Replication Data for: Využití víceúrovňových modelů při analýze kontextuálních efektů míry ekonomické aktivity na podporu přerozdělování v komparativních longitudinálních datech

This article studies the links between a country’s labour force participation rate and attitudes towards income redistribution. The article also demonstrates how to specify a multilevel model when analysing contextual effects and it presents several types of random effects structures and options for centering explanatory variables in comparative longitudinal survey data. The contextual effect is decomposed into longitudinal and cross-sectional components for time-varying contextual variables, such as the labour force participation rate. The analysis of redistribution support based on ESS data from 27 countries and nine rounds shows how fundamentally the mentioned properties can influence substantive conclusions. The analyses presented in this article do not provide any evidence for a link between redistribution support and the labour force participation rate. However, the hypothetical configurations of multilevel models presented here cover all possible substantive effects of the labour force participation rate. Contextual effects analysis may thus lead to highly unreliable results when a multilevel model fails to control for the compositional effects of individual-level predictors, when it does not specify random effects at the level to which a substantial variation of the outcome variable may be attributed, and when it does not distinguish between the longitudinal and cross-sectional effects of time-varying variables.

本文研究了国家劳动力参与率（labour force participation rate）与收入再分配态度之间的关联。本文还展示了在分析情境效应时如何设定多层次模型（multilevel model），并介绍了比较纵向调查数据中随机效应结构的几种类型以及解释变量中心化的可选方案。对于劳动力参与率等时变情境变量，情境效应被分解为纵向和横截面两个部分。基于27个国家9轮欧洲社会调查（ESS）数据的再分配支持分析表明，上述特性对实质性结论的影响程度之深。本文的分析未发现再分配支持与劳动力参与率之间存在关联的任何证据。然而，本文提出的多层次模型假设配置覆盖了劳动力参与率所有可能的实质性效应。因此，当多层次模型未能控制个体层面预测变量的组成效应、未在结果变量存在显著变异的层面设定随机效应、或未区分时变变量的纵向与横截面效应时，情境效应分析可能会得出高度不可靠的结果。