The relationships between individualism nationalism ethnocentrism and authoritarianism

The relationships between individualism (I), nationalism (N), ethnocentrism (E) and authoritarianism (A) have been discussed amongst others in the political, philosophical and sociological literature. However, empirical analyses of their interdependencies are still scarce. That is why the main purpose of this thesis is to analyse empirically these interdependencies on the basis of the General Election Study for Belgium in 1991, 1995 and 1999.

Chapter 2 provides essential background information on Belgium, the country to which the case study relates. Belgium became an independent country in 1830. It is located in the mid-western part of the European continent; it consists of the three federal regions: Flanders, Wallonia and Brussels. The reforms of the 1970s and afterwards gradually transformed Belgium into a federal state, giving the majority of essential governmental powers to the three regions.

Each region is divided into provinces which in their turn are divided into municipalities. In Flanders, most of the people (known as Flemish) speak Dutch; in Wallonia, most of the people (known as Walloons) speak French. In Brussels, both French and Dutch are official languages. Along the eastern border, German is the official language of a small minority. There are also three cultural communities: the Flemish, the French and the German-speaking community. The communities have powers in areas where public services are highly dependent on language use, such as education, health and culture. The communities and regions each have their own Parliaments and their own Governments. Each region has a great deal of autonomy but frictions about language, ethnicity, and national identity between Flemings and Walloons continues to the present day, especially in the Brussels-Halle-Vilvoorde region.

The voting right in Belgium is a “one man, one vote” system: every Belgian national, male or female, who has reached the age of 18 has the right, and is obliged to cast one vote (unless this right has been suspended or the individual is ineligible) in the elections at the six different levels.

Today, there basically are no longer national parties in Belgium, except for some small unionist parties. All parties are homogeneous Flemish or Francophone and are present either in the Flemish or in the French-speaking constituencies, or else in the undivided bilingual electoral district of Brussels-Halle-Vilvoorde. The Belgian multi-party system usually leads to a coalition government.

In a cross-sectional study, Billiet (1995) found a relationship between individualism and ethnocentrism among Flemish Roman Catholics while Meloen (1996) empirically addressed the issues of authoritarianism and modern political racism in a survey of 900 Flemish high school students. In a sample of adults, also administered in Flanders, Van Hiel and Mervielde (2005) found that right wing authoritarianism was positively related to prejudice. In a longitudinal study Billiet et al. (2005) found a moderate and rather constant across time correlation between nationalism and ethnocentrismin in Flanders. The author did not address the question whether nationalism leads to ethnocentrism, ethnocentrism to nationalism, or both effects operate simultaneously in a reciprocal causal relationship.

The background and rationale of the above studies was the voting behavior in Flanders, particularly the support for the extreme right-wing party Vlaams Blok (Fraeys, 2004). Nationalism, authoritarianism, and political protests are all supposed to play an important role in the support for the Vlaams Blok. The main problem in the ideology of the Vlaams Blok is the choice for an ethnic national state, where the ‘state’ is understood as an ‘ethnic community which is biologically determined’ (De Witte & Klandermans, 2000). The Vlaams Blok has been convicted for racism by a Belgian court in 2004.

The present study is inspired by, and extends, the research on the development of political ideologies in Flanders. Particularly, it presents longitudinal analyses of the relationships between individualism, ethnocentrism, nationalism and authoritarianism.

Chapter 3 presents the conceptual model of interdependencies between the four key concepts of individualism, ethnocentrism, nationalism, and authoritarianism on the basis of a literature review. I define the notions of individualism, ethnocentrsim, nationalism and authoritarianism as follows:

Individualism: the pursuit of personal happiness. According to this political ideology the core task of a community or state is to foster the rights, and improve the development, of individuals and to assure their freedom. Community and state are seen as devices to individuals to achieve those objectives. The community exists for the sake of its individual members. Individualism implies that the government should not unduly intervene in individuals’ lives. Instead, it should guarantee that individuals do not harm each others’ interests.

Ethnocentrism: a belief in the superiority of one’s own group and a corresponding disdain for other groups. Ethnocentrism implies a strong distinction between “ingroups” (groups with which the individual identifies him or herself), and “outgroups” (typically minority groups, toward which he or she has no sense of belonging or which are perceived as antithetical to the ingroup).

