Methodological Issues in Empirical Cross-Cultural ...

Methodological Issues in Empirical Cross-Cultural Research: A Survey of the Management Literature and a Framework Author(s): S. Tamer Cavusgil and Ajay Das Source: MIR: Management International Review, Vol. 37, No. 1 (1st Quarter, 1997), pp. 7196 Published by: Springer Stable URL: http://www.jstor.org/stable/40228388 Accessed: 10-06-2016 13:08 UTC Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://about.jstor.org/terms

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Springer is collaborating with JSTOR to digitize, preserve and extend access to MIR: Management International Review

This content downloaded from 131.96.47.145 on Fri, 10 Jun 2016 13:08:05 UTC All use subject to http://about.jstor.org/terms

mir vol.37, 1997/1, pp. 71-96 #

mir

International Review © Gabler Verlag 1997

S. Tamer Cavusgil/Ajay Das

Methodological Issues in Empirical Cross-cultural Research: A Survey of the Management Literature and a Framework

Abstract ■ Comparative empirical research methodology has been the subject of analysis and study for the past three decades. Several problems and issues however still remain largely intractable or often ignored. ■ This article reviews and classifies key studies and concerns in the field, offers appropriate remedies, and develops a detailed process framework for creating sound cross-cultural research methodology designs.

Key Results ■ Relevant problems in comparative research design, sampling, instrumentation, and data collection and analysis are identified and addressed through a prescriptive framework.

Authors Professor S. Tamer Cavusgil, The John William Byington Endowed Chair of Global Marketing; Executive Director, International Business Center, Michigan State University, East Lansing, MI,

U.S.A.

Ajay Das, Doctoral candidate in Operations and Sourcing Management, Department of Management, Eli Broad College of Business, Michigan State University, East Lansing, MI, U.S.A. Manuscript received August 1995, revised January 1996, revised May 1996.

mir

vol.

37,

1997/1


71


Methodology issues in cross-cultural empirical research have been the subject of inquiry for many scholars across several disciplines over the past two decades (Samiee and Jeong 1994, Nasifetal. 1991, Cheng 1989, Parameswaran and Yaprak 1987, Adler 1983, Winter and Prohaska 1983, Sekaran 1983, Douglas and Craig 1983, Green and White 1976). These studies comprise both analytical as well as the pause and consolidate type of investigation. While studies such as Nasif et al. (1991) focus more on the latter, scholars like Cheng (1989) have introduced additional conceptual dimensions to the domain. Our study seeks to add value to the body of existing knowledge through an intensive state-of-the-art literature review,

a comprehensive examination of potential statistical pitfalls and available analytical tools. The overall aim is to formulate a conceptual framework for guiding cross-cultural research methodology. More specifically, this article's objectives are two-fold:

1 . To construct a compendium of generic methodological problems and recommended remedies in comparative business research, and 2. To develop a methodology process model of prescriptive character for conducting cross-cultural research (Exhibit 1). We begin by identifying key studies in the area of cross-cultural business research methodology over the past two decades. Study selection criteria include source journals, cite frequency, and expert opinion as represented in syllabi for doctoral international business methodology courses. We then classify and analyse major methodological concerns in cross-cultural research. In doing so, we generate a list of generic problem areas in cross-cultural research and offer suggestions for their resolution. The article concludes with the development of a process model of cross-cultural methodology design.

Key Issues in Cross-cultural Research Methodology Sekaran (1983) delineates methodological issues in comparative research into five broad heads: functional equivalence, instrumentation, data-collection, sampling design, and data analysis. Adler (1983) lends support to the above framework, adding the elements of criterion definition and research administration to the clas-

sification structure. In their update on the same topic, Nasif et al. (1991) introduce the aspects of methodological simplicity and level of analysis. As with any classification framework, individual issues and factors are combined into broader composites based on (a) inter-relatedness and (b) considerations of parsimony. We follow the same logic in structuring our methodology classification framework into four broad heads: Basic Research Design, Sampling

72

mlr

vol.

37,

1997/1


Methodology in Comparative Research

s * ll" A i I Si

is i i p ipii

ii U I ilii l

Ii l|g i if e

S3 la» g Si*g I § I I N ip p ^|is i S 2 i §e« g «-pK i § £ & BeS gea I g Mii 8 S a £ i

III III J I 111! 1 1 8 I

. i s isii s«b 3-i is in i I n ib5j giii %u gi i| i « h 2 5 O J

I A 8 >h I 2 § 5 1 I

1 . 8,i 1 I I 1

i I i . 1 b

■ pr-'-a i s I < n |

: s i 1 1 s s s i I

4

w

mir

vol.

37,

^

1997/1


73


Issues, Instrumentation & Data Collection, and Data Analysis. Within each category are included a variety of pertinent sub-issues. It is important to remember here that while successful cross-cultural research design may follow a sequential pattern in terms of first developing solid theoretical constructs followed by sampling, instrumentation, data collection and data analysis, some aspects of research design may be more important than others. An otherwise well designed research program may fail just because of ill-planned instrumentation or ill-executed data collection. But it will certainly be easier to recover from lapses in say, data analysis, than from failures caused by specification error or unreliable instrumentation. However, each step in the research plan integrates with the other, and the whole is obviously dependent on the parts. This theme is expanded later in the paper in developing a generic process model for conducting cross-cultural research (Exhibit 1) which presents an integrated overview of the research plan. We next review the major research studies in the area of cross-cultural methodology and highlight their salient features. Key contributions to the literature are

profiled in Table 1. We adopt Cox's (1974) typology to explain the first three columns of the table. Data analysis forms a separate topic for discussion. Cox (1974) characterises potential sources of error in research projects into (a) errors of definition, (b) errors of estimation and (c) errors of explanation. This classification framework is presented in Table 2.

Errors of Definition

Mis-statement of the problem or mis-statement of relevant variables create specification error. The relative complexity of cross-cultural research renders it especially susceptible to such pitfalls. Several prescriptions have been advanced in

this regard. Verba (1967, 1971) suggests a two stage approach to obtain meaningful tests of complex phenomena. He suggests to first look for the relationship between dependent and independent variables within each single nation and then compare this relationship between countries. Cheng (1989) focuses on the organisation as the level of analysis and suggest a logical contextual approach to crosscultural research. He recommends the initial selection of dependent (organisational) variables followed by the selection of theory linked independent (societal) variables. Cultures are selected last based on their relative standing on each of the independent variables. Results are interpreted as a function of the statistical rela-

tionship between the societal level independent and organisational level dependent variables. The translation of abstract and fuzzy cultural differences into different levels of more specific theory-derived societal variables is in line with Adler's (1984) admonition to select cultures on a theoretical basis. The active consideration of theory in research design reduces the changes of making gratuitous specification errors in the basic constructs. In addition, the frequently discussed

74

mir

vol.

37,

1997/1



C/3

a)

c«

oj

*?

tj

2

^

1 a) tl= c« oj *? I l tj **! ^ 8 5 « .a 5 S r I u .^ 5 | S '§ QuiSuO

D

D

§CXZ

&

£

s§ g^si « 25.^ 'sis.*-! & si

1 1 -c 1 ™ 1 « S >. 2 §■ 1 5-1 8 > £ I 3 | g •■§

ii Him ii i in in tf hi 1

,£

1

I

1

I -11 » 1 1 ii f I

I i |l I II I 11 I ! I I III I II 1 III! I 1 s

«»

& a

"5

•§

I §

>

1 a Is 1 n I 8| *oc3§E

2

j-

A

«

g

m

.

