Kohn Melvin Cross National.pdf0

download Kohn Melvin Cross National.pdf0

of 26

Transcript of Kohn Melvin Cross National.pdf0

  • 7/30/2019 Kohn Melvin Cross National.pdf0

    1/26

    Cross-National Research as an Analytic Strategy: American Sociological

    Association, 1987 Presidential Address

    Melvin L. Kohn

    American Sociological Review, Vol. 52, No. 6. (Dec., 1987), pp. 713-731.

    Stable URL:

    http://links.jstor.org/sici?sici=0003-1224%28198712%2952%3A6%3C713%3ACRAAAS%3E2.0.CO%3B2-B

    American Sociological Review is currently published by American Sociological Association.

    Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available athttp://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtainedprior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content inthe JSTOR archive only for your personal, non-commercial use.

    Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained athttp://www.jstor.org/journals/asa.html.

    Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printedpage of such transmission.

    JSTOR is an independent not-for-profit organization dedicated to and preserving a digital archive of scholarly journals. Formore information regarding JSTOR, please contact [email protected].

    http://www.jstor.orgThu May 17 18:17:11 2007

    http://links.jstor.org/sici?sici=0003-1224%28198712%2952%3A6%3C713%3ACRAAAS%3E2.0.CO%3B2-Bhttp://www.jstor.org/about/terms.htmlhttp://www.jstor.org/journals/asa.htmlhttp://www.jstor.org/journals/asa.htmlhttp://www.jstor.org/about/terms.htmlhttp://links.jstor.org/sici?sici=0003-1224%28198712%2952%3A6%3C713%3ACRAAAS%3E2.0.CO%3B2-B
  • 7/30/2019 Kohn Melvin Cross National.pdf0

    2/26

    CROSS-NATIONAL RESEARCH AS AN ANALYTIC STRATEGY*Am erican . Sociological A ssociation, 1987 Presidential AddressMELVINL. K O H N

    The Johns Hopkins U niv ers ip

    In this essay, I discuss some of the uses anddilemm as of cross-national research. I argue thatcross-national research is valuable, even indis-pensable, for establishing the generality offindings and the validity of interpretationsderived from single-nation studies. In no otherway can we be certain that what we believe tobe social- structu ral regularities are not merelyparticularities, the product of som e limited set ofhistorical or cultural or political circumstances. Ialso argue that cross-national research is equallyvaluable, perhaps even more valuable, forforcing us to revise our interpretations to takeaccount of cross-national differences and incon-sistencies that could never be uncovered insingle-nation research.My thesis is that cross-national researchprovides an especially useful method for gener-ating, testing, and further developing sociologi-cal theory. As with any research strategy,cross-national research comes at a price. It iscostly in time and money. it is difficult to do,and it often seems to raise more interpretiveproblems than it solves. Yet it is potentiallyinvaluable and, in my judgment, grossly under-utilized. This is hardly a radically new thesis.As Stein Rokkan (1 964) long ago pointed o ut, todo cross-national research is to return to thepreferred analytic, strategy of the forefathers ofsociology, a strategy that was nearly abandonedin sociology's quest for methodological rigorbut now can be pursued anew with the muchmore powerful methodological tools availabletoday.

    " Direct all correspondence to Melvin L. Kohn,Department of Sociology, The Johns Hopkins U niversity,Baltimore, MD 21218.I am indebted to my collaborators in cross-nationalresearch: Carrni Schooler, Kazimierz M. Slomczynski,Joanne Miller, Carrie Schoenbach, Atsushi Naoi, and(some years ago) Leonard I. Pearlin; to the sponsors ofthe Polish and Japanese studies: Wlodzimierz Weso-lowski and Ken'ichi Tominaga; and to colleagues whohave critiqued one or another version of this paper:Stephen G. Bunker, Christopher Chase-Dunn, Andrew J.Cherlin, Bernard M. Finifter, W illiam F o m , JonathanKelley, Janet G . Kohn, Tadeusz Krause, John W.Meyer, Joanne Miller, Jeylan T. Mortimer, Alejandro

    A sensible discussion of the uses anddilemmas of cross-national research requiresthat I first define the domain and delineate theprincipal types of cross-national research. ThenI illustrate some of these uses and dilemmas byscrutinizing the body of cross-national research Iknow best, namely my own, my rationale beingWilliam Form's (1979) cogent observation that"probably no field has generated more method-ological advice on a smaller data base withfewer results than has [cross-national] compara-tive sociology." Using my research as a sourceof illustrations makes it possible to discuss theissues concretely. I review this research insufficient detail to highlight its accom plishmen tsand its failures, my concern being only in partwith the substance of the research for its ownsake: I also want to extrapolate from thisconcrete example, to make some more generalobservations. Finally, I discuss some fundamen-tal issues about the conduct of cross-nationalresearch. In so doing, I bring in studies dealingwith quite different substantive problems fromthose that I have addressed in my own research,and using quite different methods, to seewhether my c onclusions apply as well to a muchbroader range of studies.presidents of the American Sociological [Association],among them, William Graham Sumner, W. I. Thomas,E. A. Ross, and Robert E. Park, "exhibited substantialinterest in the comparative study of other societies."Between the 1930s and 1950s, these concerns seemedmarginal to American sociologists; here they again useASA presidents as their index, noting that, of the 20presidents from 1931 to 1950, not one is known primarilyor substantially for (cross-national) comparative work.Leaving aside the obvious question of the validity ofusing the interests of ASA presidents as an index of thesubstantive concerns of U.S. sociology, I would agreewith their generalization and I am intrigued with theirexplanation. They see the "shift toward parochialism" inU.S. sociology of the 1930s and 40s as resulting from acombination of concern with scientific status, constrict-ing resources, attention to immediate social problems(primarily the Depression and World War II), and thepolitical isolationism of American society during thattime. From the vantage point of 1973, Armer andGrimshaw saw a strong revival of cross-national researchoccurring in the 1960s. So, too, did William Evan

  • 7/30/2019 Kohn Melvin Cross National.pdf0

    3/26

    AMERICAN SOCIOLOGICAL REVIEWT Y P E S O F C R O S S - N A T IO N A L R E S E A R C HThe broadest possible definition of cross-national research is any research that transcendsnational boundaries. This definition is somew hatam b~ guo us , hough, because many s tudies ofsingle societies are implicitly cross-national. inthat the investigators interpret their findings bycontrasting what they learn about the countrythey actually study with what is known or isbelieved to be true about some other country orcountries. I prefer to restrict the term, cross-national, to studies that are explicitly compara-tive, that is, studies that utilize systematicallycomparable data from two or more nations.In restricting the term to explicitly compara-tive studies, I do not mean to belittle theimportance of studies that are only implicitlycomparative. Such studies contribute impor-tantly to our understanding; witness, for exam-ple, the distinguished series of studies ofAmerican society by foreign observers, begin-ning with Alexis de Tocquev ille's Democracy inAmerica. Consider, too, s tudies in which theselection of some one country is particularlyappropriate for testing a general p roposition-asin Kelley and K lein's (19 81) use of the Bolivianrevolution of 1 952 to test their theory that"radical revolutions" inevitably lead to anincrease in inequality, or Chirot and Ragin's(1975) use of the Romanian peasant rebellionsof 190 7 to test compe ting interpretations of theintensity of peasant rebellions. And consider,finally. those pivotal studies-Stephen Bu nke r's(1985) Underdeveloping tlze Amuzorz is aparticularly good example- where som e coun-try or region of a country is selected for studyprecisely because it exemplifies a more generalsocial phenomenon. I leave such research out ofmy not because it is unimportant, butbecause to include it would make the bounds of"cross-national" so large and ambiguous that itwould be difficult to say what, other thanresearch focused single-mind edly on a particularcountry. is 110t cross-national.Within the large genre of research that isexplicitly comparative. I would further distin-guish four types of cross-national research ofsomewhat differing intent. The four types arethose in which nation is object of study; those inwhich nation is corztest of study ; those in whichnation is rrnit of arzulysis; and those that aretrarzsrzutiorzal in c h a r a ~ t e r . ~lthough these four

    ? I make n o c l a m that this classification is theoreti-cally superior to other classifications of cross-nationalresearch, only that it serves my analytic purposes betterthan others do. Compared to Tilly's (1984) well-known

    types of research shade into one another, theirpurposes a re distinguishable an d their theoreticalimplications somewhat different. My analysiswill apply mainly to the second of the fourtypes, in which nation is context of study.In the first type of cross-national research,where nations are the object of s tudy, theinvestigator's interest is primarily in the partic-ular countries studied: how Germany comparesto the United S tates, France to the Soviet Union.or India to Pakistan. Alternatively, the investi-gator may be interested in comparing particularinstitutions in these countries: the social securitysystems of the U.S. and Australia; the educa-tional systems of the German DemocraticRepublic and the Federal Republic of G erman y.At their best, as in the systematic comparisonsof Finland and Poland by Erik Allardt, Wlodzim-ierz W esolow ski. and their collaborators (197 8),such studies can lead to well-informed interpre-tations that apply far beyond the particularcountries studied. What distinguishes suchresearch, though, is its primary interest inunderstanding the particular countries. In thisresearch, one wants to know about Finland andPoland fo r their ow n sakes; the investigator doesnot select them for study just because theyhappen to be useful settings for pursuing somegeneral hypothesis.By contrast. I wish to focus on cross-nationalstudies in which, to borrow from ErwinScheuch's (1967) apt phrase. nation is context.In such research, one is primarily interested intesting the generality of findings and interpreta-tions about how certain social institutionsoperate or about how certain aspects of socialstructure impinge on personality. In Burawoyand Lukacs ' (1987 ) comparison of a U.S.machine shop with a Hungarian machine shop,for example, their primary interest is not in theUnited States and Hungary for their own sa kes,nor certainly in the particular machine shops,but in these machine shops as exemplifying therelative efficiency of capitalist and socialistindustrial enterprises. Admittedly. it may bedifficult to differentiate research in which nationis object from research in which nation iscontext. When Robin W illiams (1985 ) s tudiesthe use of threats in USIUSSR relations, he

    sons" (what he sees as two distinct striiregies of researchI see as attempts to interpret two distinct types offindings); my "nation as unit of analysis" category isignored in his classification; and rny "transnational"categoly may be a little broader than his "encompassingcomparisons," which are limited to studies that seenations as cotnponents of encompassing international