Nationalism: an ideology, a sentiment, a form of culture, or a social movement that focuses on the nation. As an ideology, nationalism holds that ‘the people’ are the nation, and, that as a result, only nation-states founded on the principle of national self-determination are legitimate. In many cases nationalist pursuit of self-determination has caused conflict between people and states including war (both external and domestic), secession; and in extreme cases, genocide.

Authoritarianism: a political philosophy that negates democracy and an ideology that accepts a political system that is not based on the consent of the governed but on the will of the rulers. Moreover, it accepts a monopoly of power, and discussion and voting are replaced by the decisions of leaders.

The postulated relationships among the above concepts are depicted in Figure 7.1. Specifically, the hypotheses are:

Individualism has a negative effect on authoritarianism.

Authoritarianism has a positive effect on ethnocentrism.

Nationalism has a positive effect on authoritarianism.

Individualism has a negative impact on nationalism.

Ethnocentrism has a positive impact on nationalism.

Figure 7.1. The postulated recursive structure between latent state variables

7.3 Methodology and data set

In this section, I explain why I have used continuous time modeling (CT) and structural equation modeling (SEM) to estimate the parameters of the CT model. At the end of the section I describe the data set.

Most studies of the interdependencies among individualism, nationalism, ethnocentrism and authoritarianism (or of a subset of these variables) are cross-sectional, e.g. Billiet et al (2005), Fraeys (2004), Spruyt (1995) and Billiet (1995). However, cross-sectional research has a disadvantage in that there is no control over the autoregression effects and the directions of cross-effects between the variables are difficult to assess. Longitudinal research based on repeated measurements of the same variables at different points in time make it possible to reduce or overcome these problems.

The arguments for continuous time modeling have been summarized by amongst others Bergstrom (1988). The rationale of continuous time modeling is that a social or political system does not only function at quarterly or annual observation points in time, but also during intermediate intervals. Hence, the model should also relate to the intermediate intervals. Gandolfo (1993) added that the results of a model should not depend on the length of the observation interval and must remain the same when the interval is doubled or halved. If the results should not depend on the period length, he concluded, they should remain valid when this length tends to zero (that is, when one switches over from discrete to continuous time analysis). According to Oud (2007), the most compelling reason for analyzing cross-effects in continuous time is that equal effects found in discrete time do not guarantee at all that the underlying continuous time effects are equal. Particularly, equality at a single point in time may be consistent with quite different cross-lagged effect functions across time. For example, the cross-lagged effect functions of a pair of reciprocal effects, say indivdualism on ethnocentrism and vice versa, although having equal values at one specific point in time, may have quite different forms and maxima across time.

Continuous time models are estimated on the basis of observations in discrete time (in this study the General Election Studies in 1991,1995 and 1999). Estimation requires a tool that links the discrete observations to the continuous time model. One possible tool is the approximate discrete model (ADM). An advantage of the ADM is that it utilizes only simple linear restrictions to approximate the differential equation model and allows estimation by means less nonlinearly oriented SEM programs like LISREL.

An alternative to the ADM is the Exact Discrete Model (EDM). The EDM links in an exact way the discrete time model parameters to the underlying continuous time model parameters by means of nonlinear restrictions (Bergstrom, 1988),. Oud and Jansen (2000) showed how the nonlinear SEM program Mx (Neale, et al., 1999) can be employed for maximum likelihood estimation of the continuous time state space model parameters. Oud and Jansen (2000) also generalized the EDM to cover not only time-invariant parameters, but also the parameters that vary continuously over time according to a general polynomial scheme.

The data set in this thesis is obtained from the General Election Study for Belgium in 1991, 1995 and 1999. The data set contains two types of respondents, Flemish respondents and Dutch speaking respondents of the Brussels-Capital Region. The sample was selected as a two stage sample with equal probabilities of the secondary units. The total sample size avaiable for all three waves is 1274. The geographical distribution of the respondents is given in Table 7.1.

