Working Paper: NBER ID: w2845
Authors: Daniel Feenberg; Jonathan Skinner
Abstract: To address the question of whether IRA5 contribute to capital formation, we use the IRS/University of Michigan taxpayer sample for income tax returns during 1980-84. By matching families across a five-year period, we can estimate the dynamic interactions of IRA purchases and other types of saving, correct for individual differences, and test whether IRA purchases are in part offset by other (net) asset sales. The "reshuffling" hypothesis implies that taxpayers who enroll in IRAs should, over time, experience a drop in net taxable interest and dividend income as their taxable assets (or new loans) are used to purchase IRAs. Conversely, the "new saving" view of IRAs implies that taxable interest and dividend income should be unaffected by IRA purchases. We find little or no evidence which favors the view that IRAs are funded by cashing out existing taxable assets. In fact, individuals who purchased IRAs in each year between 1982-84 increased their asset holdings by more than those who did not purchase IRAs. In one sense, our results strongly confirm the studies by Venti and Wise and Hubbard that IRA saving represents new saving. But shuffling could still occur, albeit on a secondary level: families who are accumulating both taxable assets and IRAs might have accumulated even more taxable assets had IRA5 not been available
Keywords: IRA; saving; capital formation; tax incentives
JEL Codes: No JEL codes provided
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| IRA purchases (H26) | drop in net taxable interest and dividend income (H29) |
| IRA purchases (H26) | no effect on taxable income (H24) |
| taxpayers who owe money to IRS (H26) | purchase IRAs (D14) |
| IRA contributions (H20) | overall saving behavior (D14) |
| cashing out existing taxable assets (D14) | IRA contributions (H20) |
| income and wealth levels (D31) | IRA contributions (H20) |