  • 7/30/2019 Kohn Melvin Cross National.pdf0

    4/26

    715'KObS NATIONAL RESEARCH AS AN ANALYTIC STRATEGYcleari> is interested in the US and the USSRboth for itleir own sakes and as exemplifyingsupeiyowers in ;I nuclear age; there is no way of\epasating the two purposes. It is neverthelessgenerally useful 10 distinguish between researchwhose primary purpose is to tell us more aboutthe particular countries studied and researchwhose pritnary purpose is to use these countriesas the vehicle for. investigating the contexts inwhich wcial institutions operate. My examina-iiorl of cross.-national research as an analytic\ t r a t e g ~ ,will be addressed mainly to research,\:.here xition is context.'Thii domain includes such diverse studies asThe:!;! Skocpol's ( 1979) comparative analysis of~rcx,.ol~~tion, also, quite a differentnd fromtileoretical perspective, Michael Burton andiohr i Fiipley'\ (1987) analysis of the condit ionsunder which competing elites settle their differ-ences i l l grand political compromises: DonaldTreinlan'i ( 1977) analysis of the stratifications\.stems of the industrialized world; WilliamFornl's (1976) study of the complexity ofindustrial ~echnolog) workers' skill levels. andthe cluality of norkers' social interactions: JanetC'hafetr and Anthon) Dworkin's (1986) analysisof the determinants of the size and range ofideoiopie5 of women's movements throughoutthe world; and my colleagues' and my compar-atix;e research on social stratification andps) cl~ological unctioning in Poland. Japan. andthe United States (Slomczynski, Miller, andKohci 1981 ; Naoi and Schooler 1985).It is useful to differentiate research wherenation is context from two other types ofcross-national research that are not central to mydiicussion here. In the first, where nation is thet t r l i r of analysis. investigators seek to establishrelationship5 among characteristics of nationsqua nations. In such research, one no longerspeaks uf countries by name, but insteadclassifie5 countries along one or more dimen-sioni-their zross national product, or averagelevel of educational attainment, or positionalong some scale of income inequality. Aprototypic example is Bornschier and Chase-Dunn's (1985) analysis of the relationshipbetween the penetration of national economiesliy transnational corporations and the hypothe-s i ~ c d ong-run itagnation of those economies.Other pertinent examples are Blumberg andWinch's (1972) analysis of the relationshipbetween societal con~plexity nd familial com-plexity: and Ellis, Lee, and Petersen's (1978)test of the hypothesis that there is a positiverelationship between how closely adults aresupervised in ,i society and the degree to which

    with understanding hour social institutions andprocesses are systematically related to variationsin national characteristics. Such analyses neednot treat each nation as a homogeneous entity,but may study intranation institutions andprocesses, as Meyer. Hannan, and their col-leagues (1979) have done in their analyses ofnational development. Nor need research thattreats nation as unit of analvsis assume that eachnation exists in an international vacuum. AsBornschier and Chase-Dunn (1985, p. 65) put it,". . . we do not contend that nation-states areclosed systems. A unit of analysis does not needto be a closed system. When we compareindividuals or schools we know that these unitsinteract with one another and are parts of alarger social context. The unit of analysis incomparative research is any unit in which theprocess of interest is known to operate."In distinguishing research that treats nation asthe unit of analysis from research that treatsnation as the context for analysis, we are againdealing with gradations, not sharp differences.As will become evident later, attempts tounderstand cross-national differences sooner orlater require one to search for the pertinentdimensions that differentiate the nations quanations. One can. in fact. argue that research inwhich nation is treated as context is simply away-station to more general analyses in whichthe pivotal distinguishing characteristics ofnations become variables in the analysis. Inprinciple, as Rokkan (1964), Przeworski andTeune (1970), Hopkins and Wallerstein ( 1967),and Chase-Dunn (1982) all argue, one can andshould convert descriptive differences betweencountries into analytic variables. I have noquarrel with this objective. only a belief that inmany fields of sociological inquiry there ismuch to learn from research in which nation istreated as context before we are ready totranslate "nations" into "variables."Research that treats nations as the unit ofanalysis requires that one be able to discernwhich of the many differences between coun-tries are the pertinent analytic variables: that onebe able to formulate meaningful hypotheses atthe appropriate level of abstraction: and-if oneis ever to test such interpretations-that onehave at hand or have the potential to collect datafrom a sizable sample of countries. It alsorequires much better data than are generallyavailable in multination data sources. I hope thatan essay on cross-national research written tenor twenty years from now will be able to focusmuch more on such research than I believe iswarranted today.

  • 7/30/2019 Kohn Melvin Cross National.pdf0

    5/26

    tions (and I hope not distorting their usage of theterm), I call this transnational research. Im-manuel Wallerstein's (19 74, 1980) analysis ofthe capitalist world-system and Fernan do Cardo soand Enzo Faletto's (1979) analysis of depen-dency and development in Latin America areprominent examples. We are at a rather earlystage in the developmen t of approp riate method -ologies for transnational research (Meyer andHannan 1979; Chase-Dunn 1979; Chase-Dunn,Pallas, and Kentor 1982). Even now, though,transnational research has prove d its importanceby dem onstrating that the nations we com pare inall types of cross-national research are notisolated entities but are systematically interre-lated.I see all four types of cross-nation al inquiry asuseful, each fo r particular substantive problems.I focus on research that uses nation as context,not because I consider this type of cross-nationalresearch inherently more valuable than theothers, but because I think that for manysociological problems-particularly, I must ad-mit, for those in which I have the greatestsubstantive interest-this type of research hasespecially great utility in the present state ofknowledge. In particular, such research affordsthe opportunity to study each of the countrieswith sufficient thoroughness for intensive com-parison.ESTABLISHING TH E GENERALITYO F RELATIONSHIPS AND THE LIMITSO F G E N E R A L IT YMany discussions of cross-national research(Ragin and Zaret [I9831 is a thoughtfulexample) contrast two research strategies-onethat looks fo r statistical regularities, another thatsearches for cultural or historical differences. Iprefer to pose the distinction, not in terms ofresearch strategies, nor of methodological pref-erences, nor even of theoretical proclivitiestoward "transhistorical" generalizations or "his-torically contextualized knowledge, " but interms of interpreting the two basic types ofresearch findings- similarities and differences.Granted, investigators' theoretical and method-ological preferences make it more or less likelythat they will discover cross-national similari-ties; granted, too, what can be treated as asimilarity at on e level of analysis can be thoughtof as a myriad of differences at more detailedlevels of analysis. Still, the critical issue is howto interpret similarities, and how to interpretdifferences, when you f ind them.Finding cross-national similarities greatly

    AMERICAN SOCIOLOGICAL REVIEWcult to interpret. As Kazimierz Slomczynski,Joanne Miller, and I argued (albeit a little toocategorically) in our first comparative analysisof the United States and Poland:

    Insofar as cross-national analyses of socialstructure and personality yield similar find-ings in the countries studied, our interpreta-tion can ignore whatever differences therema y be in the cultures, political an d economicsystems, and historical circumstances of theparticular countries, to deal instead withsocial-structural universals. But when therelationships between social structure andpersonality differ from country to country,then we must look to what is idiosyncraticabout the particular cou ntries for our interpre-tation. (1981, p. 740)The first half of this formulation asserts thatwhen the relationship between social structureand personality is the same in two or morecountries, then the unique historical experiencesof each country, their distinctive cultures, andtheir particular political sy stems are no t of focalimportance for interpreting the relationship. Theformulation do es not assert that history, culture,and political context have been irrelevant inshaping social structures, but that the resultantsocial structures have a cross-nationally consis-

    tent impact on people. The explanation of thisimpact should be sought in terms of how peopleexperience the resultant social structures, ratherthan in the historical or cultural processes thatshaped those structures. ~ d m i t t e d l ~ ,his maynot always be the best interpretive strategy.Apparent similarities can mask profound differ-ences; what seems to call for a unitaryinterpretation may actually require entirelydifferent explanations. Nevertheless, I believethat where we find cross-national similarities,the most efficient strategy in searching for anexplanation is to focus on what is structurallysimilar in the countr ies being co mpared , not o nthe often divergent historical processes thatproduced these social-structural similarities. Thebasic and very simple point is that social-structural similarities may have been broughtabout by very different historical processes andyet have essentially similar social and psycho-logical consequences.The second half of the formulation directs usto interpret cross-n ation al dzfferences in terms ofhistorical, cultural, political, or economic idio-syncrasies. Przeworski and Teu ne (1970) arguedthat what app ear to be cross-national differencesmay really be instances of lawful regularities, ifthought of in terms of some larger, more