1 I II 1 sill II IS

2

1 i 1 18! !i_ Ii f|

3.1? i i Hi 1 ill

III I I !1I I « li SI! mir

vol.

37,

1997/1


75


C/5

>

a>

.

c

°

«

°

L jt i r*i * ifsfii 1*1

jj s Hi it mum

c o

! Si ilillilt tfifl 111! J -

a.

«

c

P

o

« |« 1* £* .8. sf-i'l-

I f 111 II! !* i i I si 3i? ||i ii ! "g s i

u

IS!

If

sill

i

il

|| 1 g I Hi? I I 8 s ex)

c

„

"S

oo

$ ?i 1 fig I 1=1 1 ? _J

CO 00

T3CCON

c^^^

O

-

3^"^

>

*O

-a 2 2 a "§ § Q. "IS

2

13

76

=

S

mir

o

eo

J

vol.

>

c

37,

8

1997/1


Methodology in Comparative Research Table 2. Potential Sources of Error

Errors of Definition Specification Error Conceptual non-equivalence Functional non-equivalence

Wrong causal directions Inclusion of non-relevant variables Errors of Estimation Frame Error Failure to include all elements in the chosen population Failure to exclude elements belonging to extraneous populations

Selection Error -

Incorrect sampling procedures

Measurement Error -

Flaws in scale design Flaws in scale mathematical properties Instrument invalidity Incorrect administration of instrument

Non-equivalence of administration Non-equivalence of response Non-equivalence of language Non-equivalence of instrumentation Sampling Error

Errors of Explanation Causal inference error Internal invalidity Universality assumption error External invalidity

Adapted from Cox (1974).

surrogacy problem of considering country as culture (Nasif et al. 1991, Samiee and Jeong 1994) is thus avoided. Cheng's (1989) model of cross-cultural research also takes care of another commonly discussed pitfall, i.e., treating an intra-culturally heterogeneous sample frame as a single culture. Since culture is framed in terms of selected societal (independent) variables, each offering a multi-level choice response format (e. g. Likert scales), it would be theoretically feasible to detect and interpret both between and within cultural nuances at a much finer level of detail, should the need arise.

The above arguments do not however, eliminate the need for what Sander (1994) calls substantive knowledge about the issues being studied. He describes the fundamental pre-requisite of 'good' comparative research as the possession and exercise of sound substantive knowledge of the intricacies of the systems being compared by the researcher. His statements find support in the views of other scholars like Douglas and Craig (1983) who advocate collaboration between remir

vol.

37,

1997/1


77


searchers in different countries, each scholar possessing the requisite substantive knowledge of his/her respective culture, country, and organisation. Hofstede (1980) prefers bi- or multicultural researchers, adept in the ways of different cultures, to composing a team of researchers drawn from multiple cultures. Such an approach helps increase functional equivalence by reducing the chances of intercultural misunderstandings or passive submission to the chief investigators' paradigm. At more operational levels, concepts and phenomena are required to be equivalent in terms of importance and appropriateness across cultures. Conceptual and functional equivalence is a critical feature of sound comparative research especially in the 'softer' areas of management such as consumer and organisational behaviour. The problem does not appear to be as acute for the study of industrial themes such as manufacturing technology and strategy, since concepts here are less likely to be culture bound or functionally different across cultures. Be that as it may, international collaboration among researchers undoubtedly assists in avoiding contentious issues of ethnocentrism, functional and conceptual equivalence in comparative research (Hofstede 1983, Adler 1983). From another perspective, multicultural research teams also lend balance and equivalence to local opinion. General systems theory (Boulding 1956) states that while the researcher is at level 7 of the General Hierarchy of Systems, the unit of analysis is the human organisation, and society at level 8, a higher and more complex level. Since the investigator is less complex than and also a part of the system he/she studies, interpretation difficulties may arise (Hofstede 1980). Multiple viewpoints help resolve such problems. Some additional problems of cross-cultural research design remain in the timing effects of data collection in case studies (Roberts and Boyacigiller 1984), uni-

dimensionality of discipline and paradigm (Hofstede 1980, 1983), and in researcher negligence to establish intra-country comparison groups as a pre-requisite to inter-country effect interpretations. The last point being a salient conclusion of England and Negandhi (1979) who have a word of caution against becoming overly excited about observed national differences, unless 'they are rather large in magnitude, in an absolute sense and in a relative sense, when compared to observed differences within a given country'. The issue of what constitutes a 'large' difference is of course dependent on the domain being studied, the measures being used and the extent of substantive knowledge possessed by the researcher. England and Harpaz (1983), in a subsequent article discuss the use of scaled, relative and inferred scales to gauge inter-country differences relative to intra-country differences. All scales are normalised using the average standard deviation of the measure in the different countries. All of which, of course, brings

us back to the question of recognising intra-country cultural differences and accommodating them in the research design in various ways - the contextual model discussed earlier (Cheng 1989) being one of them.

78

mir

vol.

37,

1997/1



Errors of Estimation

An error of estimation is the discrepancy between the true value of a measure of some object or construct and its observed value. Both uni-variable and multi-variable measures are susceptible to this type of research error. The domain of errors of estimation spans across errors of selection, frame error, non-response error, er-

ror of measurement, and sampling error. Only the last being amenable to measurement with any degree of real precision (not to be confused with the accuracy or validity of results). We begin with the treatment of frame and selection errors in cross-cultural research and go on to consider errors of measurement, non-response and sampling, in turn. Any kind of error can be a fatal flaw in a research design, depending on its nature and severity. However, as we shall see, certain errors are more amenable to a-priori damage avoidance and ex-post facto damage control than others.

Frame and Selection Errors

Frame error occurs when the researcher fails in including all the elements of the population or universe under study, or excluding those elements, which belong to a different universe. The impact of frame error is felt in two ways: first, due to inaccuracy in the actual frame itself and second, through the existence of systematic differences between that part of the population accurately included in the

frame and those subjects who have been incorrectly included or excluded from the frame. Samiee and Jeong (1994) advocate paying close attention to the sampling frame in the countries being studied and ensuring that subject selection max-

imizes within-subsample homogeneity. The broader question of which countries to select for the sample frame should be guided by three criteria. One, the number of cultures selected should be large enough to both randomize the variance on non-matched cultural variables and eliminate distracting hypotheses (Brislin, Lonner and Thorndike 1973). In fact, as Adler (1983) suggests, two- or three-culture studies should be treated as pilot studies, contributing to the development of a greater cumulative body of knowledge in later years. Secondly, the research objective has to do a lot with establishment of frame boundaries (Sanders 1994). Popper's (1959) position on falsification as a requisite to theory validation encourages the inclusion of more countries in the sampling frame if the research purpose is theory testing across cultures. Care should be taken to ensure sufficient variance between countries and cultures

in this regard. On the other hand, the scope of the frame need not be as wide if the research objective is to compare and contrast countries or cultures to obtain benchmarks or points of explanatory reference on such issues as specific management practices. The final criterion is also the one most fulfilled by researchers,

mir

vol.

