Mittwoch, 25. Mai 2016

How Does Widowhood Affect Wellbeing? Results From 14 European Countries and Israel (Propensity Score Matching)

Maja Adena (WZB), Michal Myck (CenEA), and Monika Oczkowska (CenEA)

We study health related, social and material consequences related to the loss of a partner. The panel data for the current study is based on „Survey of Health, Ageing and Retirement in Europe (SHARE).“  Causal inferences from observational data concerning the impact of a given treatment are often tricky since the assignment of treatments is typically not random. Matching methods attempt to approximate randomized controlled trials by specifying an “appropriate” control and treatment group. The main goal is to reduce imbalance and, as result, to reduce model dependence in the statistical analysis. However, given the multitude of available matching methods and other choices the researcher must do, the likelihood of bias is not necessarily reduced. In an attempt to understand the consequences of a partner loss on wellbeing we apply the following matching strategy: first we conduct exact matching on broad categories (gender, age and education category, country) and then propensity score matching within each strata. Those who lose their partner show more frequently symptoms of depression and a decrease in life quality. Especially females see a reduction in their material wellbeing.

Literature:     

RH Dehejia, S Wahba, Propensity score-matching methods for nonexperimental causal studies,  Review of Economics and statistics, 2002 - MIT Press 

M Caliendo, S Kopeinig, Some practical guidance for the implementation of propensity score matching,  Journal of economic surveys, 2008  

Gary King and Richard Nielsen. 2016. “Why Propensity Scores Should Not Be Used for Matching” 

Guo, Shenyang und Fraser, Mark W, Propensity score analysis: statistical methods and applications, Sage 2014