  • 7/30/2019 Kohn Melvin Cross National.pdf0

    6/26

    CROSS-NATIONAL RESEARCH AS AN ANALYTIC STRATEGYdifferences in history, culture, and political oreconomic systems that distinguish any twocountries are pertinent to explaining the differ-ences w e find in their social structures or in howthese social structures affect people's lives. I donot contend that cross-national differencescannot be lawfully explained-q uite the con -trary-but only that the lawful explanation ofcross-national differences requires more explicitconsideration of historical, cultu ral, and political-economic particularities than does the lawfulexplanation of cross-national similarities.Ultimately, the distinction between cross-national similarities and differences breaksdown, and the issues cannot be so s imply andneatly dichotomized. Nonetheless, it is a usefulway to think about these issues. Therefore, Ishall discuss the two types of cross-nationalresearch findings separately, beginning withcross-national similarities. I use the U . S .-Polishand U.S.-Japanese comparisons that my collab-orators and I have carried out as my principalillustrations of both cross-national similaritiesand differences, my substantive concern in thispart of the essay being the relationship betweensoc ia l s t ruc tu re and pe r~ona l i ty .~he conclu-sions I draw are by no means limited to thissubstantive area.

    ' My concern is not with cross-national similarities ordifferences in personality but with cross-national similar-ities or differences in the relationship between socialstructure and personality. I do not believe that currentmethods are adequate for assessing whether Poles aremore or less intellectually flexible than are Americans, orwhether Japanese value self-direction m ore or less highlythan do Americans. Methodological experts whom Igreatly respect disagree with this judgme nt. They believethat if you construct confirmatory factor-analytic modelsof the same concept for representative samples of twocountries, using not only the same indicators of theconcep t, but also the sam e reference indicator to establishthe metric in both countries, you can compare, e.g., themean level of authoritarian conservatism for U.S. andPolish adults (Schoenberg 1982). This assumes not onlyan exact equivalence of meaning, an issue about whichconfirmatory factor analysis does give us considerableconfidence, but also exact equivalence in the frames ofreference that people employ in answering questions. Idoubt, though, that "strongly disagree" has the sameconnotations in a Polish interview as in an Americaninterview; the survey specialists of the Polish Academyof Sciences believe that it is difficult for Polishrespondents to overcome their cultural tendency to bepolite to their guest, the interviewer. We do not have azero-point for our scales, nor any other basis for meancomparisons. This, however, in no way prevents us fromaccurately assessing whether, for example, the relation-ship between social stratification and authoritarian beliefs

    Cross-na tional Sim ilaritiesOver the course of three decades of research inthe United States , Carmi Schooler and I, incollaboration with Joanne Miller, Karen A.Miller, Carrie Schoenbach, and Ronald Schoen-berg, have intensively studied the psychologicalimpact of social stratification-by which wemean the hierarchical distribution of power,privilege, and prestige (Kohn 1969; Koh n andSchooler 1983). We interpret the consistentrelationships that we have fo und between socialstratification and such facets of personality asvalues, orientations to self and others, andcognitive functioning as the product, in largepart, of the intimate relationship between socialstratification and particular job conditions.People o f higher social-stratification position (asindexed by educational attainment, occupationalstatus, an d job income) enjoy greater opportuni-ties to be self-directed in their work-that is, towork at jobs that are substantively com plex , freefrom close supervision, and not highly routin-ized. The experience of occupational self-direction, in turn, is conducive to valuingself-direction, both for oneself and for one'schildren, to having self-conceptions and socialorientations consonant with such values, and toeffective intellectual functioning. It is evenconducive to seeking out opportunities forengaging in intellectually active leisure-timepursuits (K. Miller and Kohn 1983). All this istrue both for employed men and for employedwomen (J . Mil ler , Schooler , Kohn, and K.Miller 1979; Koh n and Schooler 1983; Kohn,Slomczynski, and Schoenbach 1986).Structural-equation analyses of longitudinaldata have enabled us to confirm even that part ofthe interpretation that posits a causal impact ofjob cond itions on personality (Kohn a nd Schooler1978, 1982; Kohn and Schoenbach 1983).These analyses show the relationships to bereciprocal, with job conditions both affectingand being affected by personali ty. Moreover,analyses of housework (Schooler, Kohn, K.Miller, and K . Miller 1983) and of education (J .Miller, Kohn, and Schooler 1985, 1986)demonstrate that the experience of self-direction, not only in paid emp loym ent, but alsoin housework and schoolwork, decidedly affectspeople's self-conceptions, social orientations,and cognitive functioning. The interpretationhas considerable generality.

    In the absence of appropriate cross-nationalevidence, though, there would be no way ofknowing whether this (or any other) interpreta-tion applies outside the particular historical,

  • 7/30/2019 Kohn Melvin Cross National.pdf0

    7/26

    AMERICAN SOCIOLOGICAL REVIEWintegral part of the social-stratification systemtypical of industrial societies, o r are to be foundonly in the United States, or only in countriesthat have capitalis t economies, or only incountries characterized by Western culture, withits purportedly higher valuation of self-direction. Replications of our research bycolleagues in other countries (for a review, seeKohn and Schooler 1 983, chap. 12 ), particularlythe comprehensive replications that have beencarried out by our Polish and Japanese col-leagues (Slomczy nski et al. 1981; Naoi andSchooler 1985), have made possible tests of thegenerality of the U.S. findings and of thevalidity of our interpretation. In the main, thesefindings are highly co nsistent with tho se for theUnited States, thus greatly enlarging the powerof the interpretation.Of pivotal importance here are the Polish-U.S. comparisons, particularly the comparativeanalyses of men, for whom the Polish studycontains more complete occupational data. Theprincipal issue to which these analyses areaddressed is the specificity or generality of theU.S. findings about the l inkages of socialstratification to job conditions, and of jobconditions to personality. Are these linkagesspecific to the econom ic an d social structures ofcapitalist society, or do they obtain as well insocialist society?We have found, for Poland as for the UnitedStates, that higher social-stratification positionis associated with valuing self-direction, withholding social orientations consonant with sucha value-nam ely, a nonauthoritarian, open-minded orientation, personally responsible stan-dards of morality, and trustfulness (Slomczynskiet al. 1981)-a nd with effectiv e intellectualfunctioning (Slomczynski and Kohn in press).We have further found a strong reciprocalrelationship, for Poland as for the United States ,between social-stratification position and occu-pational self-direction (Slomczy nski et al. 1981).Finally, insofar as possible with cross-sectionaldata, we have shown for Poland, too, a causalimpact of occupational self-direction on values,social orientations, and intellectual functioning(Slomczynski et al . 1981; Slomczynski andKohn in press). Self-direction in one's workleads to valuing self-direction for one's chil-dren , to having a more op en, flexible orientationto society, an d to effective intellectual func tion-ing. Lack of opportunity for self-direction inone's work leads to valuing conformity toexternal authority for one's children, to viewingsocial reality as hostile and threatening, and todiminished intellectual flexibility. The effects of

    This does not mean that these processes arenecessarily the same in all socialist and allcapitalist societies, but it does mean that theU.S. findings are not restricted to capitalistcountries. There is solid evidence, instead, thatthe interpretive model developed for the UnitedStates applies to at least one socialist society."The United States and Poland, of course, areboth Western societies. Are the processessimilar in non-Western societies? The Japanesestudy provides an excellent test of whether ourinterpretation of the U.S. and Polish findingsapplies as well to a non-Western industrializedsociety. In the main, the findings for Japan aremarkedly consistent with those for the UnitedStates and Poland. Social-stratification positionis related to values, to social orientations. and tocognitive functioning in the same w ay , althoughperhaps not to quite the same degree, as in theUni ted Sta tes and Poland (Kohn, Naoi ,Schoenbach, Schooler, and Slomczynski 1987).Occupational self-direction has markedly similareffects on psychological functioning in Japan asin the West (Naoi and Schooler 1985). Thus,despite pronounced cultural differences, anddespite the sharper division between the primaryand secondary sectors of the economy in Japan,the linkages of social stratification to occupa-tional self-direction, and of occupational self-direction to personality, are much the same inJapan as in the United States and Poland. TheU.S. and Polish findings are not l imited toWestern society. Here, again, a single cross-national comparison yields immense benefits forour ability to test the generality of a set ofempirical relationships and their interpretation.Moreover, since the United States, Poland,and Japan are such diverse societies, the set ofthree studies provides prinza facie evidence thatthe psychological impact of social stratificationis much the same, and for much the same

