Issues of Complex Hierarchical Data and Multilevel Analysis : Applications in Empirical Economics

• Main Author: Karlsson, Joel
• Format: Doctoral Thesis
• Language: English
• Published: Linnéuniversitetet, Ekonomihögskolan, ELNU 2012
• Subjects: Multilevel modelling; Hierarchical modelling; Single farm payments; Part-time unemployment; Educational attainment; Performance evaluation
• Online Access: http://urn.kb.se/resolve?urn=urn:nbn:se:lnu:diva-18569; http://nbn-resolving.de/urn:isbn:978-91-86983-64-2

This thesis consists of four individual essays and an introduction chapter. The essays are in the field of multilevel analysis of economic data. The first essay estimates capitalisation effects of farm attributes, with a particular focus on single farm payments (SFP), into the price of farms. Using a sample of Swedish farm transactions sold all across the country, the results from a spatial multiple-membership model suggests that the local effect of SFP is negative while there is a positive between-region effect of SFP, on farm prices.   The second essay investigates the extent to which differences in the probability to exit from part-time unemployment to a full-time job can be accounted for by spatial contextual factors and individual characteristics. To correctly incorporate contextual effects, a multilevel analysis was applied to explore whether contextual factors account for differences in the probability of transition to full-time employment between individuals with different characteristics. The results indicate that there is a contextual effect and that there are some spatial spill-over effects from neighbouring municipalities.   The third essay investigates the determinants of educational attainment for third-generation immigrants and natives in Sweden. Using a mixed-effects model that includes unobserved family heterogeneity, for linked register data, the main result is that the effect of parent’s educational attainment is mainly due to the between-parental education effect of family income.   The fourth and last essay presents a new robust strategy for performance evaluation in the case of panel data that is based on routinely collected variables or indicators. The suggested strategy applies a cross-classified, mixed-effect model. The strategy is implemented in two illustrative empirical examples, and the robustness is investigated in a Monte Carlo study.