Table 7.1 Geographical distribution of panel respondents in Flanders and Brussels

interviewed in 1991, 1995 and 1999

Province
Respondents

Antwerp

331

Flemish Brabant

187

Limburg

187

East Flanders

311

West Flanders

223

Brussels

35

Total

1274

(Source: Interuniversitair Steunpunt Politieke-Opinieonderzoek K.U. Leuven, General Election Study 1991, 1995 and 1999)

7.4 The main empirical results

The first empirical chapter 4 Measuring authoritarianism with different sets of items in a longitudinal study deals with measurement of authoritarianism. As defined in chapter 3, authoritarianism is a form of social behavior characterized by strict obedience to the authority of a state or organization and adherence to enforcing and maintaining control through the use of oppressive measures. It refers to a complex of nine sub-syndromes (Adorno et al., 1950), of which conventionalism (strict adherence to conventional values), aggression and submission are the most important (Meloen, 1991).

The sets of items in the data set at the three time points (1991, 1995 and 1999) were not the same. In total 12 different items were used over the three waves. The purpose of chapter 4 was the identification of the items that consistently and adequately measure authoritarianism over time. The items 1 and 2 are the core items and were used in all threee waves (1991 C 1995 C 1999). These items are: “Obedience and respect for authority are the two most important virtues children have to learn”, and “Most of our social problems could be solved, if we could somehow get rid of the immoral, crooked people”. Items 3-6 and item 9 were used only twice (in 1991 C 1995 and in 1995 C 1999, respectively); the remaining five items only once (in 1991 or 1999).

I applied the congeneric measurement model by Joreskog (1971, 1974). Congenericness between items means that their underlying latent variables have a correlation coefficient approximately equal to 1 and thus can be considered to measure the same underlying phenomenon. I found that Joreskog’s model performed well. The main empirical result was that the core items measure authoritarianism well in all three waves

The second empirical chapter 5 is Assessing the relationships between nationalism, ethnocentrism, and individualism in Flanders using Bergstrom’s approximate discrete model. The reciprocal relationships between the three concepts individualism nationalism, and ethnocentrism was analyzed by a cross-lagged panel model. I hypothesized strong autoregressions for individualism, nationalism and ethnocentrism and, on the basis of the theoretical considerations in Billiet (1995) and Billiet et al. (2005) and the conceptual model presented above, the following causal cross-lagged structure:

A negative impact of individualism and a positive impact of nationalism on ethnocentrism

A negative impact of individualism on nationalism.

The recursive cross-lagged structure is summarized in Figure 7.2.

Figure 7.2 The recursive cross-lagged structure among individualism, nationalism and ethnocentrism

As a starting point I did not hypothesize reciprocal cross-lagged effects between the variables.

Individualism was measured by five items with 5-point-scales; ethnocentrism by eight items with 5-point-scales, and the third latent variable, nationalism, was measured by four items in a somewhat more complicated fashion.

Estimation was done by means of the LISREL program by estimating the approximate discrete model (ADM), from which the exact discrete model (EDM) was derived and used in subsequent computations. For the measurement model, all loadings turned out to be highly significant, indicating that every item contributes to the latent variable. Moreover, the reliabilities as measured by R2 ranged from 0.230 to 0.670.

The autoregressive effects for all three variables turned out to be rather strong, as hypothesized.

Regarding cross-effects, there are substantial differences between Figures 7.2 and 7.3, particularly:

The hypothesized negative relationships from individualism on nationalism and ethnocemtrism turned out to be positive. A possible explanation is that by changes in society individualism got a less liberal character and developed into a more nationalistic and ethnocentric direction.

There is a reciprocal relationship between individualism and ethnocentrism whereas we hypothesized a unidirectional relationship from individualism to ethnocentrism. A possible explanation is that the growing ethnocentrism in the society of Flanders stimulated the less liberal kind of individualism mentioned above.

Figure 7.3 The estimated relationships between individualism, ethnocentrism, and nationalism

Furthermore, both individualism and ethnocentrism have small effects on nationalism. Nationalism was found to be dependent only with no significant effect on the two other latent variables.