    'The Polish study provides m any further examples ofcross-national similarity. We have found, for example,that in both Poland and the United States, occupationalself-direction not only affects intellective process, butdoes so consistently for younger, middle-aged, and olderworkers (J . Miller, Slomczynski, and Kohn 1985). Wehave further found that, in both the United States andPoland, the social-stratification position of the parentalfamily has a considerable impact on the values of itsadolescent and young-adult offspring (Kohn et al. 1986).The family's stratification position affects both father'sand mother's occupational self-direction; each parent'soccupational self-direction affects that parent's values;the parents' values affect their children's values. Forpresent purposes, these findings are important primarilybecause they show how cross-national evidence strength-

  • 7/30/2019 Kohn Melvin Cross National.pdf0

    8/26

    719R O S S - N A T IO N A L R E S E A R C H A S A N A N A L Y T I C S T R A T E G Yreasons, in all industrialized societies. Admit-tedly, negative evidence from research in anyindustrialized society would require a modifica-tion of this hypothesis or a restriction of itsgenerality. Admittedly, too, the interpretationspeaks only to existing societies . We can saynothing from this evidence as to whether itwould be possible to have an industrializedsociety in which one or another link in theexplanatory chain is broken-a society with aless pronounced system of social stratification; asociety in which social-stratification position isnot so intimately linked with opportunities foroccupationa l self-direction; ev en a society whereoccupational self-direction has less impact onpersonality.5 Nevertheless, the Polish and Japa-nese studies d o tell us that in decidedly d iversesocieties-arguably, in all industrialized socie-ties-social stratification is assoc iated withvalues, social orientations, and cognitive func-tioning, in large part because people of higherposit ion have greater opportunity to be self-directed in their work.W hether o r not this interpretation is correct, itdoe s illustrate my central point: W here on e findscross-national similarities, then the explanationneed not, indeed should not, be focused on theparticular histories, cultures, or political oreconomic circumstances of each of the coun-tries, but instead should focus on social-structural regularities common to them all.In studying social stratification, I am ofcourse dealing with a feature of social structurethat is notably similar in all industrializedsocieties (Treiman, 1977). I would like toextend the argument a bit, to suggest that evenwhere some feature of social structure is not"identical" in all the countries being comp ared ,but only "eq uivalent," it is still possible to findcross-nationally consistent relationships betw eencontemporaneous social structure and personal-ity. More than that, it is still appropriate tointerpret these consistent relationships in termsof contemporaneous social structure, howevermuch that feature of social structure has beenshaped by the particular histories and cultures ofthose countries.My illustration here come s from o ur analysisof position in the class structure an d personalityin the United States , Japan, and Poland (Kohn etal . 1987). For all three countries , we haveadapted the sam e basic idea-that social classes

    Michael Buraw oy's (19 79, p. 13) warning ispertinent, even though our research transcends capitalistsociety: "By taking the particular experiences ofcapitalist society and shaping them into universal

    are to be distinguished in terms of ownershipand control of the means of production, andcontrol over the labor power of others-to theparticular historical, cultural, economic, andpolitical circumstances of the country. (ForPoland, where ownership of the means ofproduction is not a primary desideratum ofclass, control over the means of production andover the labor power of others is our primarycriterion of class posit ion.) Th e guiding hypoth-esis is that social class would bear a similarrelationship to personality as doe s social stratifi-cation. Hence, we hypothesized that, in all threecountries , those who are more advantageouslysituated in the class structure are more self-directed in their values and orientations, and aremore intellectually flexible, than are those whoare less advantageously situated. Our furtherhypothesis, again paralleling what we havelearned for social stratification, is that, in allthree countries, the explanation lies mainly inthe greater opportunities for occupational self-direction enjoyed by those who are moreadvantaged in class posit ion. The hypotheses,then, are simple extrapolations to social classfrom what we have consistently found to be thepsychological impact of social stratification; thenew element is the much greater country-to-country variability of class structures than ofstratification systems.Both hypotheses are confirmed. All threecountries can be meaningfully thought to haveclass structures; class position has similar effectson cognitive functioning, values, and orientationin all three countries; and class affects thesefacets of psychological functioning for essen-tially the sam e reason-becau se of the intimaterelationship between position in the classstructure and opportunities afforded for occupa-tional self-direction. Hence, to extrapolate, it isno bar to structural interpretation that socialstructures have been shaped by distinctlydifferent historical processes.Cross-national DifferencesInterpreting differences, as I said earlier, iswhere things become much less certain andmuch more difficult . The key, of course, is thetruism that if consistent findings have to beinterpreted in tenns of what is common to thecountries studied, then inconsistent findingshave to be interpreted in terms of how thecountries-or the studies-differ. This truism ,unfortunately, gives no clue as to which of themany differences between countries or betweenstudies lies at the heart of the differences in

  • 7/30/2019 Kohn Melvin Cross National.pdf0

    9/26

    AMERICAN SOCIOLOGICAL REVIEWThere is a curious inconsistency in the wayresearchers interpret results from attemptedreplications when discrepancies crop up.Failure to reproduce a finding in the sameculture usually leads the investigator toquestion the reliability, validity, and compa-rability of the research procedures used in thetwo studies for possible method artifacts. Butfailure to corroborate the same finding in ad~fferent culture often leads to claims ofhaving discovered "cultural" differences, andsubstantive interpretations are promptly de-vised to account for the apparent differences.(1977, p. 155)Issues of method. The most fundamentalmethodological issue is whether the conceptsemployed in the analyses are truly equivalent.

    Stefan Nowak posed the issue with characteris-tic clarity:How d o we know w e are studying " the samephenomena" in different contexts; how d o weknow that our observations and conclusionsdo not actually refer to "quite differentthings," which we unjustifiably include intothe same conceptual categories? Or if theyseem to be different, are they really differentwith respect to the same (qualitatively orquantitatively understood) variable, or is ourconclusion about the difference between themscientifically meaningless? (1976, p. 105)(See also Almond and Verba 1963, pp.57-72; Scheuc h 196 7, 1968; Sm elser 1968;Nowak 1977; Marsh 1967; and Armer 1973).

    The issue is so complex that a thoroughtreatment would require quite another essay. Inthis essay, instead, I simply assume equivalenceof concepts and go on to consider moremundane methodological differences.In principle, methodological differences be-tween studies could produce either consistent orinconsistent findings (Finifter 1977). Still, whenone finds cross-national similarities despite dif-ferences in research design, even despite defectsin some of the studies, it is unlikely that thesimilar findings were actually produced by themethodological differences. Substantive similar-ity in the face of methodological dissimilaritymight even argue for the robustness of the find-ings. But when one finds cross-national differ-ences, then d issimilarities and defects in researchdesign make for an interpretive quagmire-thereis no way to be certain whether the apparentcross-national differences are real or artifactual.It can be terribly perplexing not to knowwhether an apparent cross-national difference is

    industrialized societies, neither of which showsthe pattern that has been consistently found infully industrialized societies. One study wasconducted in Taiwan before that is land becameas industrialized as it is today (Stephen Olsen1971), the other in Peru (Scurrah and Montalvo1975). In Taiwan, the relationship betweensocial stratification and parental valuation ofself-direction was essentially the same as hasbeen found in more industrialized societies, butoccupational self-direction fails to explain thisrelationship. In Peru, the correlations of socialstratification with such aspects of personality asfatalism, trust, and anxiety are similar to thosefound in more industrialized societies, butoccupational self-direction explains only amodest portion of these correlations.Should we therefore restrict the interpretationthat occupational self-direction is of centralimportance for explaining the psychologicalimpact of social stratification to apply only tofully industrialized societies? P erhaps w e shou ld,and one can readily think of reasons why theinterpretation might not apply to partiallyindustrialized societies-for exa mp le, the linkbetween social stratification and occupationalself-direction may be weaker in such societies.Bu t, since neither the Taiwan nor the Peru studyis truly comparable to those done in industrial-ized societies (see the discussion in Kohn andSchooler 1983, pp. 293-94), the issue is verymuch in doubt. The Taiwan and Peru studiesleave us in a quandary: They raise doubts as towhethe r the interpretation does apply to partiallyindustrialized societies, but they do not provideconvincing evidence that i t does not.To obviate the possibility that differences infindings are merely an artifact of differences inmethod-in the nature of the samples, in themean ing of the questions asked, in the comp lete-ness of data , in measu rement-o ne tries todesign the studies to be comp arable, to establishboth linguistic and conceptual equivalence inquestions and in coding answers, and toestablish truly equ ivalent indices of the underly-ing concepts (Scheuch 1968). Edward Suchman(1964, p. 135) long ago stated the matter withelegant simplicity: "A good design for thecollection of comparative data should permitone to assume as much as possible that thedifferences observed . . . cannot be attributedto the differences in the method being used."Unfortunately, one can never be certain. Thebest that is possible is to try to e stablish dam agecontrol, to present whatever evidence one canthat methodological incomparables are not sogreat as to explain the differences in findings.

