)

Lalwani et al. (
2006
) also demonstrated that individuals with more individualistic cultural orientations engaged in more self-deceptive enhancement, while individuals with more collectivistic orientations engaged in more impression management. In a related vein, Matsumoto (

2006b

) showed that differences between Americans and Japanese university students’ individualistic versus collectivistic cultural orientations disappeared once socially desirable responding was statistically controlled.

Two other types of response bias are 

acquiescence bias

, which is the tendency to agree rather than disagree with items on questionnaires, and 

extreme response bias

, which is the tendency to use the ends of a scale regardless of item content. Van Herk, Poortinga, and Verhallen (

2004

) examined responses on marketing surveys regarding household behaviors (e.g., cooking, use of products, shaving, washing clothes) in six European countries. They reported that countries near the Mediterranean (Greece, Italy, and Spain) exhibited more of both acquiescence bias and extreme response bias than countries in northwestern Europe (France, Germany, and the United Kingdom). Interestingly, their degree of the two response biases were not correlated with national differences in actual behaviors with regard to the issues raised. (If there were differences in rates of actual behaviors, it could be argued that the response styles were not biases, but were reflective of actual differences in behaviors, but this was not the case.)

A final type of response bias is the 

reference group effect

 (Heine, Lehman, Peng, & Greenholz, 

2002

). This idea is based on the notion that people make implicit social comparisons with others when making ratings on scales, rather than relying on direct inferences about a private, personal value system (Peng, Nisbett, & Wong, 

1997

). That is, when completing rating scales, people will implicitly compare themselves to others in their group. For example, Japanese individuals may appear to be fairly individualistic on questionnaires, even more so than Americans. But Heine et al. (
2002
) argue that this may be because the Japanese implicitly compare themselves to their own groups, who are actually fairly collectivistic, when making such ratings, and thus inflate their ratings of individualism. Likewise, Americans may inflate their ratings of collectivism because they implicitly compare themselves to others, who are actually fairly individualistic. Peng et al. (
1997
) examined four different value survey methods: the traditional ranking, rating, attitude scaling procedures, and a behavioral scenario rating method. The only method that yielded reasonable validity estimates was the behavioral scenario rating method, the most uncommon of all the measures tested.

What aspects of culture account for response biases? Johnson, Kulesa, Cho, and Shavitt (
2004
) examined these biases in 19 countries around the world and correlated indices of the biases with each country’s score on Hofstede’s cultural dimensions (see 

Chapter 1

 for a review). (This study is an example of an ecological-level study.) On one hand, extreme response bias occurred more in cultures that encourage masculinity, power, and status. They suggested that this response style achieves clarity, precision, and decisiveness in one’s explicit verbal statements, characteristics that are valued in these cultures. On the other hand, respondents from individualistic cultures were less likely to engage in acquiescence bias, probably because maintaining harmony and conveying agreeableness and deference are less emphasized in these cultures.

In the past, response biases were viewed as methodological artifacts that need to be controlled in order to get to “true” responses. Today, however, there is a growing view of them as an important part of cultural influence on data. Regardless of how researchers choose to view this issue, their effects should be acknowledged and incorporated in data analysis in cross-cultural comparisons.

Interpretational Bias

Analyzing Data

In testing cultural differences on target variables of interest, researchers often use inferential statistics such as chi-square or analysis of variance (ANOVA) and engage in what is known as null hypothesis significance testing. These statistics compare the differences observed between the groups to the differences one would normally expect on the basis of chance alone and then compute the probability that the results would have been obtained solely by chance. If the probability of obtaining the findings they did is very low (less than five percent), then researchers infer that the findings did not occur because of chance—that is, that the findings reflect actual differences between the cultural groups from which their samples were drawn. This “proof by negation of the opposite” is at the heart of the logic underlying hypothesis testing and statistical inference.

In the past, researchers were quick to take “statistically significant” results and interpret them as if they were “practically meaningful to all or most members of the groups being compared.” That is, researchers (and consumers of research) often assume that most people of those groups differ in ways corresponding to the mean values. Thus, if a statistically significant difference is found between Americans and Japanese, for instance, on emotional expressivity such that Americans had statistically significantly higher scores than the Japanese, people often conclude that all Americans are more expressive than all Japanese.

But the fact that the differences between group means are statistically significant does not by itself give an indication of the degree of practical meaningfulness of the difference between the groups. Group means may be statistically different even though there is considerable overlap among the scores of individuals comprising the two groups. The tendency to make glossy, broad-sweeping statements based on “statistically significant” results is a mistake in interpretation that is fueled by the field’s fascination and single-minded concern with statistical significance and perhaps stereotypes.

Fortunately, statistical procedures are available that help to determine the degree to which differences in mean values reflect meaningful differences among individuals. The general class of statistics is called “effect size statistics”; when used in a cross-cultural setting, Matsumoto and his colleagues call them “cultural effect size statistics” (Matsumoto, Grissom, & Dinnel, 

2001

). There are a number of different types of such statistics that can help researchers and readers get an idea of the degree to which the between-group cultural differences actually reflect differences among the individuals tested, helping to break the hold of stereotypic interpretations based on group difference findings.