But, as section 2.2.1 pointed out, differences in investment behavior can be a
completely rational reaction to differences in preferences or endowments. For
example, a highly risk-averse individual may be completely happy with low expected
final wealth, if the volatility of the wealth distribution is also sufficiently low.
Thus, a comparison of investment strategies should incorporate the rational
contribution of preferences and endowments. For this, the literature uses the concept
of lifetime utility. The actual behavior (in this study, the behavior is asset allocation)
thus is compared with some benchmark behavior derived using the normative
lifetime utility model.
Research on comparisons of investment strategies and behaviors (including this
contribution) measures utility costs—that is, losses in welfare—associated with
deviations from the normatively determined optimal behavior. Among the important
works are Dammon, Spatt, and Zhang (2004); Cocco, Gomes, and Maenhout (2005;,
and Yao and Zhang (2005). Dammon, Spatt, and Zhang (2004) compare the expected
lifetime utility of commonly observed investment choices with optimal choices. They
undertake these measurements with respect to taxable and tax-deferred accounts.
Cocco, Gomes, and Maenhout (2005) compare common investment advisers’
recommendations with the optimal choices and furthermore evaluate the effect of
ignoring labor income or labor income risk while deriving the optimal choice. The
welfare costs of not following an optimal renting versus owning a house strategy are
calculated by Yao and Zhang (2005).9
3 Optimizing Asset Allocations and Wealth over the Life-Cycle - The Normative
Benchmark Model
In this section, the normative benchmark model is defined and calibrated with
empirical data. We derive the optimal asset allocation—depending on the
9 Poterba et al. (2006) also use a utility-based benchmark for investigating the performance
of DB and DC plans, but restrict the measurement of utility to the wealth distribution at
one single point in time (age 63), thus abstracting from life-cycle effects.