  • 7/30/2019 Kohn Melvin Cross National.pdf0

    10/26

    C R O S S - N A T IO N A L R E S E A R C H A S A N A N A L Y T I C S T R A T E G Y 7 2 1involved in interviewing and in index construc-tion (J. Miller, Slomczynski, and Schoenberg1981; Slomczynski et al . 1981; J . Miller et al .1985; Kohn et al . 1986). So, too, have manyother scholars (see, in particular, Scheuch 1968;Przeworski and Teune 1970; Armer 1973; Elder1976; Kuechler 1986). Therefore, I do notdiscuss these issues further here. Instead, Iassume comparability of methods (as well ascomparabili ty of concepts) and go on to theequally perplexing substantive issues in interpret-ing cross-national differences.Substantive interpretations of cross-nationaldifferences. Finding a cross-national differenceoften requires that we curtail the scope of aninterpretation, by lim iting our generalizations toexclude implicated variables or relationships ortypes of countries from a more encompassinggeneralization. Ultimately, though, we want toinclude the discrepant findings in a morecomprehensive interpretation by reformulatingthe interpretation on a more general level thataccounts for both similarities and differences.Thus, although the discovery of cross-nationaldifferences may initially require that we make aless sweeping interpretation, in time and withthought, i t can lead to more general and morepowerful interpretations.I wish that I could offer from my research anexample of a powerful reinterpretation derivedfrom coming to terms with cross-nationaldifferences. Instead, I can only share with youmy dilemma in s t i l l not fully understandingsome differences that I have been struggling tounderstand for some years . I may not convinceyou that discovering cross-national differencesnecessarily leads to new understanding, but Ishall certainly convince you that the discoveryof such differences forces one to questiongeneralizations made on the basis of studyingonly one country. To i l lustrate, I use the mostperplexing cross-national inconsistencies thatwe have found in the U.S.-Polish-Japanesecomparisons (Kohn et al . 1987).Quite in contrast to our consistent findingsabout the relationship of social stratification toother facets of personality, w e have fou nd a d e-cided inconsistency in the relationship betweensocial stratification and a principal underlyingdimension of orientations to self and others-asens e of well being versus distress. In the UnitedStates, higher stratification position decreasesfeelings of distress; in Ja pan , there is virtually norelationship between social stratification and feel-ings of distress; and in Po land , higher stratifica-tion position increases f ee lin gs o f d i s t r e ~ s . ~he

    magnitude of the correlation is not great in anycountry, but the inconsistency in direction of re-lationship is striking. Similarly for social class:In the United States , members of more advan-taged social classes, mana gers in particular, hav ea greater sense of well-being; members of lessadvantage d social classes, blue-collar work ers inparticular, have a greater sense of distress. InPoland, quite the opposite: It is the managerswho are more distressed, the blue-collar workerswho have a greater sense of well-being. In Ja-pan, as in the United States , managers have astrong sense of well-being, but it is the white-col-lar-not the blue-collar-workers who are mostdistressed.Why don' t advantageous posit ions in thestratification and class systems have cross-nationally consistent effects on the sense ofdistress? On one level, this question is readilyanswered: Our analyses show that stratificationand class matter for psychological functioningprimarily because people of more advantagedposition have greater opportunity to be self-directed in their work. But we find, in causalmodels of the reciprocal effects of occupationalself-direction and distress, that althou gh occupa-aspects of social orientation, but affects some aspects ofself-conception differently. In particular, in the UnitedStates, higher stratification position is associated withgreater self-confidence and less anxiety; in Poland, quitethe opposite."Social orientation" and "self-conception," howe ver,are merely convenient rubrics; they are not underlyingdimensions of orientation. Schooler and I (Kohn andSchooler 1982; 19 83 ,' Chapter 6) subsequently did asecond-order confirmatory factor analysis of the severalfirst-order dimensions of orientation, using U.S. data, todemonstrate that there are two underlying dimensions:self-directedness of orientation versus conformity toexternal authority, and a sense of well-being versus asense of distress. Self-directedness of orientation impliesthe beliefs that one has the personal capacity to takeresponsibility for one's actions and that society is soconstituted as to make self-direction possible. It isreflected in not having authoritarian conservative beliefs,in having personally responsible standards of morality, inbeing trustful of others, in not being self-deprecatory, innot being conformist in one's ideas, and in not beingfatalistic. Distress is reflected in anxiety, self-deprecation, lack of self-confidence, nonconformity inone's ideas, and distrust. We have since shown that thesesame two dimensions underlie the several facets oforientation in Poland and in Japan (Kohn et al. 1987).The basic parameters of the Polish and Japanese models,in particular the relationships between second-order andfirst-order factors, are quite similar to those for the U.S .model. In all three countries, there is a strong positiverelationship between social stratification and self-directedness of orientation. The relationship between

  • 7/30/2019 Kohn Melvin Cross National.pdf0

    11/26

    AMERICAN SOCIOLOGICAL REVIEWtional $elf-direction has a statistically significanteffect (negative, of course) on the sense ofdistress for the United States and Japan, it hasno effect at all for Poland. This is in markedcontrast to the cross-nationally consistent effectsof occupational self-direction on intellectualflexibility. values, and self-directedness oforientation. One can, in fact, incorporate thecross-national inconsistency into an encompass-ing generalization: Where occupational self-direction has cross-nationally consistent effectson psychological functioning, so too do socialstratification and social class; where occupa-tional self-direction fails to have consistenteffccts, stratification and class also have incon-sistent effects.On another level, though, the questionpersists: CV'II! doesn't occupational self-directionmitigate against distress in Poland, as it does inthe United States and Japan? Moreover, occupa-tional self-direction does not provide as effec-tive an explanation of the relationships ofstratificarion and class with distress in any of thethree countries as it does for their relationshipswith other facets of personality in all threecountries. Given the rather substantial effect ofoccupational self-disection on dirtress for theUnited States, we might well expect a highercorrelation of social stratification with distressthan the - 0. 18 that we actually do find. Weshould c,ertainlx expect a higher correlation thanthe -0 .0 1 that we actually do find for Japan.We should expect n o relationship, not a positiverelationship, for Poland. Clearly, more thanoccupational self-direction is involved in explain-ing the relationships of stratification and class todistress. My formulation, which implies thatoccupational self-direction, and therefore alsostratification and class, would have an impact onfeelings of distress consistent with its impact onvalues, self-directedness of orientation, andcognitive functioning, must be revised.[t is not at all certain from the evidence athand. though. whether the interpretation re-quires nlinor revisior~or extensive overhaul. Iam reasonably certain that the cross-nationaldifferences are not merely a methodologicalartifact, for example in the conceptualization ormeasurement of distress. In particular, thecross-national differences are found, not only inanalyses using the "higher-order" concept,distress, but also in analyses using the "first-order" concepts, notably self-confidence andanxiety (see note 6). The issues are substantive,not ~nethodological.In any reformulation, it is essential that wenot lose sight of the fundamental principle that

    cies only for the sense of distress, not for values.for self-directedness of orientation, or forcognitive functioning. Explanations so broadlyframed as to lead one to expect Polish orJapanese men of more advantaged position tovalue conformity for their children. to have aconformist orientation to self and society, or notto be intellectually flexible, could not be valid.Nor would it make any sense to explain thefindings in terms of a weaker linkage of socialstratification or of social class to occupationalself-direction in Poland or in Japan than in theUnited States, or in terms of occupationalself-direction being any less important for Polishor Japanese men than for U.S. men.As I see it, there are at least five ways that myinterpretation might be reformulated:The simplest reformulation would be to limitthe scope of the interpretation to exclude thesense of distress; for as-yet unknown reasons, aninterpretation that does apply to cognitivefunctioning, values, and self-directedness oforientation seems not to apply to the affectiverealm. This reformulation simply curtails thescope of my interpretation, until such time as weare able to develop a more general interpretationthat incorporates cross-national differences alongwith cross-national similarities.A second type of reformulation would positthat the psychological mechanisms by which jobconditions affect distress may be different fromthose by which job conditions affect cognitivefunctioning, values, and self-directedness oforientation. Such a reformulation might ormight not emphasize job conditions differentfrom those that I have emphasized; it certainlywould posit different processes by which jobconditions affect personality. Mine is a learning-generalization model: People learn from theirjob experiences and apply those lessons tonon-occupational realms of life (Kohn 1985).One could argue that the inconsistent effects ofoccupational self-direction on the sense ofdistress raise questions as to whether a learning-generalization model applies to this facet ofpersonality. Perhaps, instead, one should em-ploy some other model of psychological pro-cess-a "stress" model is the obvious candi-date-for understanding the effects of job on thesense of distress. The "stress" model posits thatjob conditions affect personality, in whole or inpart, because they induce feelings of stress,which in turn have longer-term, off-the-jobpsychological consequences, such as anxietyand distress. Clearly, "stress" is a plausible linkfrom job conditions to distress. But I think theevidence for a "stress" model, even when

