Figure 9: Welfare Losses, ∆W0 / W0, for German (EVS) Data; Gender = 0
(male), γ = 2, δ = 0.97, Age = 50, Education = Low, Middle, or High
--- put Figure 9 here ---
Although the regression coefficients for the low and high education dummy variables
show the expected sign, welfare losses are the higher the lower education is. Higher
losses for those with lower education are driven by their larger savings (due to the
lower expected income stream for a given current income) overcompensating the
correct sign of the regression coefficient. The resulting larger basis, on which the
deviation to the optimal risky asset share works, leads to larger welfare losses.
The final variation in age, depicted in Figure 10, shows similar results as for the
United States, except for individuals age 65.
Figure 10: Welfare Losses, ∆W0 / W0, for German (EVS) Data, Gender = 0
(male), γ = 2, δ = 0.97, age = 30, 50, or 65, Education = Middle
--- put Figure 10 here ---
For individuals age 65, the welfare losses are often higher than they are for younger
individuals. Here, opposite the U.S. results, the steeply falling empirical age-risky-
asset allocation profile leads to larger losses. The empirical risky asset share falls
much faster in age than required by the benchmark model.
6.3 Comparison of the United States and Germany and Derivation of General
Results
Comparing Figures 3 to 7 with Figures 8 to 10, we observe that, for most
combinations of parameters, Americans experience larger welfare losses than
Germans due to suboptimal asset allocation. Exceptions can be found in some but not
all parameter combinations for women, those with lower education, and individuals
age 65. The major explanation for this finding is the larger gap between the U.S.
29