Finding the Patterns of IRA Investment Decision Among Europeans Aged 50+ with Formal Education and Primary Residence Before the Fall of Communism. Evidence from SHARE-ERIC (Wave 7)
23 Pages Posted: 23 Mar 2021
Date Written: March 22, 2021
This paper deals with the analysis of determinants of the decision of investing in Individual Retirement Accounts (IRA) among Europeans aged 50 and over, starting from the SHARE-ERIC data set (Wave 7), filtered on dominant residences and full-time education before the fall of the Iron Curtain in Europe. Using more than 33.000 records, it validates the assumption that schooling and living place in former communist countries count for such financial behavior. Further, it brings two particular models with good accuracy of classification starting from the latter criterion. We applied many methods and techniques based on data mining and variable selection tools, probit and binary logistic regression analysis with average marginal effects, automatic cross-validations, mixed-effects modeling with random effects on countries, and prediction nomograms. We found that some influences from the same financial category as the dependent variable, such as having life insurance or ever investing in mutual funds or stocks count the most when dealing with investing in IRA. More, the younger respondents, those with computer skills required by their jobs, and those who have gone through at least a period of high stress were more likely to invest using such an instrument, no matter their residence and education. Besides, more educated people, those who suffered in their past by some form of discrimination, the relaxed, happier, more independent, and more interpersonal confident individuals were more prone to choose to invest this way. Moreover, this paper also confirms some country-level influences related to stock market capitalization to GDP ratio and Worldwide Governance Indicators.
Keywords: Retirement Income, Communist Country, Data Mining and Variable Selection Techniques, Naive Bayes and LASSO, Logistic and Probit Models, Marginal and Mixed-effects, Prediction Capability.
JEL Classification: C58, G17, O57
Suggested Citation: Suggested Citation