  • 7/30/2019 Kohn Melvin Cross National.pdf0

    12/26

    72 3ROSS-NATIONAL RESEARCH AS AN ANALYTIC STRATEGYA related possibility, one that is much m ore tomy liking, retains the learning-generalizationmodel but expands the range of pertinent jobconditions. This reformulation begins with theU .S . finding that job conditions other than thosedirectly involved in occupational self-direction

    are more important for distress than for otherfacets of personality (Kohn and Schooler 1982;1983, Chapter 6). Some of these job conditionsare related to stratification and class, hencemight explain the effects-or lack of effects-of stratification and class on distress. The cruxof this reformulation is the hypothesis that theeffects of these other job conditions on distressmay be at odds with, and perhaps moreimportant than, those of occupational self-direction. We have some pertinent, albeitlimited, evidence that lends credence to thispossibility (Kohn et al. 1987). In the UnitedStates, for example, job protections (such asseniority provisions in union contracts) mitigateagainst distress. Nonetheless, the very peoplewho at the time of our interviews enjoyed thegreatest job protections-the blue-collar work-ers-were also the most distressed. Blue-collarworkers were distressed because they lackedopportunities for occupational self-direction anddespite the job protections that many of them,particularly union members, enjoyed. Occupa-tional self-direction a nd job p rotections se em tohave countervailing effects, which may accountfor the relatively modest relationships of bothsocial stratification and social class with dis-tress, even in the United States.For Japan, we f ind that believing that oneworks under considerable pressure of time, andbelieving that people in one's occupation are atrisk of being held responsible for things outsideof their control, are both related to distress.Although these findings may merely reflect apropensity of distressed people to overestimatethe pressures and uncertainties of their jobs, it isat least a plausible hypothesis that such jobconditions do increase distress. Our causalmodels suggest as well that either educationitself, or job conditions related to ed ucation,increases distress. The countervailing effects ofoccupational self-direction, education, and otherjob conditions correlated with them both, mayhelp explain wh y stratification and class have s olittle net effect on distress in Japan.For Poland, we lack information about jobconditions other than those directly pertinent tooccupational self-direction. We do, however,have one fascinating bit of information that mayhelp explain what it is about the conditions oflife experienced by Polish managers that makes

    class is particularly distressed-those manag erswho are not members of the Polish UnitedWorkers (Communist) Party. There are too fewnon-Party managers for this finding to bedefinitive, but I think it suggestive that thenon-Party managers have decidedly higherlevels of distress, compared not only tomanagers who are members of the Party, butalso compared to members of any other socialclass, Party members or not. The implication, Ithink, is that being a non-Party manager in thePolish system of centralized planning entailsuncertainties, risks, a nd insecurities greater thanthose experienced by managers who are mem-bers of the Party, and greater than thoseexperienced by managers in the less centralizedsystems of capitalist countries. The Polishsystem may hold these managers responsible foraccomplishments they have neither the leewaynor the resources to achieve. By the sam e token,the U.S. and Japanese systems may leadmanag ers to fee l- more in control of theconditions of their lives than they really are.Our evidence suggests , then, that not onlydoes occupational self-direction fail to have thecross-nationally consistent effect on distress thatit has on other facets of psychological function-ing, but also, that other job conditions associ-ated with stratification and class may havecountervailing effects. What is lacking isadequate information about these other jobconditions.A fourth type of reformulation would takegreater account of the processes by whichpeople attain their occupational positions and ofthe meaning these positions have to them.Slomczynski, Miller , and Kohn (1981) specu-lated at length about the implications ofpost-World War I1 historical developments thatresulted in differences between the Un ited Statesand Poland in structural mobility, job-selection

    processes, and the symbolic importance attachedto class position-differences that might explainwhy social stratification bears a differentrelationship to distress in the two countries.These speculations still seem to me to beplausible and they are certainly potentiallytestable. One could similarly point to differ-ences between Japan and the West in thestructure of industry, particularly in the sharperdivision in Japan between primary and secon-dary sectors of the economy, that might bepertinent to explaining why stratification has solittle relationship to distress in Japan, and whyJapanese white-collar workers are more dis-tressed than are membe rs of other social classes.Finally, one could broaden the scope of the

  • 7/30/2019 Kohn Melvin Cross National.pdf0

    13/26

    AMERICAN SOCIOLOGICAL REVIEWin religious belief, or in whether the urbanpopulation is primarily rural in origin, or in"national culture" bear on the sense of distress.The pivotal questions, though, are not whetherfamily, religion, rural origins, or culture ac-count for differences in Polish, Japanese, andAmerican men's sense of distress, but whethersuch non-occupational conditions help explainwhy social stratification and social class beardifferent relationships to the sense of distress inPoland, Japan, and the United States .We do not have the evidence to test any ofthese interpretations. Each type of reformulation(other than simply limiting the scope of theinterpretation to exclude distress) would requirea different type of data. To test a "stress"formulation would require more informationabout the relationship between objective jobconditions and the subjective sense of "stress"in one's work, and about the relationshipbetween job stress and off-the-job distress.Similarly, to test any other model of psycholog-ical process would require data directly pertinentto that formulation. To test the hypothesis thatjob conditions other than those involved inoccupational self-direction help explain therelationships of social stratification and socialclass to distress would require that we obtainmuch fuller information in all three countriesabout those job conditions thought to beproductive of a sense of distress. T o test thehypothesis that different processes of educa-tional and occupational attainment account forthe differential effects of stratification and classon the sense of distress would require informa-tion of yet another type: historical informationabout the impact of changes in the educationaland occupational s tructures of Poland, Japan,and the United States s ince World W ar I1 as theyimpinged on particular cohorts of Polish,Japanese, and American workers . And then,finally, to test the rather vaguely formulatedhypothesis that non-job conditions explain thecross-nationally inconsistent relationships ofboth class and stratification with distress wouldrequire information about the interrelationshipof stratification and class with these other linesof social and cultural demarcation, in all threecountries.In any case, on the basis of presentlyavailable evidence, I still do not have a fullyadequate exp lanation of w hy social stratificationand social class have cross-nationally inconsis-tent effects on the sense of distress. Perplexedthough I am, I value the cross-national evidencefor making clear where my interpretation appliesand where it does not, thus defining what is at

    clearly does to values, self-directedness oforientation, and cognitive functioning.

    S O M E G E N E R A L C O N S I D E R A T IO N SI can now address some more general issuesabout cross-national research that I deliberatelydeferred until I had offered some concreteexamples. These remarks are primarily ad-dressed to research in which nation is treated ascontext.

    I . In whose interest is cross-national re-search? This seemingly innocuous questioncontains a range of serious ethical and profes-sional issues. At its worst, as in the infamousCam elot affair (Horow itz 1 967 ), cross-nationalresearch has been used in the service of politicaloppression. In a less dramatic way, cross-national research has too often been a mecha-nism by which scholars from affluent countrieshave employed scholars in less affluent coun-tries as data-gatherers, to secure information tobe processed, analyzed, and published else-where, with little benefit either in training or inprofessional recognition for those who collectedthe data (Portes 1975; Scheuch 1967). Th ese arecomplex issues, w here surface appearances maybe misleading. But, certainly, the history ofcross-national research has not been entirelybenign.Past sins and mistakes notwithstanding, cross-national research need not be employed in theservice of academic or other imperialisms. Myow n research is again illustrative. As a ma tter ofhistorical record, it was not I but WlodzimierzWesolowski (1975, p. 98 ) who proposed thePolish-U.S. comparative s tudy. He did so forprecisely the same reason I found the prospectso attractive when he suggested it to me: to seewhether the U.S. findings would apply to asocialist society. The study was funded andcarried out by the Polish Academy of Sciences,who thought the issues important for Polishsociology and Polish society. The extension ofthe U.S.-Polish com parison to encompass Japancame about because Ken' ichi Tominaga, hisJapanese colleagues, and the Japanese universi-ties and foundations that funded this researchwere as interested as were the Americans andthe Poles in seeing whether these phenom ena aresimilar in that non-Western society.The opportunities for genuine cross-nationalcollaboration today, when there is a thriving,highly professional sociology in many parts ofthe world, are much greater than they were onlya few years ago. Today it is quite possible, and

  • 7/30/2019 Kohn Melvin Cross National.pdf0

    14/26

    725R O S S - N A T IO N A L R E S E A R C H A S A N A N A L Y T I C S T R A T E G Yought to be-equally important for sociologistsof all the countries concerned.2 . Is cross-national research distinctly dzffer-ent @om research that compare s social classes,or ethnic groups, or genders in a singlecountry? I see cross-national research as onetype of comparative research. In many discus-sions, though (see, for example, Armer andGrimshaw 1973), the term "comparative re-search" is treated as synonymous with cross-national research, as if the only possiblecomparison were inter-national comparison; thisI regard as hubris on the part of the internation-alists. In other discussions (e.g., Hopkins andWallerstein 19 67) the term "com parative" isused more broadly and "cross-national" islimited to w hat I consider to be only on e type ofcross-national research, transnational research.And in s t i ll other discussions (e. g. , Ragin1982), comparative research is seen as thatparticular type of cross-national research where"society" is used as the explanatory unit. 'These varying usages seem to me to impedemeaningful discourse. I think it best to use thecommonsense meanings of both "comparative"and "cross-national. "My own research shows that cross-nationalresearch is no different in principle from othercomparative research, although in practice it islikely to be more complex, especially as onetries to interpret cross-national inconsistencies.What makes it worth distinguishing cross-national research from other types of compara-tive research is that a much broader range ofcomparisons can be made: comparisons ofpolitical and economic systems, of cultures, andof social s tructures. Any comparisons we makewithin a sin gle country a re necessarily limited tothe one set of political, economic, cultural, andhistorical contexts represented by that particularcountry. I simply cannot imagine any study ofthe psychological impact of class and stratifica-tion, done entirely within the United States, thatcould have extended the scope of our knowl-edge, or the power of our interpretation, asgreatly as did the Polish and Japanese studies.

