The methods used in this thesis are multilevel logistic regression (Chapter 2) and multinomial logistic regression (Chapters 3, 4, and5) models. We work with different units of analysis: older person-geographically close child dyads (Chapter 2), older person-distant child dyad-years (Chapter 3), older person and all their children as a group-years (Chapter 4), and older person-distant sibling dyad-years (Chapter 5). In Chapters 2, 3, and 5, the dyadic perspective was chosen because it enables one to take account of characteristics of the older person and each family member of interest and, at the same time, include information on the family group. To avoid double counting and correlated outcomes between partners, we run separate models for older women and older men (Chapters 2 -4). An important feature that is taken into account is that, in all the chapters, the units of analysis are clustered within dyads, within older persons, or both, causing the standard assumption of independence of observations to be violated.Additionally, in Chapter 2, and in the sensitivity analysis of Chapter 4, we treat the municipality as an extra level on which differences in migration, immobility, or institutionalization can occur. To adequately account for the data hierarchy, we estimate multilevel models or one-level models with multiway clustering of standard errors.
AcknowledgementsThis research has been conducted within the research project "Family ties that bind: A new view of internal migration, immobility, and labor-market outcomes." The Family Ties project, led by Clara H. Mulder, has received funding from the European Research Council (ERC) under the European Union Horizon 2020 research and innovation program (grant agreement No. 740113). This thesis contributes to two objectives of the project: (i) identifying the role of family ties as a deterrent to migration and key determinant of immobility and (ii) explaining migration toward family in relation to migration in other directions. References 24 Collaborations with researchers from the "Geographic distribution of health care" project (GeoHealth) and the Aging Well research program have been initiated in the process of working on this thesis. Data obtained for the GeoHealth project from Statistics Norway have been essential for the analyses presented in Chapter 2. The project received funding from the Norwegian Research Council (grant agreement No. 256678). The GeoHealth project explores the causes and consequences of substantial regional differences in the utilization of health care. Data obtained for the Aging Well research program from Statistics Sweden have been essential for the analyses presented in Chapters 3 to 5. The program has received funding from the Swedish Research Council for Health, Working Life and Welfare (grant agreement No. 2016-07115). The overarching aim of the Aging Well program is to address the shifting dependency threshold and its implications for the burden of care associated with individual and societal aging. Chapters 3 to 5 link to two Aging ...