Fig. 1. Summary of the marginal tax rate patterns in 1983, 1989, and 1998.

Table 1. Probability of ownership (percent) of each of eight asset classes: 1983–1998

Source: Authors’ tabulations based on data in Surveys of Consumer Finances. Households are weighted by sample weights in each year.

Table 2. Conditional ownership probabilities (percent) for eight asset classes: 1998

Notes: Each entry indicates the probability that a household owns the asset class indicated at the column head, conditional on owning the asset class indicated a the beginning of the row. Entries are based on authors’ tabulations using the 1998 Survey of Consumer Finances, weighting households by their sampling weights.

Table 3. Average portfolio shares (percent): 1983–1998

Notes: Unconditional averages refer to all households, while conditional averages refer to the households with positive holdings of the indicated asset class. Tabulations are based on authors’ calculations using data from the Survey of Consumer Finances, weighting each household by its sampling weight.

Table 4. Aggregate portfolio shares (percent): 1983–1998

Source: Authors’ tabulations using various years of the Survey of Consumer Finances. Households are weighted by sample weights in each year.

Table 5. Summary statistics on Survey of Consumer Finances sample: 1983–1998

Notes: Authors’ tabulations based on Surveys of Consumer Finances, weighting each household by its sampling weight. Demographic characteristics pertain to the head of the household.

Table 6. Probit estimates for financial asset ownership: 1998

Source: Authors’ estimates based on 1998 Survey of Consumer of Finances data. See text for further discussion.

Table 7. Marginal effects of changes in marginal tax rate on probability of asset ownership probit models: 1983–1998

Notes: The top panel shows the marginal effect of a unit increase in the marginal tax rate on the expected probability of ownership. The lower panel divides 0.10* (top panel estimate) by the probability that a household owns the asset class, as shown in Table 1. Asterisks indicate statistical significance at the 5% (**) and 10% (*) significance levels. See text for further discussion.

Table 8. Estimated correlation matrices from bivariate probits: 1998

Note: The entry for each cell is the estimated correlation parameter from a bivariate probit estimated on the pair of asset classes listed in the row and column headers. See text for further discussion.

Table 9. Tobit estimates of financial asset shares: 1998 (Adding up constraints imposed on marginal effects)

Source: Authors’ estimates based on 1998 SCF. Each tobit model allows for censoring at portfolio shares of zero and one. See text for further discussion.

Table 10. Marginal effects of changes in marginal tax rate on portfolio shares Tobit models 1983–1998

Notes: The top panel shows the marginal effect of a unit increase in the marginal tax rate on the expected portfolio share. The middle panel divides 0.10* (top panel estimate) by the unconditional average portfolio share reported in Table 3. Asterisks indicate statistical significance at the 5% (**) and 10% (*) significance levels. The bottom panel presents the P-values from the test of the joint significance of the tax variables in the eight equations and Lagrange multipliers on each of the adding up constraints for the explanatory variables. See text for further discussion.

Table 11. Tests of robustness of Tobit estimates of financial asset shares: 1998

Notes: Authors’ estimates based on data from 1998 Survey of Consumer Finances. The top panel shows the marginal effect of a unit increase in the marginal tax rate on the expected portfolio share. The first column repeats the results from the first column of Table 10. The specification in the second column expands the set of income categories from 7 to 20. The third column replaces the expanded set of dummy variables with a piecewise linear spline through the same 20 categories. The last three columns pertain to a specification in which the marginal tax rate is interacted with dummies for whether the household rents and for whether the household is a homeowner. The columns report the marginal effects of the two marginal tax rate interactions and the dummy for whether the household is a homeowner. Asterisks indicate statistical significance at the 5% (**) and 10% (*) significance levels. The bottom panel presents the P-values from the test of the joint significance of the tax variables in the eight equations and Lagrange multipliers on each of the adding up constraints for the explanatory variables. See text for further discussion.

Table 12. Estimated correlation matrices of bivariate tobits, SCF 1998

Note: The entry for each cell is the value of the correlation parameter that maximizes the likelihood function for a bivariate tobit for the assets listed in the corresponding row and column of the table when the coefficients are fixed at their values in Table 9.