37,

1997/1


79


i.e., the criterion of convenience. Although Adler (1983) bemoans the selection of cultures based on opportunistic availability, a parsimonious approach to factors of time and cost should guide any research endeavour. This is not to suggest neglect of theory or intellectual compromise as much as to emphasise the need to adopt a cost/benefit policy in the delineation of the sample frame. Selection errors occur as a consequence of sampling procedure, 'if certain elements in the frame have a greater chance of falling into the sample than others, and if this difference is not corrected by subsequent weighting' (Mayer 1971). Sample representativeness may be in question if this happens. Randomization through probability sampling is a luxury afforded to few cross-cultural researchers and non-probabilistic approaches abound (see Douglas and Craig 1983, for an

excellent exposition on sampling techniques and issues). Under such circumstances, the best feasible course is to describe the sample characteristics in detail with particular reference to those factors which may impact the results of their interpretation. Given the difficulty of conducting probability sampling in crosscultural research, the next sampling goal should be the achievement of sampling equivalence. One way to increase sampling equivalence though not necessarily representativeness, is through the use of 'matched' samples across cultures (Sekaran 1983). There is another sampling issue which merits mention here, often referred to as 'Galton's problem', i.e., the non-independence of samples, quite obviously impossible to guarantee, without resorting to random sampling at some

level of the sampling plan. Both Adler (1983) and Sekaran (1983) do not view non-independence of samples as a major concern and in fact, question the desirability and feasibility of sample independence, considering the global interrelatedness of contemporary business. However, if non-independent cultures are chosen to reflect specific independent cultural variables in a study as suggested by Cheng (1989), non-independence of samples could give rise to several well known statistical artifacts. On the other hand, deliberate, stratified sampling to include desired cultures could also induce questions of both internal and external validity. Randomization, at some level of the sampling plan appears to be the best way out, if at all feasible. Otherwise, statistical options should be explored and accompanying clarifications on outcome limitations should be made a part of the study.

Errors of frame or selection are not easily remedied. The best one can usually do on a post-facto basis is to explicitly detail the research circumstances and highlight factors which may impact on data analysis, interpretation and external validity. Study re-design or replication is usually not a very feasible alternative for obvious reasons. Similarly, statistical techniques exist to deal with imperfect probability sampling methods but a non-probability sampling procedure renders it impossible to measure sampling error. The extent of sampling error, of course, can-

not be said to be greater in a non-probability sampling procedure than in a probability sampling procedure. Conversely, it is unrealistic to expect to be able to de-

80

mir

vol.

37,

1997/1



sign out frame-error entirely or conduct pure probability sampling in a cross-cultural research study. What is possible and feasible is the minimization of such errors through a careful consideration of all the factors mentioned in our discussion above, during the design stage of the study.

Errors of Measurement

Measurement error occurs when the measure or scale employed fails to reflect the

correct extent to which the subject possesses the attribute being measured. Measurement errors can arise because of flaws in scale design, flaws in scale mathe-

matical properties, instrument invalidity or incorrect application of the scale. While these are general methodological problems with well established remedies, the complexities of cross-national research require special attention to the issue of reliability in terms of equivalence of language and equivalence of instrumen-

tation (Douglas and Craig 1983, Samiee and Jeong 1994). In addition, the researcher also needs to be aware of the need for equivalence of administration and equivalence of response, both of which concerns are discussed later in this paper. Our mental models and understanding of world phenomena are affected by the nature, richness, range and versatility of our language. The 'whorfian' hypothesis endorses this view by positing similarity in language as a pre-condition for

similarity in mental model (Fishman 1974). Equivalence of language implies that the words used in the measuring instrument convey equivalent concepts and mean-

ings across cultures. Some of the ways such equivalence can be achieved are through the use of high frequency words, equivalence of syntax and grammar, the use of multiple measures and the avoidance of idiomatic expressions (Hofstede 1980, Adler 1983). To enable experiential (equivalent inference across cultures)

equivalence, Sekaran (1983) and others (Adler 1983, Douglas and Craig 1983) suggest the use of iterative back translation and parallel translation techniques with native speakers and inhabitants. However, one needs to be cognisant of the danger of obliterating cultural nuances through excessive equivalence. Cross-cultural translation encompasses four basic types (Casagrande 1954, Brislin 1980) pragmatic, linguistic, aesthetic-poetic and ethnographic. We are more concerned with translations of the last category, our objective being to explicate the cultural context of the source and target language versions. This, however, may be the most difficult to accomplish among the four different types of translations since the translator must take information, grammar, aesthetic content and cultural considerations into account while striving for equivalence in the translated version of the instrument. Hofstede (1980) also suggests using bi- or multilingual translators, preferably organising affairs such that translators translate into their preferred language, as it takes greater linguistic familiarity to be able to articulate nuances to others than to merely be cognisant of them. Panel

mir

vol.

37,

1997/1


81


analysis by multi-lingual experts and the comparison of correlations and loadings among individual language data-sets through the use of factor analysis are some other methods for enhancing linguistic equivalence. Translation errors can also be randomised by increasing the number of languages in the study although this may increase the cost and complexity of the research project.

Equivalence of instrumentation involves equivalence of variables, equivalence of metrics (scale), and equivalence of experimental manipulation. Basing variables on conceptually and theoretically sound foundations (Cheng 1989) resolves the first concern to some extent. Once functional and conceptual equivalence has been established, the issue becomes one of designing an instrument which adequately reflects the attribute being measured across different cultures. Different scales have displayed different degrees of validity (Sekaran and Trafton 1978, Sekaran and Martin 1982) and even a 3-point increase over a 4-point scale was found to enhance instrument sensitivity noticeably (Barry 1969). The validation of measures in a cross-cultural setting has been traditionally addressed in two ways. The development of individual emic (culturally specific) measures for each culture or the creation of etic (culturally universal) measures for multiple cultures (Green and White 1976, Van Raaij 1978). Etic instruments (properly translated) can be administered in identical form in a number of nations.

Each approach has its documented problems (see Brislin, Lonner and Thorndike 1973 and Triandis, Malpass and Davidson 1971, for a detailed discussion of the use of emic and etic measures in cross-cultural studies). Emic measures support functional equivalence, i.e., they have functionally equivalent interpretation for respondents in each culture. Etic measures support what Green and White (1976) term as 'formal equivalence', i. e., identical questionnaires or items for all respondents in a multicultural study. We refer the reader to Van Raaij 's (1978) taxonomy of measure equivalence which essentially compartmentalizes measures into four distinct types, i.e., etic, ipsatized, modified and emic, the second and third types being culturally modified versions of absolute etic and emic measures. An alternative way to visualise instrumentation validity may be as a continuum of equivalence, anchored by pure emic and etic measures at each end. Virtually all domestic research employs emic measures which had been customarily administered unilaterally or with slight modification to foreign cultures (Green and White 1976), thus possibly confounding cross-cultural comparisons. Some attempts have been made to develop etic instruments. One example is the semantic differential scale type (Osgood, May and Miron 1975) which has been tested internationally with consistent results. Similar exercises to create etic measures have been attempted thorough the adaptation and acculturization of originally emic scales, the Myers and Warner Colloquial Instrument scale modified for use in a French setting (Douglas and Craig 1983) being a case in point. We do not foresee the development of a totally culture-free instrument in the social sciences, even for such relatively objective, quantifiable areas like manufacturing or oper-

82

mir

vol.

