Abstract
The purpose of this paper is to investigate whether the choice of household informant for psychological variables included in models of risky household financial behavior matters to the empirical researcher. Five research hypotheses are posited in relation to this purpose, which concentrate on evaluating results from different correlation and regression analyses based on behavior measured at the household level, but with psychological data drawn from either the family financial officer (FFO) or the spouse in family households (N = 807). A sample of one-person households from the same database was used as control group (N = 211). It could not be shown directly that the amount of explained variance differed significantly between multiple regression analyses, in which the psychological data were drawn from different informants. However, other tests and analyses strongly indicate that including FFO data increased the validity of the model, while the inclusion of spouse data gave a marginally positive, albeit statistically significant, effect. The interpretation of the model also differs when different informants' data are used. One-person household data used to estimate an identical model seemed to produce a better fit than family household data. Finally, measures of "couple" variables showed stronger agreement between spouses than "individual" variables. Zero-order correlations between psychological variables and measures of risky financial behavior differed significantly between spouses in a few cases. The implication is that in this behavioral domain, psychological data must be collected from the family financial officer, while the spouse can be excluded without any severe consequences. This will also reduce the need to eliminate households from the analysis because of partial non-response.