Standardized cross-lagged effect functions (unit-impulse responses) revealed the maximum impact of individualism on ethnocentrism (0.235) to occur after 17 years and the effect in the opposite direction (0.190) after 16.4 years. The smaller maximum impacts of individualism on nationalism (0.105) and ethnocentrism (0.099) are expected to occur later, after 22 and 23.2 years, respectively.

Chapter 6 The relationships between individualism, nationalism,

ethnocentrism, and authoritarianism in Flanders by means of

the continuous time EDM/SEM model extends the analysis presented in Chapter 5 to all four key variables presented in the conceptual model. The EDM/SEM is estimated by the Mx program with all four concepts handled as latent state variables that influence each other continuously across time. Although nationalism, ethnocentrism, individualism, and authoritarianism in Flanders have been the subject of several studies before, a longitudinal analysis has not been performed on all four concepts simultaneously nor have their relationships and the direction of their relationships been studied in continuous time.

Figure 7.4 The hypothesized relationships between individualism, nationalism, ethocentrism and authoritarianism

The basic hypotheses in Figure 7.4 are:

Individualism has a negative effect on authoritarianism.

Authoritarianism has a positive effect on ethnocentrism.

Nationalism has a positive effect on authoritarianism.

One important result of the chapter is that the SEM model is identified with three time points.

The estimated measurement model showed that the latent variables are well measured. The structural model is presented in Figure 7.5:

I

Individualism

A

Authoritarianism

0,0319

(0,0065)

0,0357

(0,0093)

0,0386

(0,0062)

E

Ethnocentrism

N

Nationalism

Figure 7.5 The estimated relationships between nationalism, individualism, ethnocentrism and authoritarianism

As in chapter 5, we find substantial differences the hypothesized model (7.4) and the empirical findings (7.5). Particularly:

Nationalism is an “isolated” variable in that it neither has an impact on any of the other variables in the model nor is impacted by any of them. This is a consequence of the introduction in the model of the fourth concept authoritarianism. Apparently, authoritarianism is the the key variable in the model. It impacts on individualism and ethnocentrism and, in its turn, is impacted by ethnocentrism. In the present-day Flemish context, the traditional notions of nationalism, individualism, ethnocentrism and authoritarianism and their relationships do no longer apply. Tradtional nationalism does not fit into this model any longer

In the conceptual model we specified a relationship from authoritarianism to

ethnocentrism. The empirical results confirm this relationship but also reveal an even stronger relationship from

ethnocentrism to authoritarianism. Apparently, ethnocentrism is the key driver in the model .

7.5 Reflections on the methodology and the implications of the empirical results

The empirical analysis has dealt with four latent variables, each measured by different sets of indicators. For one variable, authoritarianism, I used Joreskog’s (1971, 1974) congeneric model to identify the indicators that measure this variable consistently over time. I found that this model performs well to test whether or not different items used in a longitudinal analysis can be used to measure the same underlying latent variable.

The next important methodological result is that continuous time modeling is appropriate to analyze the development of the interdependencies among individualism, ethnocentrism, nationalism and authoritarianism over time. Continuous time makes it possible to fill the gaps between the discrete time points and evaluates the auto-effects and cross-effects of the variables for intermediate time intervals. It thus allows comparing effects for unequal time intervals between waves, obtained in the same study or in different studies. The reason is that the autoregressions as well as the cross-effects do not depend on the time interval used . In particular, in a stable model autoregression goes down for an increasing time interval, while cross-effects first increase and next go down (Oud, 2002).

A third methodological finding is that a model with four latent variables observed at three points in time is identified.

A fourth major methodological finding relates to the performance of structural equation modeling (SEM). A SEM is made up of two submodels: a measurement model where the relations between the observed and the latent variables are specified and estimated; and a structural model that presents the relationships among the latent variables. In SEM, we can use cross-sectional and longitudinal data. The present study has shown the advantages of SEM in longitudinal analysis, particularly adequate estimation of autoregressions and of cross-lagged effects controlling for autoregressions.

The main result as regards content is that we have gained more insight into the relationships among individualism, ethnocentrism, nationalism and authoritarianism. Particularly:

Compared to Billiet (1995), a major achievement of this thesis is the finding of a reciprocal relationship between the authoritarianism and ethnocentrism and not just a one-way relationship. This is due to the fact that we used a longitudinal analysis while Billiet only used a cross-sectional analysis..