37,

1997/1



ations. Basic cultural values differ along too many dimensions and cultural convergence remains more a subject of academic debate than a real-life phenomenon (Fram and Ajami 1994, Ralston et al. 1993, Adler, Doktor and Redding 1986). Several methods exist to alleviate the instrumentation equivalence problem. Separate emic measures could be employed in each cultural setting and the correlations between scores and other study variables examined in the cultures under study (Triandis et al. 1971). Statistical reliability, stability and validity tools such as principal components analysis, factor analysis, tests of internal consistency and parallelism etc. are also available for validating cross-cultural instruments. Another option may be to use an emic instrument in a multi-cultural study and try to

identify and rationalise the differences in the resulting scores for obtaining some level of basic understanding about the concept under study (Frijda and Jahoda

1966). Van Raaij (1978) also discusses the feasibility of combining emic and etic items in a single scale, such a method has shown encouraging results for overall scale reliability. The use of expert opinion, conducting independent research and subsequent comparison, concurrent and participative development of equivalent instruments and appropriate training of instrument administrators are some of the

other suggestions available in the literature. Statistically speaking, errors of measurement can be identified and corrected

through the use of specific statistical formulae. Several analytical techniques (e. g. structural modelling and path analysis) allow for the use of corrected effects in parameter estimation. Mechanisms exist to assess the reliability of item measures (factor analysis, Cronbach's alpha, etc.) for use in error correction formulae. The 'coefficient of congruence' computation (Cavusgil 1985) represents another method for estimating reliabilities through the comparisons of inter-cultural factor loadings. The availability of such corrective resources does not however obviate the need for obtaining highly reliable measures in the first place. Although one may be tempted to rely on the correction formulae to compensate for poor measurement, such a strategy is less than optimal because the standard error of a corrected correlation is larger than that of an unconnected correlation. Equivalence of administration represents another concern in reducing errors of measurement. Inaccuracy in data collection can be a function of subject bias or error as well as administrator, setting or timing bias or error. Administration equivalency targets the latter group of potential sources of errors. Essentially, the goal should be to ensure that research settings, instructions and timings are equivalent not identical or divergent. Roberts and Boyacigiller (1984) warn against the

lapse of much time between data collection in different countries. The social and educational status and race of the administrator may also affect data collection in many less egalitarian cultures. Administrator motivation problems may also im-

pact data quality (Cox 1974). Researchers present several solutions to issues of cross-cultural administration equivalence. Sekaran (1983) emphasises the use of local administrators and

mir

vol.

37,

1997/1


83


researchers in such studies. Douglas and Craig (1983) offer an excellent analysis of the appropriateness of different data collection techniques (surveys, interviews etc.) and associated instrument administrator training needs in different cultural milieu. The context of the research of course, determines the extent of data vulnerability to administration error. Studies in retail consumer behaviour will likely see data collection through large scale site-interviews and phone surveys as compared to a study of industrial corporate sourcing practices, where investigations are more likely to be undertaken directly by the concerned research scholar(s). Administrator objectivity, expertise and motivation levels will be different in each

of the above scenarios and thus affect administration equivalence differently. Errors of measurement are also caused by lack of equivalence of response. Equivalence of response can be thought to have two dimensions in cross-cultural research. One concerns the response rates across the study locales, both in an absolute and relative sense. The second is related to the first and involves the presence of response bias in the data. Low and/or widely divergent response rates in a multi-cultural study can jeopardize the statistical and external validity of the research as well as affect comparability across samples. The presence of systematic differences between respondents and non-respondents in any research study can corrupt findings seriously. Researchers have proposed a variety of techniques to deal with response equivalence issues. These can be broadly clubbed under two headings: ways to increase response rates and ways to minimize non-response error and bias. The greater the sample size and the greater the response rate, the more robust the statistical conclusions and the more accurate the population parameter estimations. While sample representativeness and size is a function of selection accuracy and temporal/financial constraints, several studies have been conducted into the issue of increasing response rates, mostly in mail survey format. Although the superiority of tele-

phone and personal interviews over mail surveys for generating responses is generally established (Yu and Cooper 1983), cost and time restraints present significant barriers to using the former two procedures in cross-cultural research. Much of cross-cultural data collection is therefore operationalized through the mail survey, despite the presence of risk factors such as undependable postal systems, illiterate respondents and excessive time.

Findings in regard to enhancing response rates in surveys are mixed. Yu and Cooper (1983) in their exhaustive review of response rate research concluded that monetary incentives, preliminary notification, personalization and follow-up had significant positive effects on response rates. Neither personal appeals nor appeals to the social conscience were found to affect response rates significantly. Surprisingly, questionnaire length was not found significantly related to the response rate. In a subsequent study, Albaum and Strandskov (1989) found preliminary notification and detailed explanation of little value to survey data collection strategy. Specific research into cross-cultural survey response rates indicates that

84

mir

vol.

37,

1997/1



incentives and strategies can affect response rates but in varying degree both within

and among countries (Eisenger et al. 1974, Keown 1985, Ayal and Hornik 1986). Jobber, Mirza and Wee (1991) underscored the efficacy of enclosed (sent with the instrument) incentives relative to promised (survey results will be mailed to you) incentives. One important point to keep in mind in this regard is the potential of offers to share information to negatively impact response rates because respondent confidentiality concerns. Another important, though tentative conclusion of the study was that local language instruments may encourage response. Mintu, Calantone and Gassenheimer (1993) investigated the subject further and recommended the use of mail surveys in international research, provided the appropriate procedures are applied. Questionnaire item formulation and source of origin, subject relevance and organisational endorsement/sponsorship constitute some of these.

Scholars have also paid considerable attention to the second dimension of the

response equivalency question. Douglas and Craig (1983) discuss enhancing response accuracy through the application of non-verbal scales in low literacy countries and different point scales in different countries, although problems exist with

such procedures too. Sekaran (1983) suggests using uniform data collection procedures including identical methods of introduction, task instructions, concluding remarks etc., though the caveat of 'not identical but equivalent' remains un-

diminished. Others have suggested study design and administration in ways directed to ensure equivalence of response. Subject familiarity with the instrument and social environment, similar pre-test psychological levels, similar experimenter effects and similar presentation strategy are some specifics in this regard (Adler 1983). Attention also needs to be paid to some of the common issues in response error and bias such as fatigue levels, commitment levels, ego/humility errors/biases and consistency in responses (Hulbert and Lehmann 1975), all of which are subject to cultural affects. Extreme response patterns, with responses

congregating at the ends of the scale, can be experienced in cross-cultural research, affecting internal consistencies, making group mean differences uninterpretable and confounding output analysis (Samiee and Jeong 1994). Such effects can be traced to other than culture factors, such as the presence of a 'ceiling effect'

in the scale. Such response patterns can be attributed to cross-cultural differences only by the careful elimination of alternative explanations through the use of alternative response formats or multiple measures in the research instrument. In summary, response equivalence in cross-cultural research can be achieved

through the strategic employment of culture-tailored domestic response rate incentives and equivalence of administration and experimental condition. Equivalence, again, is not to be confused with uniformity or standardisation. It is possible to detect and evaluate non-response bias along key frame char-

acteristics through a variety of statistical mechanisms - the 'KolmogorowSmirnov' test (Siegel and Castellan 1988) and uni/multivariate tests of variance,

mfr

vol.

37,

1997/1


85


for instance. It may be harder however, to correct for the same, once found in any

large measure. One can attempt to increase the response rate through follow-ups but any such effort essentially represents a trade-off between sample representativeness and sample size. Cajoling responses from initially reticent subjects may only succeed in eliciting data from a special sub-section of the population with specific physical or psychological attributes significantly disproportionate to their actual distribution in the population. As advised in earlier situations, we recommend that generalizability claims have to be made with due care and explicit boundaries under such circumstances.

