At the extreme, if Joe were in the 15 percent federal tax bracket, he would pay zero percent on capital gains. In that case, Joe should harvest gains, and reset his basis, lowering future tax bills. He should never harvest losses.
The type of gains offset matters too. If Joe happened to face short-term capital gains taxes, and were in a high tax bracket, he’d save real cash by harvesting losses and avoid paying ordinary income rates, a 3.8 percent Medicare surcharge, and state taxes.
If not, Joe might have long-term gains to offset. He can also offset up to $3,000 of ordinary income. Otherwise, Joe would just carry the loss forward. It has no immediate cash value, and therefore produces no tax savings to invest.
The second-most-important variable is time, because the longer Joe can invest the tax savings, the greater the power of compounding. Both Wealthfront and Betterment use time to their advantage in their tax-loss harvesting white papers. By starting their analysis in 2000, with the tech crash, they produce most of their tax savings at the outset of their backtests, and allow it to compound through 2013.
Lastly, there’s the pattern of investment returns to consider. Tax-loss harvesting only works if you can sell investments at a loss. Investing itself works under the opposite conditions. For both to work, you need volatility and a strong stomach, and you also need your losses to precede the gains.
The takeaway here is that an awful lot has to go right for investors to realize sizable benefits from tax-loss harvesting. And some of it isn’t really predictable ahead of time.
The robo advisors acknowledge this—kind of.
Betterment publishes a “who benefits most/least” analysis, plus a substantial disclosure section in its tax-loss harvesting white paper, laying out the many variables that affect its returns. Wealthfront's is a bit coy, emphasizing cases where tax-loss harvesting works well, but burying mentions of variability in the fine print disclosure section.
Wealthfront’s tax-loss harvesting white paper shows actual client results of 0.40 percent “tax alpha” over a 13-month period from October 2012-2013, not the 1.00 percentage point that Wealthfront posts on its home page.
And still, there’s a catch: Wealthfront’s tax alpha examples ignore the eventual sale of the assets, for which Michael Kitces, the blogger behind “Nerd’s Eye View,” has taken them to task. Presumably, Kitces would do the same to FutureAdvisor, which also presents tax alphas rather than internal rates of return that capture terminal tax liabilities.
Tax alpha measures the savings from the first stages of the tax-loss harvest, but not the increased tax liabilities from the eventual sale of any lowered-basis securities. Wealthfront switches back to the primary after 30 days. Unless the secondary regained all of the primary’s losses during that 30 days, the primary’s cost basis will be lower than it was before the harvest.
Betterment’s tax-loss harvesting white paper runs a range of liquidation examples, giving more information than Wealthfront’s. However, Betterment did not supply actual client results, but instead presented a backtest.