National Identity and the Interplay between National Pride and Ethnic Exclusionism: The Exceptional Case of the Czech Republic

Abstract

This article focuses on national identity in post-communist Czech Republic. It empirically analyses attitudes of national pride and ethnic exclusionism and their interrelations in the years 1995 and 2003. Comparative studies dealing with national identity usually refer to the Czech Republic as an exceptional case. By focusing on the Czech case, this study contributes to a better understanding of local variations of concepts such as national pride and ethnic exclusionism. Confirmatory factor analysis is used to establish the multidimensionality of national pride and ethnic exclusionism in the Czech Republic at two time points. The interrelationships between different dimensions of national pride and of ethnic exclusionism are analysed using structural equation modelling. Data are drawn from the International Social Survey Programme National Identity Modules 1995 and 2003. Following empirical results, the paper suggests more valid dimensions for the concepts of national pride and ethnic exclusionism in the distinctive Czech context and discusses their mutual relationships.

Acknowledgements

An earlier version of this article was presented at the conference ‘Nationalism and Conflict: Interdisciplinary Methodological Approaches’, Warsaw, December 2012. The author is grateful to the participants of the conference and the two anonymous reviewers for invaluable comments and suggestions. Special thanks go to Bernhard Wessels for his general feedback on this article and Roy Kimmey for precious help with the final version.

Notes

1 Consequently, the concept of patriotism used in this article is in fact interchangeable with the terms ‘constructive patriotism’ and ‘constitutional patriotism’.

2 A detailed overview of Czech–Romani relations is provided, for example, by Fawn (2001).

3 The scoring of the answer categories has been reversed from the original version.

4 Means, standard deviations and percentage of missing values for all observed items and Cronbach's alphas for the exogenous and endogenous latent constructs are available from the author on request.

5 The Kolmogorov-Smirnov test for both years (1995 and 2003) revealed univariate non-normality—all items p < 0.001 for both years. Moreover, examination of skewness and kurtosis in 1995 indicated eight severely skewed or kurtotic items greater than |1|, with one item indicating 2.069 kurtosis. In 2003, nine skewed or kurtotic items were greater than |1|, with one item indicating 2.136. I evaluated the assumptions of multivariate normality and linearity and observed nine multivariate outliers (p < 0.001) for 1995 and 16 multivariate outliers for 2003. Thus, I removed the nine outliers from the analysis and compared the results (N = 859 for 1995) with the uncorrected data set (N = 868 for 1995). Although removing the outliers improved the model fits as expected, the goodness-of-fit statistics were not substantially better. Hence, the outliers do not alter the results dramatically and all subsequent models in the present study are estimated with the inclusion of outliers.

6 For both time points, approximately 70% (71.2% for 1995 and 72.3% for 2003) of the respondents provided valid responses to all items.

7 The MLR chi-square test statistic is asymptotically equivalent to the Yuan–Bentler χ² test statistic (Yuan & Bentler, 2000).

8 The final CFA model was also estimated with the robust weighted least squares (WLSMV) estimator in order to address the categorical (ordinal) nature of the data and goodness-of-fit indices were compared with those estimated by the MLR estimator. The χ² statistics and other fit indices were not substantially different, hence indicating that the number of categories is large enough so that the failure to address the ordinality of the data is probably negligible (Byrne, 2012, p. 129).

9 Chi-square difference tests between nested models are inappropriate when using the MLR estimators, because the chi-square-difference values are not distributed as chi-square (Muthén & Muthén, 1998–2007; Byrne, 2012). For the MLR estimators, χ² difference testing must be done using the scaling correction factor (Satorra & Bentler, 2001). Although χ² statistics are reported here, the χ² high sensitivity to sample size and its consequent problematic nature should be kept in mind (de Figueiro & Elkins, 2003; Byrne, 2012). Therefore, it is often preferred to evaluate model fit based on other fit statistics.

10 The correlations between covariates and the exogenous variables are not shown here, for space restrictions, but are available from the author on request.