Sampling Error We do not spend much time in describing this category of errors of estimation, since it has been a traditional focus of parametric statistics. Robust parametric tests have been developed which tolerate deviations from randomization, normality and homogeneity of variance requirements to a certain degree. Sampling error is the only type of error conducive to measurement with any degree of precision (Cox 1974). By this virtue, however spring concerns of a larger nature, which we re-visit briefly. There is considerable controversy about the importance of the significance test (Carver 1987, Folger 1989, Chow 1989). It is generally accepted that statistical significance does not imply theoretical or practical significance. Statistical significance is affected by the power of the test, the size of the sample, the number of effect estimates, the reliability of the measures and the size of the effect and its standard error. Sampling error estimation obtained through effect

magnitude assessment by drawing confidence intervals contains considerable more information than the significance test. Of course, large samples are required to reduce the size of the confidence intervals and increase the estimate accuracy. Significance cannot be used to judge effect size either, in the absence of substantive knowledge of the situational context. Statistical significance tests, effect sizes,

power estimates, confidence intervals and the like are meaningful only when one has some knowledge of the substantive considerations involved. None of these tests constitute a claim for causality. These arguments are readily extendable to the cross-cultural domain and in much stronger form. Randomization becomes a far more difficult objective to accomplish when dealing with multiple levels of analysis in multiple research milieus. Similarly, normality and homogeneity of variance requirements are hard to maintain within permitted limits across multiple cultures and countries. The inherent and additional complexity of multi-national research renders identification of errors, maintenance of assumptions and interpretation of statistical results more

difficult than in domestic or uni-location studies. Sampling error estimations in comparative studies should be referenced and standardized with respect to the ex-

86

mlr

vol.

37,

1997/1



tent of the presence of the underlying statistical assumptions in each culture or country. Results should be presented in terms of both statistical significance and magnitude estimates which are bounded by confidence intervals. Causal inferences must not be drawn from the mere presence of 'significant' effects in the data.

Errors of Explanation Errors of explanation are errors of causal inference and errors of universality assumption. The elements affecting internal and external validity and the trade-offs implicit between the two in the design of experimental vs. field studies have been

extensively described in the literature (Campbell and Stanley 1963, Cook and Campbell 1979). Suffice to say that any conclusion regarding causality should fulfil Mills (Cook and Campbell 1979) threefold criteria of covariation, temporal precedence and the absence of alternative explanation. The frequent unavailability of randomization as a tool for controlling threats to internal validity also requires the researcher to explicate and eliminate each possible threat individually. The enormity of such an undertaking in the context of a multi-cultural study does

not condone the absence of such analysis in research interpretation. Comparative research is especially susceptible to threats of selection and mortality. Similarly researchers should clearly identify any factor capable of restricting the generaliz-

ability of the findings. Since much of cross-cultural research is undertaken as quasi-experimental and field research, the threats to external validity may be more

easily resolved or at a minimum, more definitely expressed than those of internal origin.

Data Analysis The need for more analytical rigor in comparative research has been underscored in past literature (Schollhammer 1973, Douglas and Craig 1983). A common point of many observations was the recognition of the limited capability of univariate techniques to unravel complex cross-cultural relationships and the advocacy of multi-variate techniques in such situations (Sekaran 1983, England and Harpaz 1983). The methodological difficulties of conducting empirical comparative research have led to a proliferation of conceptual and qualitative studies (Nasif et al. 1991). The use of parametric techniques has been inhibited by the lack of availability of objective data. On a more positive note, the sophistication of scaling methods and procedures has increased considerably over the past two decades. Sanders (1994) points to the recent developments in multi-level modelling which allow for the

mir

vol.

37,

1997/1


87


simultaneous estimation of individual, regional and regional effects, with 'history and geography included rather than ignored'. The factor congruency method proposed by Cavusgil (1985) for cross-cultural research application represents another useful tool for data analysis. Essentially one uses the technique to compute a 'congruence' coefficient among individual cross-cultural sample factor loadings -a perfect +1 denoting a perfect match among the factors and a perfect 0, perfect unrelatedness among the factors. Similarly, Padmanabhan and Cho (1995) in their insightful analysis of the significance of ostensibly 'insignificant' findings offer valuable advice on specific statistical (factor analysis and regression) methods to deal with the effects of multi-collinearity and multi-dimensionality in comparative research. Current advances in statistical software, PC hardware, and networking have made cross-cultural data sharing and concurrent data analysis very much possible. Ready access to global expertise and multi-cultural perspectives through the World Wide Web represents a valuable resource for particularly knotty analytical or inter-

pretation issues. The internet could potentially become an important medium for cross-cultural primary data collection through the international E-mail facility. Electronic business directories with organisational and executive E-mail addresses are even now available. Any technique, of course, needs to be used in conjunction with good research practice. The preceding discussion of research contributions to the area of cross-cultural research methodology does not dwell on the deficiencies observed therein. For instance, we found no mention of assumptions regarding the relative importance or rank-ordering of the methodological issues identified by the studies. It is expected that certain methodological problems may be more acute in some cultures than others. Western cultures emphasise individuality over collectivism (Hof-

stede 1980). Also, although most cultures incorporate value-orientations in dimensions such as time and nature (Kluckhohn and Strodtbeck 1961), they may rand-order these value-orientations differently. A high PDI (power distance index) society may elicit special problems in such areas as undue subject deference to the interviewer or vice-versa if the subject perceives himself/herself to be super-

ior to the researcher. An investigation of methodological issues of specific significance to specific cultures would be a valuable contribution to cross-cultural re-

search. Another general shortcoming we found among the studies was their emphasis on problems rather than solutions. We need better solutions on how to differentiate between cultural and other environmental variables in cross-cultural

studies - the separation of culture as an exogenous, independent measurable variable.

We also found very little attention being paid to cross-cultural methodology at the within-country level. Distinct and very palpable differences exist at the cul-

tural level across different parts and communities in highly heterogeneous countries such as Malaysia and India. Conceivably, issues of functional and concep-

88

mir

vol.

37,

1997/1



tual equivalence may arise within the same country if widely different sections of society are addressed in the same study. As a case in point, India has 16 official languages in addition to numerous dialects. Translations and researcher collaboration are required for any study to ensure equivalence and generalizability. On a final note, we also found a relative absence of methodological rigor in exploring cause-effect relationships. Techniques such as structural modelling and path analysis are available to suggest causal directions and can be used very fruitfully when underlaid with adequate theory. Also desirable are longitudinal studies which can help separate cause-effect patterns in more definite ways. The expense and time involved may not be totally insurmountable if the research effort can be made a collaborative endeavour.

Our discussions have so far dealt with the methodological problems in comparative research. What we propose next is a step-by-step process which may be of some use in pre-empting or resolving these problems in conducting cross-cultural research. This research process framework is illustrated in Exhibit 1 .

A Process Model for Cross-cultural Research

The first and most fundamental step in the suggested seven-stage procedure involves specifying the theoretical domains of the research construct. One approach (Cheng 1989) suggests the careful selection of dependent organisational variables followed by the choice of theory driven predictor sauce-cultural variables. Only then are specific research sites chosen, based on their relative representation on each of the independent variables. Factors common to the research sites may be treated as environmental parameters. This method enables the researcher to inter-

pret findings in more concrete, theory-linked terms and also increases detection of detail in similarities or differences across samples. Irrespective of the method used however, the cross-cultural researcher has the responsibility of clearly explicating the theoretical underpinnings of the research constructs as well as any emic nuances in the concepts employed. Thoroughness in literature review is absolutely necessary at this stage of the research.

The second step in our process involves the acquisition and application of substantive knowledge about the conceptual and functional equivalence of the constructs. Inter-cultural researcher collaboration affords variety and breadth in the knowledge base. In situations where international research teaming is not possible, the researcher should try to update his/her knowledge of alien cultures by seeking bi-cultural, in-country assistance and by in-depth readings of relevant literature. Particular efforts should be made to obtain literature perspectives from the cultures involved in the study.

ntlr

vol.

