The Analysis of Risk Attitude Amongst Family Members
Volume 5
Jena Economic Research Papers
2011-12-23
2011
2007
2011-069
Multiple Equation Models
Panel Data
Risk Attitude
The determinants of risk attitude amongst family members are explored using data from the German Socio-Economic Panel over the period 2004 to 2009. The focus of the analysis is the repeated responses to the survey question about general willingness to take risk. Responses to this question are provided on a 0-10 Likert scale. We respect both the ordinality and the panel structure of the data by estimating the random effects ordered probit model. We divide household members into thee types: heads, spouses and offspring. Of the three types, we find that spouses are the most risk averse, and offspring the least risk averse. In view of these findings, we estimate the model separately for the three groups and find different results between the three. For example, household income has a positive effect (on risk-taking) for heads and spouses (particularly strong for spouses), but no effect on offspring. Some effects are similar between the the three groups; for example, risk aversion always increases with age. In the offspring equation, we include both the head’s and the spouse’s risk attitude as explanatory variables, and find that both have a significantly positive effect on the offspring’s risk-attitude, indicating that children tend to inherit the risk attitude of their guardians. We then focus on couples in the data set, and we apply the random effects bivariate ordered probit model to the analysis of the simultaneous determination of the male’s and the female’s risk attitude. In this model, the individual-specific effects for the male and the female are assumed to have a non-zero correlation, which is estimated to be +0.412. This significantly positive correlation is interpreted as a form of homophily: individuals tend to form partnerships with others having a similar risk attitude. The importance of respecting the ordinality of the data is confirmed when a straightforward (linear) seemingly unrelated model is applied to the same problem; this gives a correlation of only +0.27: a seriously downward-biased estimate of this key parameter.
jportal_jpjournal_00000016
2012-01-16T14:23:50.073Z
2012-01-16T14:23:14.614Z