Germany, the German results confirm the predictions. The U.S. results are rather
mixed, the occupation retired dummy variable has a negative and insignificant
coefficient and the age effect alone is only from age 54 on compatible with the
predictions.
The gender dummy variable has the expected sign for Germans: women invest 4
percentage points less into risky assets after controlling for other effects, indicating
more risk-averse behavior. The U.S. results are surprising: women invest 5
percentage points more in risky assets. Thus, our analysis to this respect contradicts
previous empirical evidence and also stands in contrast to the predictions from the
normative model. Similarly, the German education coefficients show the expected
sign, but the opposite is true for the U.S. education coefficients.
The combined effect of labor income and cash on hand is clearly diametrical to the
normative model for U.S. individuals. For any combinations of labor income and net
worth, the risky asset share is a negative function in the labor income-to-net worth
ratio (compare Figure 1). For German investors, the results are mixed. While varying
the labor income-to-net worth ratio (and consequently labor income or/and net
worth at the same time), the resulting function of the risky asset share is u-shaped;
thus, only after reaching the minimum did it match the predictions of the benchmark
model. Because both the coefficients for the logs of labor income and net worth are
unequal to zero, the location of the minimum depends on the absolute values of the
two variables and not only on their ratio. In general, we need ratios of labor income
to net worth of a magnitude de of 200 to reach the minimum (from the left). Thus the
predictions of the benchmark model hold in tendency rather for very small net worth
holdings.
The hypothesis that U.S. investors should invest a higher share into risky assets than
Germans (due to the for CRRA utility better risk-return trade-off of the risky asset)
does not hold. The coefficients for high education, log(net worth), and
labor Income/net worth allow for many combinations of the explanatory variables
that result in a lower risky asset share for U.S. investors.
21