37,

1997/1


89


In our opinion, the first two steps are the most critical in the research process.

Errors committed at this stage are potentially the most damaging and irreversible in nature.

The third step involves the creation of an effective and cost-efficient sampling design. The purpose of the study should dictate sampling breath choice. Cross-cultural theory verification calls for more extensive inter-cultural sampling than research which has the objective of best practice comparison. We view the frame selection process as a trade-off between cost and time considerations on the one hand and the demands of variance randomization and generalizability on the other. Another trade-off involves balancing cost, equivalence, interpretation and sampling error concerns while determining sample size in each culture. We propose that the researcher would do well to start with a two sample study in exploratory situations. More mature research should try to add more cultures and dimen-

sions to the research question, although financial constraints often limit broad coverage cross-cultural research. A Hofstede (1980) type study is a relative rarity. We also recommend randomization to the extent possible, at whatever level of analysis feasible, with deviations articulated clearly in the data analysis and results portion of the study.

The fourth step requires developing a sound instrumentation design. If validated scales exist, the researcher should use them. A new measure should become necessary only when existing scales cannot be adapted for use or in the rare situation where scales do not exist at all for the construct under study. We recommend

Churchill's (1979) paradigm for developing new measures and scales, employing Delphi meets, literature reviews and focus groups for item generation, followed by factor analysis, reliability assessment and multitrait-multimethod construct val-

idation. Language equivalency becomes another issue in instrument equivalency. The instrument can be back/parallel translated or offered for a multi-lingual panel

analysis. The assistance of international research collaborators can expedite and accredit the process. To meet response equivalence objectives, we suggest enclosing incentives alongwith the instrument and using common forms of presentation and introduction as far as possible. Research has shown the efficacy of incentives (university embossed bookmark), which may not necessarily be of a pecuniary or particularly expensive nature. The fifth step in our seven tenet program concerns the collection of data. Specific care should be taken to see that data collection in the different research sites

takes place as concurrently as possible. The researcher should also place due emphasis on the subject dimension in the data collection process. Many countries

have markedly different social systems, literacy rates, cultural values and status/prestige symbols from the familiar western norm. Data accuracy is usually the first casualty of subject bias. Researchers can use local administrators, con-

ceal the true origin of the research (if ethically permissible), apply multiple measures, reverse code items etc. to counteract response bias in data collection.

90

mir

vol.

37,

1997/1



Administrator bias too can be a confound at times, though this should not pose a major problem if one works with the international academic community. Otherwise, the best recourse for avoiding administrator bias lies in careful selection and training of personnel in the early stages of the research program. One can also consider replicating selected cases using different administrators, as a measure of detection and control for such sources of bias.

We deem steps three to five as of significant importance to the researcher. Any

errors or shortfalls at this point can jeopardize the ensuing steps in the research design. However, the presence of solid theoretical back-ups (steps 1 and 2) should make it easier for the researcher to recover from misjudgements at these stages.

Steps number six and seven encompass data analysis and interpretation and are usually the most amenable to post-facto corrective action. However, the sophistication of data analysis is dependent on a number of factors including the initial selection of the scale type. Ratio and interval scales lend themselves to statistical testing of a power not available to the ordinal measure. The design and implementation of the former is obviously much more difficult than the latter, es-

pecially considering the additional complexities of cross-cultural research. But the researcher should make a systematic effort to incorporate higher order scales in the instrumentation to the extent possible. Provided the data allows, the use of multivariate methods is recommended. We also stress the use of confirmatory factor analysis for underlying dimension identification across cultures. The data can be subsequently aggregated and compared employing additional statistical

techniques such as congruency analysis. Exploratory factor analysis can be avoided by the a-priori selection of applicable theory. Once relevant factors are confirmed and compared, the researcher can move on to derive and test causal relationships through structural modelling and path analysis. Both local and global fit and significance tests can be performed with due adjustments of errors of measurement.

The seventh and final step in our model focuses on data interpretation. While much of the interpretation will be necessarily theory-based, given properly designed constructs, we wish to emphasise three important requirements in this regard. One is the establishment of in-country/culture benchmarks for independent-

dependent variable linkage effect sizes. Inter-cultural comparisons become meaningful only when the differences are large, relatively and absolutely speaking, when compared to within-culture effect size. Earlier studies may provide some indications for this purpose. The second issue is the question of external validity. Frame, sample and situational parameters determine the degree of generalizability of findings. What is more important to us is the use of more discretion in cross-

applying and aggregating conclusions across countries. Many countries in Asia, for instance, represent cultural pot-pourris in themselves. India with more than a dozen official languages and far more variety in garb, cuisine, values and behaviour exemplifies a case in point. Studies in the northern part of the country can-

mir

vol.

37,

1997/1


91


not be extended to the south, without some major qualifications. Also, countries like Singapore and Thailand are too far apart in a number of ways to be clubbed together for research purposes, except perhaps in very circumscribed dimensions. Preliminary validation of research frame and sampling assumptions can be facilitated using foreign student bodies and bi-cultural local help, if cross-cultural collaboration cannot be done. Finally and by the same token, we advice that researcher

interpretation of cultural data be second-checked by similar methods if results cannot be interpreted in cross-cultural concert. We also incorporate feedback loops in the proposed framework. Actual research will involve iterations in several sub-routines. Perhaps the most important area being the establishment of relevant theory and construct of conceptual and functional equivalence. The model has also been kept relatively simple for illustrative reasons. Additional or intervening stages are entirely feasible as also cases where particular areas may be emphasised over others. The basic steps described however, are drawn from established cross-cultural methodology literature and can enhance the research process, if properly adhered to.

Conclusion

We have traversed substantial ground in this survey and analysis of comparative

research methodology. What remains perhaps is an evaluation of individual streams of cross-cultural research in each major discipline using the methodological criteria developed here. Cross-cultural methodology research has a multi-disciplinary and rich tradition (Psychology - Whiting 1968, Economics - Wagner 1969, Marketing - Green and White 1976). Much of the focus in recent years, however, has been in the area of marketing and consumer research (Samiee and Jeong 1994, Mullen, Milne and Doney 1995, Netemeyer, Durvasula and Lichten-

stein 1991, Winter and Prohaska 1983). While this is not entirely unexpected, given the increasing globalization of the marketing function, commensurate attention has not been extended to other important functional areas equally affected by cross-cultural factors. One reason we ascribe for this uneven treatment is the fact that certain aspects of organisational operations are more subject to environmental forces than others (Katz and Kahn 1978). As such, some aspects of firm activities are inherently more affected by cultural differences than others (human

relations vs. production). What surprised us however, was the relative neglect of cross-cultural issues in such critically important and potentially culturally sensitive areas as global sourcing, for example. The field is ripe for appropriate crosscultural studies of sound conceptual and functional design in such relatively neglected areas of comparative research.

92

mlr

vol.

37,

1997/1



Another issue worth looking at, especially in the cross-cultural context, is the quality and scope of comparative research being undertaken by the non-english speaking part of the academic world. With exports being emphasised as the route to economic nirvana, Taiwan sits on top of a foreign exchange cash pile of well over U.S.-$ 100 billion, while Japan is running a potentially permanent trade surplus with the rest of the world. Such export success mandates the careful examination and analysis of customer economies and cultures - a task which certainly must have been shouldered, at least in part, by in-country scholars. It would be interesting and educative to learn about the issues being investigated by native academia and thinktanks as also about the processes and procedures employed in such investigations. Comparative research design and methodology is a dynamic and an increasingly complex phenomenon. We do not expect the complexity to diminish but advancements in research methodology and analytical techniques should hopefully increase the validity of research findings and certainly reduce the traditional uncertainties associated with international research endeavours.

