Since then I’ve started several threads on the forum on the topic of Roth conversions, and published a paper in the J

*ournal of Financial Planning*on the arithmetic of Roth conversions.

That journal has just published my latest effort, which I don’t believe is as yet behind a paywall: https://www.financialplanningassociatio ... sions-OPEN.

The title is self-explanatory:

**Net Present Value Analysis of Roth Conversions**.

Here is the Executive Summary

• In most conventional treatments, the advisability of a Roth conversion hinges on a comparison of present to future tax rates. If future tax rates will be higher, conversion is indicated.

• This paper argues that future tax rates are unknowable, and that this uncertainty needs to be incorporated into the decision process around Roth conversions.

• The time course over which tax savings accrue has also been ignored. For the mass affluent, conversions typically serve to reduce required minimum distributions. Because RMDs continue for life, tax savings from reducing future RMDs may take decades to accumulate.

• RMD-reducing Roth conversions fit the classic model for net present value analyses. There is a large one-time tax payment at conversion, followed by years and years of tax savings. Savings postponed until the distant future must be discounted. The conversion pays off only if the discounted value of future savings exceeds the cost to convert.

• This paper develops the implications of the long, slow, uncertain payoff for conversions. Clients need to be comfortable with a payoff that may not be complete until 20 or 30 years have passed, and which may be comparatively small in present value terms.

Here is a key passage that will convey the implications of the NPV analysis:

**Discounting Future Tax Savings**

The proper discount rate must be the expected annualized rate of appreciation on the portfolio targeted for conversion. That is the rate that makes the client indifferent to paying tax now versus later (assuming constant tax rates throughout the distribution period). Here is an example: Suppose a $10,000 portfolio in a tax-deferred account is invested in stocks returning 10 percent and that the tax rate on withdrawals is 25 percent. If the account is converted today, tax of $2,500 will be due. If the account is liquidated after one year of appreciation, $11,000 will be withdrawn, and the tax due will be $2,750. Discounted at the portfolio rate of appreciation, paying $2,750 after one year is no different from paying $2,500 today.

If any lower rate of discount were to be applied to future tax savings (whether inflation, the risk-free rate, or no discount at all), then Roth conversions would always pay off spectacularly if held for long enough. Left in stocks for 30 years, that $10,000 portfolio has an expected future value of about $175,000; undiscounted, the future tax savings would be about $43,625. Absent any discounting, a rational taxpayer would always convert immediately, paying only $2,500 in tax today to save tens of thousands of dollars in future tax.

Clients in the top bracket are particularly vulnerable to this form of money illusion. In the top bracket, clients can convert an arbitrarily large amount at a single tax rate, say, $10 million at 37 percent, paying $3.70 million in tax today. Again, invested in stocks at 10 percent for 30 years, that conversion avoids a future distribution of $175 million, thus averting tax of $64.75 million. Who could resist saving over $60 million in tax through an immediate conversion?

By contrast, the rational taxpayer not subject to money illusion will discount tax savings not received until some future date, and will discount more heavily the more distant that future.

Last, I append a sample of some of the recommendations, all of which are presented as rules of thumb.

Here is a set of circumstances where a Roth conversion becomes a high-risk, low-payoff bet.

1. In general, the earlier the conversion, the more risky the bet; the future is necessarily uncertain, and the more distant the future the more so. Too much can change, legislatively or personally. No one who converted at the first opportunity in 1998, after the creation of Roth accounts, at the then 28 percent rate, had any reason to expect that income in that bracket would see a rate reduction first to 25 percent under EGTRRA and then to 22 percent under TCJA.

2. The higher the tax rate paid on conversion, the bigger the loan made to the government and the greater the vulnerability to a change in personal circumstances or in the legislative environment.

3. Conversions that represent a pure bet on future legislation—as in the baseline example of converting at 22 percent to avoid tax at 25 percent post-TCJA—typically have small payoffs, high uncertainty, and considerable downside risk. The taxpayer has to get both future congressional action and their own future tax situation correct, to win a modest payoff; if they get both wrong, the cost may be steep. Interestingly, converting at 24 percent to avert 28 percent post-TCJA is a somewhat safer bet. If this client gets both predictions wrong, they paid 24 percent to avert tax of 22 percent, a smaller downside. [

*assumes no IRMAA*]

4. Conversions intended to reduce RMDs and deliver a payoff during the client’s life are almost certain to be disappointing unless the conversion tax rate is in single digits and the tax gap is a multiple of it and in double digits. The base case in this paper violated both conditions. The middle bracket converter has to live long enough, with no untoward personal events or unexpected legislation, for a payoff to be received while alive. Most of the time, these conversions will represent a loan to the government not fully paid off even at the taxpayer’s demise, with any payoff from the conversion going to the heirs.

In short: the new paper reverts to a more jaundiced view of the prospects for a Roth conversion in the 22% or 24% bracket, as in the initial working paper.