    3. W hy put the emphasis on cross-national?W hy not cross-cultural or cross-societal orcross-systemic? Doesn't the term cross-nationalascribe a greater importance to the nation-statethan it deserves? I use the term cross-nationalmainly because nation has a relatively u namb ig-uous meaning. Cross-cultural can mean any-thing from comparing subcultures within a'The issues in distinguishing cross-national researchfrom other comparative research are discussed thought-

    single nation, for example, comparing Mexican-American and Anglo-American subcultures inthe Southwest region of the United States, tocomparing very large groupings of nations thatshare broadly similar cultures, as in WilliamGoode's (1963) comparative analyses of histor-ical changes in family patterns in "the West,"Arabic Is lam, Sub-Saharan Africa, India, China,and Japan. Similarly, as Charles Tilly (1984)cogently argues, it is extremely difficult todefine what is a "society." And the termcross-systemic is so vague as to have littleresearch utility.

    I do not think that this usage of nationnecessarily implies anything about the impor-tance of nation, or the nation-state, as such, anymore than cross-cultural implies (or, at any rate,should imply) that culture is the explanatorydesideratum. Furthermore, we learn somethingabout the importance or lack of importance ofthe nation-state by discovering which processestranscend national boundaries and which pro-cesses are idiosyncratic to particular nations orto particular types of nations. In choosing whichnations to compare, sometimes we do mean tocomp are nation-states; how could Thed a Skocpol(1979) have done differently in her analyses ofrevolutions? When we deal with governments ,laws, and legally regulated institutions, thenation-state is necessarily a decisive context.But sometimes we use nation as a way ofcomparing cultures; in this case, we wouldchoose nations w ith distinctly different cultures,for example, by comparing the United States toJapan, not the Federal Republic of Germany toAustria. Sometimes we mean to comparepolitical and economic systems, as in comparingthe United States and Japan to Poland , or if onewanted to minimize cultural differences whilecontrasting political systems, in comparing theGerman Democratic Republic to the FederalRepublic of G erman y. Cross-national research isflexible, offering the advantage of makingpossible multiple types of comparison withinone general analytic framework.This flexibili ty, i t must be recognized, com esat a price: When one finds cross-nationaldifferences, it may not be clear whether thecrucial "context" that accounts for the differ-ences is nation or culture or political oreconom ic system (Scheuch 1967). Sti l l , one canat least try to assess which of these contextsmight logically be pertinent to explaining aparticular cross-national difference. And, formany types of research, one can then proceed todesign new studies to differentiate among thecontexts.

  • 7/30/2019 Kohn Melvin Cross National.pdf0

    15/26

    generalities, it is desirable to include allcountries for which pertinent data can besecured. Thus, Alex Inkeles's pioneering paper,"Industrial Man, " (1 960) gained considerab lyfrom its demonstration that the relationshipbetween social stratification and many facets ofvalues and beliefs is consistent for a wide arrayof countries. Seymo ur Martin Lipset's argumentin "Dem ocracy and Working Class Authoritarian-ism" (1959), that the working class is more"liberal" than the middle class on economicissues, but illiberal on issues of civil libertiesand civil rights, was the more forceful becausehe marshalled evidence from several countries.Donald T reiman 's (1 977) com prehensive analy-sis of the similarity of social stratificationsystems throughout the industrialized worldeffectively utilized data from many countriesand was enriched as well by information aboutthe historical past. Janet Chafetz and AnthonyDw orkin's (19 86) analysis of the size and rangeof ideologies of women's movements gainedscope and power from their use of data from aconsiderab le diversity of countries. W ith similarintent, I have searched for all extant studies toestabl ish the "universa l i ty" of a se l f -directioniconformity dimension to parental val-ues in industrialized societies (Koh n and Sch oen-bach 1980). I have also searche d for evidence instudies conducted in many countries for cross-national tests of one or another link in myexplanatory schema (Kohn 1977, 1981; Kohnand Schooler 1983, Chapter 1 2) . And , asrecently as the July 1987 issue of the AmericanJournal of Sociology, Alejandro Portes andSaskia Sassen-Koob demonstrated anew theusefulness of a broad comparative sweep, inshowing that, contrary to all theoretical belief,the "inform al," "underg round" sector of theeconomy is not merely a transitional phenome-non of Third World development, but is insteada persistent and integral part of the economies ofeven advanced capitalist nations. In doingsecondary analyses it is highly advantageous toutilize data from all countries for whichpertinent information can be secured.Moreover, even in collecting primary data,there can be considerable advantage to assessingthe consistency of findings across a range ofnations, cultures, and political systems, asInkeles and Sm ith showed in Becoming Modern(1974) and as Erik Olin Wright and hiscolleagues are demonstrating anew, in a verydifferent type of research endeavor, in theirmulti-nation studies of social class.Yet, it is expensive, difficult, and time-consuming to collect data in many countries.

    AMERICAN SOCIOLOCiICAL REVIEWcomparison. It is not necessarily true that themore nations included in the analysis, the morewe learn. There is usually a tradeoff betweennumber of countries studied and amount ofinformation obtained. In this tradeoff, investiga-tors can certainly disagree about the relativeimportance of number of countries and depth ofinformation. And the same investigator mightmake different choices for different substantiveproblems. By and large, though, I would opt forfewer countries, m ore information.My own preferred strategy is the deliberatechoice of a small num ber of nations that providemaximum leverage for testing theoretical issues.One may begin with a study in one country,with subsequent extensions of the inquiry toother countries, as my collaborators and I havedone in using Poland to learn whether U.S.findings are applicable to a socialist society andJapan to learn whether such findings apply to anon-Western, industrialized society. Alterna-tively, one can select pivotal countries thatprovide maximum opportunity to test somegeneral hypothesis, as Theda Skocpol (1979)did in selecting France, Russia, and China forher study of the causes and consequences ofsocial revolutions, or as John Walton (1984) didin selecting the Philippines, Colombia, andKenya for his compar$&e analysis of nationalrevolts in underdeveloped societies. Whetherone starts with one country and then extends theinquiry to others, or begins with a small set ofcountries, does not seem to be crucial. Eitherway, the deliberate choice of a small number ofcountries for systematic, intensive study offersmaximum leverage for testing general proposi-tions about social process.How, then, does one decide which countriesto compare? The only rule of thumb I know isthat cross-national research is most useful whenit can resolve a disputed question of interpreta-tion. It follows that what is a strategiccomparison at one stage of knowledge may beoverly cautious or overly audacious at another.At an early stage of my own research, forexam ple, w hen I had established little more thanthat white middle-class parents in Washington,DC valued self-direction for their children morehighly than did white working-class parents inthat same city at that one time, the focal issuewas Washington's atypicali ty. Was the Wash-ington finding peculiar to the times andcircumstances of this relatively affluent, econom-ically secure, mainly non-industrial city in thelate 1950s, or did that finding reflect a moregeneral relationsh ip between social stratificationand parental values? Leon ard Pearlin (1 97 1;

  • 7/30/2019 Kohn Melvin Cross National.pdf0

    16/26

    727ROSS-NATIONAL RESEARC H AS AN ANAL YTIC STRATEGYeconomically secure than Washington, and witha much less conservative working-class tradi-tion. A more cautious choice of locale wouldhave been an industrial city in the United Statesor perhaps in English-speaking Canada or inAustralia. A more audacious choice would havebeen an industrial city in a non-Western countryor in a socialist country. Turin, to my mi nd, wasneither too cautious nor too audacious a choice:different enough from Washington that if thefindings proved to be similar, the increment toour knowledge would be considerable, but notso different from W ashington that if the findingshad proved to be dissimilar, we would havebeen at a complete loss to know w hy. Tu rin wasnot the only city that could have served ourpurposes; several other West European citiesmight have served as well. In that state of ourknowledge, though, I do not think that Warsawor Tokyo would have been optimal choices. Itwould have been too difficult to interpretdissimilar findings.Later, when we had solid evidence about thegenerality of our findings in Western, capitalistsocieties, studies in Poland and Japan becameespecially useful. The issue was no longerWashi'ngton's atypicality, but whether the rela-tionships among social stratification, job condi-tions, and psychological functioning were pecu-liar to capitalist society or to Western society.Here, again, we could have chosen othercountries that might have served our purposes aswell: perhaps Hunga ry instead of Poland, or if ithad been possible to do such resea rch there atthat t ime, the Soviet Union; perhaps SouthKorea instead of Japan. It is often the case thatno one country is uniquely appropriate forcross-national comparison. Other consider-ations-research feasib ility, the availability ofpotent ia l col labora tors , funding, happen-stance-ma y then legitimately enter in.