References Adler, N., A Typology of Management Studies Involving Culture, Journal of International Business Studies, 14, Fall 1983, pp. 29-47. Adler, N.J./Doktor, RVRedding, S.G., From the Atlantic to the Pacific Century: Cross-cultural Management Reviewed, Journal of Management, 12, 2, Summer 1986, pp. 295-318. Adler, N.J., Understanding the Ways ot Understanding: Cross-cultural Management Memoaoiogy Reviewed, in Farmer, R.N. (ed.), Advances in Comparative Management, Greenwich, CT: JAI

1984, pp. 31-67.

Albaum, G./Peterson, R.A., Empirical Research in International Marketing, Journal of international

Business Studies, 15, Spring/Summer 1984, pp. 161-173. Albaum, G./Strandskov, J., Participation in a Mail Survey of International Marketers: Effects of Pre-

Contact and Detailed Project Explanation, Journal of Global Marketing, 2, 4, 1989, pp. 7-23. Aulakh, P.S./Kotabe, M., An Assessment of Theoretical and Methodological Development in International Marketing: 1980- 1990, Journal of International Marketing, 1, 2, 1992, pp. 5-28.

Ayal, I./Hornik, J., Foreign Source Effects and Response Behaviour in Cross-national Mail Surveys, International Journal of Research in Marketing, 3, 1986, pp. 157-167. Barry, H., Cross-cultural Research with Matched Pair of Societies, Journal of bocial Psychology,

79, October 1969, pp. 25-33. Boulding, K.E., General Systems Theory: The Skeleton ot Science, Management science, z, ivoo, pp. 197-208. Brislin,R.W./Lonner,W.J./Thorndike,R.M.,Cra55-cM/mra//?^^arc/iMer/ioJ5, New York: Wiley 1973.

Brislin, R.W., Translation and Content Analysis of Oral and Written Material, in Tnandis, H.C./ Berry, J.W. (eds.), Handbook of Cross-cultural Psychology, Boston, MA: Allyn & Bacon 1980, pp. 389-444. Campbell, D.T./Stanley, J.C., Experimental and Quasi-Experimental Designs for Research, New York: Rand McNally and Co. 1963. Carver, R.P., The Case Against Statistical Significance Testing, Harvard Educational Review, 48,

1987, pp. 378-399.

mir

vol.

37,

1997/1


93

S. Tamer Cavusgil/Ajay Das Casagrande, J., The Ends of Translation, International Journal of American Linguistics, 20, October 1954, pp. 335-340. Cavusgil, ST., Factor Congruency Analysis: a Methodology for Cross-cultural Research, Journal of the Market Research Society, 27, 2, 1985, pp. 147-155. Cheng, J.L.C., Toward a Contextual Approach to Cross-National Organisation Research: A Macro Perspective, Advances in International Comparative Management, 4, 1989, pp. 3-18.

Chikudate, N., Cross-cultural Analysis of Cognitive Systems in Organizations: A Comparison between Japanese and American Organizations, Management International Review, 31, 1991, do. 219-231.

Chow, S.L., Significance Tests and Deduction: Reply to Folger, Psychological Bulletin, 106, 1989,

pp. 161-165.

Churchill, G.A. Jr., A Paradigm for Developing Better Measures of Marketing Constructs, Journal of Marketing Research, February 1979, pp. 64-73. Cook, T. A. /Campbell, D.T., Quasi-Experimentation: Design and Analysis Issues for Field Settings, Boston, MA: Houghton Mifflin Co. 1979.

Cox, E.P., III., Research for Business Decisions: An Interdisciplinary Approach, Bureau of Business Research, The University of Texas at Austin, Austin, TX 1974.

Davis, H.L./Douglas, S.P./Silk, A.L., Measure Unreliability: A Hidden Threat to Cross-national Marketing Research, Journal of Marketing, 45, 2, 1981, pp. 98-109. Dawson, S./Dickinson, D., Conducting International Mail Surveys: The Effect of Incentives on Response Rate with an Industry Population, Journal of International Business Studies, 19, 3, Fall

1988, pp. 491 -496.

Douglas, S.P./Craig, S., International Marketing Research, Englewood Cliffs, NJ: Prentice-Hall Inc. 1983.

Eisinger, R.A./Janicki, W.P./Stevenson, R.L./Thompson, W.L., Increasing Returns in International

Mail Surveys, Public Opinion Quarterly, 33, 1974, pp. 124-130. England, G.W./Harpaz, I., Some Methodological and Analytic Considerations in Cross-national Research, Journal of International Business Studies, Fall 1983, pp. 49-59. England, G.W./Negandhi, A.R., National Contexts and Technology as Determinants of Employees' Perceptions, in England, G.W./Negandhi, A.R./Wilpert, B. (eds.) Organisational Functioning in a Cross-cultural Perspective, Kent, OH: Kent State University Press 1979, pp. 175-190. Fishman, J.A., A Systematization of the Whorfian Hypothesis, in Berry, J.W./Dasen, PR. (eds.), Culture and Cognition, London: Methuen 1974, pp. 61-85. Flemming, H., Managerial Implications of Cross-cultural Studies of Buyer Behaviour, in Woodside, A.G./Sheth, J.N./Bennet, P.D. (eds.), Consumer and Industrial Buying Behaviour, New York: Elsevier North Holland Inc. 1977, pp. 387-395. Folger, R., Significance Tests and the Duplicity of Binary Decisions, Psychological Bulletin, 106, 989, pp. 155-160. Fram, E.H./Ajami, R., Globalization of Markets and Shopping Stress: Cross-country Comparisons, Business Horizons, 37, 1, 1994, pp. 17-23. Frijda, N./Jahoda, G., On the Scope and Methods of Cross-cultural Research, International Journal of Psychology, 2, 1966, pp. 109-127. Gardner, E.S. Jr./Ivancevich, J.M., Productivity in the U.S. and Japan: A Re-examination, Interfaces, 24, 6, November/December 1994, pp. 66-73. Green, R.T./White, P.D., Methodological Considerations in Cross-national Consumer Research, Journal of International Business Studies, 14, Fall/Winter 1976, pp. 81-87. Hofstede, G., Cultures Consequences: International Differences in Work-related Values, Beverly

Hills, CA: Sage Publications 1980.

Hofstede, G.H., Value systems in forty countries: Interpretation, validation and consequences for theory, in Eckensberger, L.H./Lonner, W.J./Poortinga, Y.H. (eds.), Cross-cultural contributions to psychology, Lisse, Netherlands: Swets and Zeitlinger 1979, pp. 389-407. Hofstede, G., The Cultural Relativity of Organisational Practices and Theories, Journal of Interna-

tional Business Studies, 14, 2, 1983, pp. 75-89. Hulbert, J./Lehmann, D.R., Reducing Error in Question and Scale Design: A Conceptual Framework, Decision Sciences, 6, 1975, pp. 166-173.

94

mir

vol.

