Table 2: 2004 Survey of Consumer Finances (SCF) and 2003 Income and
Expenditure Survey (EVS) (for definition of variables, see Appendix B)
--- put Table 2 here ---
The descriptive statistics in Table 2 include two variables indicating the amount of
risky assets. The first, risky 1, is a more traditional measure (e.g., as in Bertaut and
Starr-McCluer, 2002, or Eymann and Borsch-Supan, 2002) including stocks and
similar investments (see Appendix B). The second, more broadly defined measure,
risky 2, includes the value of real estate and, as negative investments into risky
assets, debt (assuming that debt is risky). Risky 2 will be used in our regressions
later, and thus we concentrate on this measure in the following discussion.
Comparing the United States and Germany, we observe that Americans invest a
higher share into the risky asset (58% vs. 39%). But we also see that the variables
referring to financial wealth and income vary to a great extent in both countries, and,
thus, as the benchmark model revealed, a regression model should control for that.
5 Econometric Analyses
5.1 Hypotheses about the Investment Process and Implications for Regression
Models
In the literature (e.g., Bertaut and Starr-McCluer, 2002; Eymann and Borsch-Supan,
2002), two general hypotheses about the investment process are discussed. The first
assumes that the choice to invest into risky assets is made simultaneously with the
decision on the share of wealth invested. The second hypothesis assumes a two-stage
investment process. At the first stage, the individual decides whether to invest in the
risky asset; at the second stage, the share is independently derived. The rationale for
this two-stage process is that, before investing in risky assets, costly information
must be obtained (e.g., college education); if these costs are prohibitively high, no
investment is made at all.
18