Multiple Units of Analysis in Cross-National Research

THE IMPORTANCE OF CONTEXTS: MULTIPLE UNITS OF ANALYSIS IN CROSS-NATIONAL RESEARCH

ROBERT M. KUNOVICH

Download PDF Article of Kunovich

Abstract

Although researchers often treat countries as a unit of analysis in multilevel studies of ethnic, racial, and national attitudes, there has been some debate about which is the appropriate unit of analysis in such studies. Rather than limiting ourselves to one unit of analysis, we should systematically examine and compare multiple units of analysis. In this essay I offer two recommendations for improving cross-national research in nationalism, conflict, and identity. First, researchers should measure contextual variables for multiple geographic regions. Second, researchers should consider non-geographic contexts, such as occupations, in multilevel analyses. For each theoretically relevant concept, scholars should ask: For which units of analysis is it possible to measure this concept and what are the strengths and weaknesses of selecting each?

Keywords: cross-national research, geographic data, multiple units of analysis

Full Text of Article

Cross-national research on nationalism, conflict, and identity has grown considerably over the years. One area of enduring interest is explaining geographic differences in attitudes – for example, explaining why attitudes toward immigrants are more negative in some places compared to others. Much of the research in this area seeks to establish an association between negative attitudes with labor market competition and poor economic conditions (e.g., Quillian, 1995). Other contextual conditions are also associated with people’s attitudes toward immigrants: ethnic heterogeneity and social security benefits expenditure (Coenders et al., 2004); percentage of the extreme right-wing vote (Semyonov et al. 2006, 2008; but see Wilkes, Guppy, and Farris, 2007); restrictionist immigration policies and emphasis on national culture (Hjerm, 2007); and religious diversity and the minority’s acquisition of majority language (Evans and Need, 2002).

Although researchers often treat countries as a unit of analysis in multilevel studies of ethnic, racial, and national attitudes, there has been some debate about which is the appropriate unit of analysis in such studies. Quillian (1995:592) argues that countries are an appropriate unit of analysis compared to cities or smaller regions because comparable data are more readily available for countries, because they are “important cultural, political, and economic units”, and because of their ties to national identity and threat perceptions. Fossett and Kiecolt (1989) argue that metropolitan areas are appropriate because metropolitan areas define people’s space in daily life – including intergroup competition for resources. Oliver and Mendelberg (2000:577), by contrast, argue that regions within metropolitan areas are better because they contain less internal variability, more external variability, and better describe one’s immediate racial context (see also Ha 2010).

There are costs and benefits associated within any of these choices. Rather than limiting ourselves to one unit of analysis, however, we should systematically examine and compare multiple units of analysis. Doing so will allow us to determine the appropriate level of analysis for different variables. Immigration policy and social security benefits expenditures, for example, should probably be measured at the country level. Labor market competition, though, may be better captured at the metropolitan level due to regional variation within countries in factors such as the size of foreign-born groups and economic conditions. Housing market competition may be better measured at the neighborhood level due to variation in desirability within metropolitan areas. This does not mean that we should begin to estimate complex multilevel models with respondents simultaneously nested within three or four geographic regions; such data may not meet the basic assumptions of multilevel modeling. It is possible to proceed in a more cautious way. Meuleman (2011), for example, relies on correlations and scatterplots to examine relationships between group size, economic conditions, and aggregated anti-immigrant attitudes after first establishing measurement equivalence.

It is also surprising that few have questioned whether it might be possible to measure critical concepts, such as labor market competition, in non-geographic contexts (see Kunovich forthcoming; Ortega and Polavieja 2009 for exceptions). One such possibility is occupation. Labor markets are clearly influenced by physical space. There is a distinction, however, between labor markets, which are associated with specific occupations, and labor market areas, which are associated with geographic regions. Focusing only on labor market areas is problematic because the labor markets for occupations are segmented. A variety of factors limit competition between people in different occupations within the same labor market area (e.g., skill requirements, training costs, credentialing, occupational segregation, and internal labor markets within firms). Rather than measuring labor market competition only on geographic regions, we should begin to develop and test models with occupations as units of analysis. There are many sources of occupation level data (e.g., micro census data and bureau’s of labor statistics) and there are many interesting occupation-level variables that could be included (e.g., the percentage of foreign-born workers, wage inequality between native and foreign-born workers, occupational self-direction, and occupation-specific unemployment). Occupations are not meant to replace geographic regions as units of analysis in contextual studies, but they should be included along with geographic data. Doing so opens up the possibility of exploring interactions between occupation and region characteristics. Perhaps, for example, the presence of a large foreign-born population in a region has a negligible impact on people’s attitudes who are working in an occupation with low unemployment.

In sum, I offer two recommendations for improving cross-national research in nationalism, conflict, and identity. First, researchers should measure contextual variables for multiple geographic regions. Second, researchers should consider non-geographic contexts, such as occupations, in multilevel analyses. For each theoretically relevant concept, scholars should ask: For which units of analysis is it possible to measure this concept and what are the strengths and weaknesses of selecting each? By returning to fundamental questions of measurement and design, we have the opportunity to improve our understanding of the sources of nationalism and conflict.

REFERENCES

Coenders, M., Gijsberts, M. and Scheepers, P. (2004). Resistance to the Presence of Immigrants and Refugees in 22 Countries. In M. Gijsberts, L. Hagendoorn, and P. Scheepers (eds.) Nationalism and Exclusion of Migrants: Cross-National Comparisons, Aldershot: Ashgate, pp. 97-120.

Evans, G. and Need, A. 2002. Explaining ethnic polarization over attitudes towards minority rights in Eastern Europe: a multilevel analysis. Social Science Research 31:653-80.

Fossett, M.A. and Kiecolt, K.J. (1989). The Relative Size of Minority Populations and White Racial Attitudes. Social Science Quarterly, 70:820-35.

Ha, Shang E. (2010). The Consequences of Multiracial Contexts on Public Attitudes toward Immigration. Political Research Quarterly, 63:29-42.

Hjerm, M. (2007). Do Numbers Really Count? Group Threat Theory Revisited. Journal of Ethnic and Migration Studies, 33:1253-75.

Kunovich, R.M. (Forthcoming). “Occupational Context and Anti-immigrant Prejudice.” International Migration Review.

Meuleman, B. (2011). Perceived Economic Threat and Anti-Immigration Attitudes: Effects of Immigrant Group Size and Economic Conditions Revisited. in Cross-Cultural Analysis: Methods and Applications, edited by Eldad Davidov, Peter Schmidt, and Jaak Billiet. New York: Routledge.

Advertisements