37,

1997/1


Methodology in Comparative Research Jaffe, E.D./Nebenzahl, I.D., Alternative Questionnaire Formats for Country Image Studies, Journal

of Marketing Research, 21,4, November 1984, pp. 463-471. Jobber, D./Mirza, H./Wee, K.H., Incentives and Response Rates to Cross-national Business Surveys:

A Logic Model Analysis, Journal of International Business Studies, Fourth Quarter 1991, pp. 711-721. Kanuk, L./Berenson, C, Mail Surveys and Response Rates: A Literature Review, Journal of Marketing Research, 12, November 1975, pp. 440-453.

Katz, D./Kahn, R.L., The social psychology of organizations (2nd ed.), New York: Wiley 1978. Keown, C.F., Foreign mail surveys: Response Rates Using Monetary Incentives, Journal of International Business Studies, 16, 3, Fall 1985, pp. 151 - 153. Kluckhohn, F.R./Strodtbeck, R.L., Variations in Value Orientation, Evanston, IL: Row, Peterson 1991.

Kotabe, M./Duhan, D.F./Smith, D.K./Wilson, R.D., The Perceived Veracity of PIMS Strategy Principles in Japan: An Empirical Injury, Journal of Marketing, 55, January 1991, pp. 26-41. Martin, I.M./Eroglu, S., Measuring a Multi-Dimensional Construct: Country Image, Journal of Busi-

ness Research, 28, 1993, pp. 191-210.

Mayer, C.S., Application of Bayesian Statistics to Research Design, in Albaum, G. and Venkatesan, M. (eds.), Scientific Marketing Research, New York: The Free Press 1971, pp. 86- 100. Mintu, A.TVCalantone, R.J./Gassenheimer, J.B., International Mail Surveys: Some Guidelines for Marketing Researchers, Journal of International Consumer Marketing, 5, 1, 1993, pp. 69-83.

Mitchell, R.E., Survey Materials Collected in Developing Countries: Sampling, Measurement, and Interviewing Obstacles to Intra- and Inter-national Comparisons, in Boddewyn, J. (ed.), Comparative Management and Marketing, Glenview, IL: Scott, Foresman & Co. 1969, pp. 232-252. Mullen, M.R./Milne, G.R./Doney, P.M., An International Marketing Application of Outlier Analysis for Structural Equations: A Methodological Note, Journal of International Marketing, 3, 1,

1995, pp. 45-62.

Nasif, E.G./Al-Daeaj, H./Ebrahimi, B./Thibodeaux, M.S., Methodological Problems in Cross-cultural research: An Updated Review, Management International Review, 31, 1991, pp. 79-91. Netemeyer, R.G./Durvasula, S./Lichtenstein, D.R., A Cross-National Assessment of the Reliability

and Validity of the CETSCALE, Journal of Marketing Research, XXVIII, August 1991,

dd. 320-327.

Osgood, C.E./May, W.H./Miron, M.S., Cross-cultural Universals of Affective Meaning, Urbana, IL: University of Illinois Press 1975. Padmanabhan, P./Cho, K.R., Methodological Issues in International Business Studies: the Case of Foreign Establishment Mode Decisions by Multinational Firms, International Business Review,

4,1, 1995, pp. 55-73.

Parameswaran, R./Yaprak, A., A Cross-National Comparison or Consumer Research Measures, Journal of International Business Studies, Spring 1987, pp. 35-49. Peter, J.P., Reliability: A Review of Psychometric Basics and Recent Marketing Practices, Journal of Marketing Research, XVI, February 1979, pp. 6- 17.

Popper, K., The Logic of Scientific Discovery, London: Hutchinson 1959. Przeworski, A./Teune, H., Equivalence in Cross-National Research, Public Opinion Quarterly, 30, Winter 1966. pp. 551-568.

Ralston, D.A./Gustafson, D.J./Cheung, F.M./Terpstra, R.H., Differences in Managerial Values: A Study of U.S., Hongkong & PRC Managers, Journal of International Business Studies, 24, 2, 2nd Quarter 1993, pp. 249-275. Rehder, R.R., Japanese Transplants: In Search of a Balanced and Broader Perspective, Columbia Journal of World Business, 24, 4, Winter 1989, pp. 17-28. Roberts, K.H./Boyacigiller, N.A., Cross-National Organisational Research: The Grasp of the Blind Men, Research in Organisational Behaviour, 6, 1984, pp. 423-475. Samiee, S./Jeong, I., Cross-Cultural Research in Advertising: An Assessment of Methodologies, Journal of the Academy of Marketing Science, 22, 3, 1994, pp. 205-217. Sanders, D., Methodological considerations in comparative cross-national research, ISSJ, 142, UNESCO 1994, pp. 513-521.

mir

vol.

37,

1997/1


95

S. Tamer Cavusgil/Ajay Das Schneider, S.C., Strategy Formulation: The Impact of National Culture, Organisation Studies, 10, 2, 1989, pp. 149-168. Schollhammer, H., Strategies and Methodologies in International Business and Comparative Management Research, Management International Review, 1 3, 6, 1 973, pp. 1 9 - 32; reprinted in Man-

agement International Review, 34, Special Issue 1994/1, pp. 5-20. Siegel, S./Castellan, N.J. Jr., Non-parametric Statistics for the Behavioral Sciences, 2nd ed., New York: McGraw-Hill 1988.

Sekaran, U./Martin, H.J., An Examination of the Psychometric Properties of Some Commonly Researched Individual Differences, Job and Organisational Variables in two Cultures, Journal of International Business Studies, Spring/Summer 1982, pp. 51-66. Sekaran, U./Trafton, R.S., The Dimensionality of Jobs: Back to Square One, Twenty-first Midwest

Academy of Management Proceedings, 1978, pp. 249-262. Sekaran, U., Methodological and Theoretical Issues and Advancements in Cross-Cultural Research, Journal of International Business Studies, Fall 1983, pp. 61 -73. Sullivan, JVSnodgrass, C, Tolerance of Executive Failure In American and Japanese Organizations, Asia Pacific Journal of Management, 8, 1, April 1991, pp. 15-34. Triandis, H./Malpass, R./Davidson, A., Cross-Cultural Psychology, Biennial Review of Anthropology, Annual Review, Palo Alto 1971. Tully, S., Purchasing's New Muscle, Fortune, 20th February 1995, pp. 75-83. Van Raaij, W.F., Cross-cultural Research Methodology as a Case of Construct Validity, in Hunt, H.K. (ed.), Advances in Consumer Research, Ann Arbor, MI: Association for Consumer Research

1978, pp. 693-701. Verba, S., Some Dilemmas in Comparative Research, World Politics, 20, October 1 967, pp. 1 1 1 - 1 27.

Verba, S., Cross-national Survey Research: The problem of Credibility, in Valier, I. (ed.), Compar-

ative Methods in Sociology: Essays on Trends and Applications, Berkeley, CA: University of California Press 1971, pp. 309-356. Wagner, PL., On Classifying Economics, in Boddewyn, J. (ed.), Comparative Management and Marketing, Glenview, IL: Scott, Foresman & Co. 1969, pp. 122-134. Whiting, J.W.M., Methods and Problems in Cross-cultural Research, in Lindzey, G. and Aronson,

E. (eds.), The Handbook of Social Psychology, Reading, MA: Addison- Wesley 1968, pp. 693-728.

Winter, L.G./Prohaska, C.R., Methodological Problems in the Comparative Analysis of International Marketing Systems, Journal of the Academy of Marketing Science, 11,4, 1983, pp. 417-432. Yu, J./Cooper, H., A Quantitative Review of Research Design Effects on Response Rates to Questionnaires, Journal of Marketing Research, XX, February 1983, pp. 36-44.

96

mlr

vol.

37,

1997/1