    Were I to embark on a new comparative s tudytoday, the considerations would again bedifferent, mainly because of what we now knowfrom the Polish and Japanese studies, andbecause of new interpretive problems that havearisen from these studies. It would now beuseful to study another socialist country andanother non-Western industrialized country. Itwould also be useful to study a less than fullyindustrialized country , I think preferably (for thenonce) a capitalist country with a predominantlyWestern culture, perhaps a Latin Americancountry. The possibilities for fruitful compari-son do not shrink as one learns more, butactually grow.The choice of countries should always be

    national research will certainly require (Galtung1967, p. 440). One must always ask: If I findcross-national consistencies, will this particularcross-national comparison extend the scope ofmy interpretation enough to have made theventure worthwhile? And if I find differences,will this particular cross-national comparisonshed light on crucial interpretive problems?Cross-national research is always a gamble; onemight as well gamble where the payoff iscommensurate with the risk.*5. What are the costs of doing cross-nationalresearch? If, as I have argued throughout thisessay, the advantages of cross-national researchare considerable, so too are the costs. Thesecosts are considerably greater than most investi-gators realize, great enough to make a rationalperson think twice about doing cross-nationalresearch when it is not needed or when it ispremature.Securing funds is always problematic, even(as in my own research) when financial supportis obtained in the countries that are participatingin the research. This, however, is only the firstand by no means the most serious difficulty.Establishing collaborative relationships that canbe sustained and will develop throughout thecourse of what can be counted on to be difficultresearch is much more problematic (Hill 1962;Sarapata 198 5). Both the greatest benefits andthe most difficult problems of cross-nationalresearch come from the collaborative relation-ships. If a good collaboration is like a goodmarriage, rewarding yet difficult, then a goodcross-national collaboration is akin to a cross-cultural marriage that manages to succeeddespite the spouses living much of the time indifferent countries, sometimes with considerableuncertainty about passports, visas, and thereliability and timeliness of mail delivery, anddespite wo rking in different institutional settings

    with conflicting demands and rewards. Andstill, it 's far preferable to the alternatives. Morethan that, without good collaboration, manytypes of cross-national research are simply notpossible.The methodological pitfalls are another set ofobstacles to good cross-nation al research; I havetouched on so me of them earlier in this essay. Itwould be hard to exaggerate the amoun t of time,thought, and analysis that m ust go into the effort

    A corollary is that, if one wants to gambleaudacio usly, do so where the payoff will be considerable.A splendid example is provided by Nancy Olsen (1974).She not only extended to Taiwan the scope of our U.S.findings about the relationship between closeness of

  • 7/30/2019 Kohn Melvin Cross National.pdf0

    17/26

    AMERICAN SOCIOLOGICAL REVIEWto achieve comparability of methods, concepts,and indices. There are also issues in thestandards of research employed in differentcountries. Sometimes these issues becomeacutely problematic when one least expectsthem. As a s imple yet tel l ing example: Thereason why we do not have Polish data aboutsome of the job conditions that may be pertinentto distress is that the survey research specialistsat the Polish Academy of Sciences refused toinclude questions about job conditions that didnot meet their criteria of objectivity in a surveyfor which they were professionally responsible.Even when we appealed to them that cross-national comparability required their repeatingthe defects of the earlier U.S. study, they wouldnot yield. They were as zealous in imposingtheir justifiable, yet irrelevant professionalstandards as were the clearance officers of theU .S. Department of Health, Education, andWelfare, and of the Office of Management andthe Budget, in imposing their not nearly sojustifiable requirements.And still, there are yet more difficultproblems, problems of interpretation. Particu-larly when one finds cross-national differences,an expert knowledge of all the countries isessential-a know ledge most easily achieved, ofcourse, by collaborators who have expertknowledge of their own countries (see Kuechler1986). Even when such collaboration exists ,though, sharing knowledge, interpreting withina commo n framewo rk, even having enough timetogether to think things through at the crucialjunctures, does not com e easily.Unless one has a good reason why researchshould be cross-national, it generally isn't worththe effort of makin g it cross-national. Operation-ally, this means that one should do cross-national research either when a phenomenoncannot be studied in just one country (forexample, the causes of revolutions) or else whensome phenomenon has been well substantiatedin one country and the next logical questionshave to do with the limits of generality of whathas been learned. In principle, but rarely inpractice, i t may be worth embarking on across-national study of a less well researchedproblem if you have good a priori reason tobelieve that important theoretical issues can bemore effectively addressed by conducting theresearch in more than one country. I remain astrong proponent of cross-national research, butI would not wish to mislead anyone intothinking that its very considerab le advantages donot come at equally considerable cost.6. Finally, to return to a question that has

    method for doing cross-national research. Iregard the persistent debate about the relativemerits of historical and quantitative methods incross-national research as a wasteful distraction,addressed to a false dichotomy .9 Each m ethod isappropriate for some research purposes andinappropriate for others. Best of all, as JefferyPaige (1975) demonstrated in his analysis of therelationship between agricultural organizationand social movements in 7 0 developing nations,is to combine the two. My question concerns,not historical analysis as method, but history asexplanation. At issue, of course, are thecompeting m erits of idiographic and nomotheticexplanation. I can hardly do justice to thiscomplex question in the closing paragraphs ofthis essay, but I would at least like to point outthat the issues are somewhat different whenanalyzing cross-national similarities from whatthey are when analyzing cross-national differ-ences.As I have argued throughout this essay, theinterpretation of cross-national similarities shouldnot focus on the unique historical experiences ofeach of the countries. One seeks to discover,instead, social-structural regularities that tran-scend the many differences in history, culture,and experience that occur among nations. Thisis true even in inquiries-W alton's (1984)Reluctant Rebels is a good example-where theevidence is mainly historical but the analysissearches, not for historical idiosyncrasies, butfor historical commonalities. The intent in allanalyses of cross-national similarities is todev elop generalizations that transcen d particularhistorical experiences in a search for moregeneral explanatory principles. In short, themethod may be historical, the interpretationshould be sociological.In a broader sense of history, of course,cross-national analysis, just as any other type ofsociological analysis, cannot be ahistoric, evenwhen much about history is only implicit in theinterpretation (Sztompka 1986). To compare theimpact of social stratification on personality inthe United States and Poland, for example,assumes that we are comparing industrializedstates that have shared much of Western h istory.

    The methodological debate takes place on tw o levels :the type of analysis used within each nation and the typeof analysis used for comparing nations. I see nothing ofvalue in the first part of the de bate; one uses w hatevermethods are appropriate to the task. The second part ofthe debate deals with real issues, for example, themeaningfulness of using "samples" of nations, the utilityof statistical tests when basing one's analysis on theentire set of existing countries, and the difficulties of

  • 7/30/2019 Kohn Melvin Cross National.pdf0

    18/26

    CROSS-NATIONAL RESEARCH AS AN ANALYTIC STRATEGYThat one is a ca ~i ta l is t ta te and the other a REFERENCESsocialist state can be viewed, depending on howyou read the broad sweep of history, as acomparison of different economic-political sys-tems or as a comparison of different levels ofpolitical development. In either case, eventhough history is not treated explicitly, historicalconsiderations are certainly there implicitly.And when one compares fully industrialized topartially industrialized societies, historical is-sues are necessarily at least implicit. Neverthe-less, in interpreting cross-national similarities,history need not be at the forefront of attention.In interpreting cross-national dzfferences, bycontrast, historical considerations cannot bemerely implicit; history must come to theforefront of any interpretation. For example,after demonstrating rem arkable parallels in boththe causes and consequences of the French,Russian, and Chinese revolutions, Skocpol(1979) had to explain differen ces, particularly inrevolutionary outcomes, in terms of historicallyunique circumstances. Similarly, when I findthat social stratification and social class do nothave the same impact on the sense of distress inthe United States , Poland, and Japan, I have tolook to the separate historical developments ofthe three countries, to try to discover what mayexplain the inconsistent findings. I maintain,though, that even in interpreting cross-nationaldifferences, explanation cannot consist merelyin explicating pertinent historical differences.The object is not an understanding of historyjust for history's sake, but the use of history forunderstanding more general social processes.The interpretation m ust be historically informe d,but sociological interpretations, even of cross-national differences, are quintessentially trans-historical.EPILOGUEIn the preface to Class and Conform ity, I made adeclaration of faith: "The substance of socialscience knowledge comes from the process ofspeculation, testing, new speculation, newtesting-the continuing pro cess of using data totest ideas, developing new ideas from the data,doing new studies to test those ideas" (Kohn1969, p. xii). I take this occasion to re-affirmthis fundamental tenet of my scientific faith. Itsrelevance to this essay is, I trust, obvious: In theprocess of speculating, testing, and speculatinganew, cross-national research, properly em-ployed, provides uniquely valuable evidence.There is no other evidence so useful forconfirming social-structural interpretations, or

    Allardt, Erik and Wlodzimierz Wesolowski, eds. 1978.Social Structure and Change: Finland and Poland-Comparative Perspective. Warsaw: Polish ScientificPublishers.Almond, Gabriel A. and Sidney Verba. 1963. The CivicCulture: Political Attitudes a nd Dem ocrac j~ in FiveNations. Princeton, NJ: Princeton University Press.Armer, Michael. 1973. "Methodological Problems andPossibilities in Comparative Research. " Pp. 49-79 inComparative Social Research: Methodological Prob-lems and Strategies, edited by Michael Armer andAllen D. Grimshaw. New York: Wiley.Armer, Michael and Allen D. Grimshaw, eds. 1973.Comparative Social Research: Methodological Prob-lems and Strategies. New York: Wiley.Blumberg, Rae Lesser and Robert F. Winch. 1972."Societal Complexity and Familial Complexity: Evi-dence for the Curvilinear Hypothesis." AmericanJournal of Sociology 77:898-920.Bornschier,