Comparing Ordinary Multilevel Modeling with Multilevel Structural Equation

Comparing (Ordinary) Multilevel Modeling with Multilevel Structural Equation Modeling Michael Braun GESIS – Leibniz Institute for the Social Sciences, Mannheim

Overview § § The substantive research question Two statistical models Results Conclusions

The substantive research question § anomalies found for item on civil disobedience in ISSP 2004 battery on rights in a democracy § in one group of countries, values for civil disobedience markedly lower than for index of rights in a democracy, excluding this item § suspicion this might be related to satisfaction with democracy (then civil disobedience not necessary or even harmful)

There are different opinions about people’s rights in a democracy. On a scale of 1 to 7, where 1 is not at all important and 7 is very important, how important is it … a. that all citizens have an adequate standard of living b. that government authorities respect and protect the rights of minorities c. that government authorities treat everybody equally regardless of their position in society d. that politicians take into account the views of citizens before making decisions e. that people be given more opportunities to participate in public decision-making f. that citizens may engage in acts of civil disobedience when they oppose government actions

Civil disobedience item vs. index (items a-e) Civil Disobedience 3. 8 Index (without civ. dis. ) 6. 2 Denmark 4. 1 6. 4 United States 4. 0 6. 4 Western Germany 5. 3 6. 3 Eastern Germany 5. 5 6. 4 Hungary 5. 6 6. 3 Spain 5. 1 6. 6 Country Canada

Two statistical models § (Ordinary) 3 -level Multilevel Modeling § Multilevel Structural Equation Model (ML-SEM) allowing for direct effects of satisfaction with democracy as covariate on items, given the common factor

3 -level multilevel model § dependent variable: evaluation of any of the six items on rights in a democracy. § that is: in addition to respondent and country level, there is an item level, which constitutes the lowest level.

3 -level multilevel model Country 1 Respondent 1 Standard of living Rights of minorities Country 2 Respondent 2 … Equality of treatment Responsiveness Respondent 1 Participation Respondent 2 Civil disobedience … …

3 -level multilevel model The following variables are included : § Indicators for the second to sixth item (main effects on the item level) § Satisfaction with democracy both on the individual and on the country level (main effects on the individual and country level, respectively) § Interactions of the indicators for the second to sixth item with satisfaction with democracy – both on the individual and country level (crosslevel interactions)

3 -level multilevel model Main effects rights of minorities -9. 6 equality of treatment . 08 responsiveness -. 06 participation 2. 2 civil disobedience 23. 6 Satisfaction with democracy (individual level) 6. 3 Satisfaction with democracy (country level) -2. 1

3 -level multilevel model Interactions with satisfaction (individual level) rights of minorities 3. 9 equality of treatment 3. 5 responsiveness -4. 8 participation -7. 8 civil disobedience -22. 2

3 -level multilevel model Interactions with satisfaction (country level) rights of minorities 1. 6 equality of treatment -. 7 responsiveness 2. 4 participation -5. 2 civil disobedience -45. 2 Constant 6. 69 Country-level variance . 035 Respondent-level variance . 471 Response-level variance 1. 169

3 -level multilevel model Main results: § [from model with only item identifiers: ]Civil disobedience has high item difficulty. § The higher satisfaction with democracy on the individual level the higher the support for rights in democracy in general. § The higher satisfaction with democracy on the country level the lower the support for rights in democracy in general (borderline effect).

3 -level multilevel model § The higher satisfaction with democracy on the individual level the lower agreement with civil disobedience. § The higher satisfaction with democracy on the country level the lower agreement with civil disobedience.

Multilevel SEM The following variables and relationships are included : § the six indicators as measures of common factors on the individual and on the country level § regressions of the factors (on both the individual and on the country level) on satisfaction with democracy on the individual and country level, respectively

Multilevel SEM § regressions of the second to sixth indicator of the common factors on satisfaction with democracy both on the individual and on the country level

Multilevel SEM Rights -0. 06 1. 0* Standard liv. 1. 29* R. minorities 0. 05 0. 89* 0. 66* Equality of t. Responsiv. -0. 00 Between level 0. 60* Participation -0. 00 -0. 09* 1. 17* 1. 20* 0. 18 Ci vil disobed. -0. 49* Satisfaction d. Rights Within level 0. 02 1. 0* Standard liv. 1. 18* R. minorities 0. 01 1. 13* Equality of t. Responsiveness 0. 01* -0. 02* Satisfaction d. Participation -0. 03* 0. 89* Civil disobedience -0. 07*

Multilevel SEM § loading of civil disobedience on the factors suspicious, particularly on the country level § regression of the common factors on the individual and the country level not significant § small negative direct effect of satisfaction with democracy on civil disobedience (given the common factor) on the individual and stronger negative direct effect on the country level

Conclusions § conclusions similar with regard to mean of civil disobedience, with regard to interaction with satisfaction with democracy § loading of civil disobedience on the individual level not really suspicious in multilevel CFA § positive effect of satisfaction with democracy on the individual and negative on the country level on attitudes towards rights in a democracy not significant in multilevel CFA

Conclusions Question: Which modeling strategy is best?

Thank you!