Regarding the relationship between individualism and authoritarianism, in this study we found a positive impact from authoritarianism to individualism instead of a negative effect from individualism to authoritarianism, as hypothesized on the basis of theory. In line with Billiet, et al (1996), we observe that the traditional hypothesis of a negative effect from individualism to authoritarianism does not fit anymore the present Flanders situation where many guestworkers are seen as a threat to the lower class and the bad economic situation in Wallonia is considered a threat for the people in Flanders who are worried that a substantial amount of social security goes to Wallonia. These developments have created a kind of individualistic selfishness which is not captured by the traditional theory of individualism. In this regard it would be interesting to analyze the relationship between individualism and authoritarianism by social class, as the perceptions of the future of social security may differ by social class.

Regarding the relationship between authoritarianism and ethnocentrism, we did not expect a reciprocal relationship. Our original hypothesis is confirmed but there is also another relationship: from ethnocentrism to authoritarianism. In the original grand theory by Adorno, et al (1950), there was the hypothesis that authoritarianism as character structure was the “mother” of many attitude configurations as for example anti-jewishm, ethnocentrism, conservative thingking about the economy and labour relation. In present-day Flandres it is ethnocentrism that drives, and in its turn is driven by authoritarianism.

Regarding nationalism and authoritarianism, the empirical results showed no

relationship between both variables while we postulated that nationalism has a positive effect on authoritarianism. Apparently, the notions of nationalism and authoritarianism are less closely linked in present-day Flandres than hypothesized in traditional theory.

The following policy recommendations can be derived from the empirical results:

The results can be used in educational and information programs. Specifically, when adressing the issue of ethnocentrism it is important to consider the entire complex of relationships in which it is embedded rather than considering it in isolation. So if a program aimed at the reduction ethnocentrism is entertained, individualism (in the sense of egocentrism) and authoritarianism need to be considered as well.

The positive relationship between individualism and ethnocentrism found in chapter 5 should be taken into account in the context of privatization programs which have taken place in Europe and elsewhere. Privatization and the accompanying philosophy of individualism may undermine the sense of Gemeinschaft which may lead to ethnocentrism and ultimately to racism. Of course, further research on this issue is needed

In the context of unemployment policies and reform of social security the wider political impacts on ethnocentrism and authoritarianism should be taken into account. For instance, a reduction of unemployment may go together with less ethnocentrism.

7.6 Some suggestions for further research

In this study I used maximum likelihood (ML) to estimate the parameters. However, because some degree of nonnormality was found in the data, instead of ML alternative estimators like weighted least squares (WLS) or diagonally weighted least squares (DWLS) procedures should be tested.

In this study I have applied the ADM/SEM and EDM/SEM procedures. However, there are two alternatives to solve the continuous time estimation problem. One alternative is the linear stochastic differential equation approach (LSDE) (Singer, 1991) The LSDE approach involves repeated calculation of the latent state vector for all subjects whereas in the EDM-Mx procedure, the latent state vector is derived on the basis of the loglikelihood function. It seems worthwhile to compare these different approachs in terms of estimation results. Another alternative is the multivariate latent differential equation (MLDE) (Boker et al.,2004). The degree of similarity between ADM/SEM and MLDE is dependent on the actual time interval between measurements. An interesting question is how similar the results are for intervals as in the present study.

Extension of the four variables model to include racism. As described above, ethnocentrism is a very broad variable and in the extreme it goes to racism. By extending the model the similarities and differences between ethnocentrism and ethnocentrism can be identified as well as their relationships to the other variables in the model.

This study is based on three waves. When new waves become available, it is worthwhile expanding the analysis by adding the new observations (e.g for the years 2004 and 2009). If we have more time point, then the precision of the analysis can be increased.

Reconsideration of the notions of individualism, nationalism, ethnocentrism and authoritarianism and their dependencies in the light of the empirical findings for Flanders. In this context it is also important to consider disaggregation of the analysis by social class. It could be that the relationships between variables are different for different classes and, moreover, that they vary over time.